2017-04-09 13:35:01 +00:00
--[[--
This module manages widgets .
] ]
2014-10-30 18:42:18 +00:00
local Device = require ( " device " )
2013-10-18 20:38:07 +00:00
local Event = require ( " ui/event " )
2014-11-28 20:12:54 +00:00
local Geom = require ( " ui/geometry " )
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
local TimeVal = require ( " ui/timeval " )
2016-03-29 06:37:15 +00:00
local dbg = require ( " dbg " )
2016-12-29 08:10:38 +00:00
local logger = require ( " logger " )
Revamp "flash_ui" handling (#7118)
* Wherever possible, do an actual dumb invert on the Screen BB instead of repainting the widget, *then* inverting it (which is what the "invert" flag does).
* Instead of playing with nextTick/tickAfterNext delays, explicitly fence stuff with forceRePaint
* And, in the few cases where said Mk. 7 quirk kicks in, make the fences more marked by using a well-placed WAIT_FOR_UPDATE_COMPLETE
* Fix an issue in Button where show/hide & enable/disable where actually all toggles, which meant that duplicate calls or timing issues would do the wrong thing. (This broke dimming some icons, and mistakenly dropped the background from FM chevrons, for example).
* Speaking of, fix Button's hide/show to actually restore the background properly (there was a stupid typo in the variable name)
* Still in Button, fix the insanity of the double repaint on rounded buttons. Turns out it made sense, after all (and was related to said missing background, and bad interaction with invert & text with no background).
* KeyValuePage suffered from a similar issue with broken highlights (all black) because of missing backgrounds.
* In ConfigDialog, only instanciate IconButtons once (because every tab switch causes a full instantiation; and the initial display implies a full instanciation and an initial tab switch). Otherwise, both instances linger, and catch taps, and as such, double highlights.
* ConfigDialog: Restore the "don't repaint ReaderUI" when switching between similarly sized tabs (re #6131). I never could reproduce that on eInk, and I can't now on the emulator, so I'm assuming @poire-z fixed it during the swap to SVG icons.
* KeyValuePage: Only instanciate Buttons once (again, this is a widget that goes through a full init every page). Again, caused highlight/dimming issues because buttons were stacked.
* Menu: Ditto.
* TouchMenu: Now home of the gnarliest unhilight heuristics, because of the sheer amount of different things that can happen (and/or thanks to stuff not flagged covers_fullscreen properly ;p).
* Bump base
https://github.com/koreader/koreader-base/pull/1280
https://github.com/koreader/koreader-base/pull/1282
https://github.com/koreader/koreader-base/pull/1283
https://github.com/koreader/koreader-base/pull/1284
* Bump android-luajit-launcher
https://github.com/koreader/android-luajit-launcher/pull/284
https://github.com/koreader/android-luajit-launcher/pull/285
https://github.com/koreader/android-luajit-launcher/pull/286
https://github.com/koreader/android-luajit-launcher/pull/287
2021-01-10 00:51:09 +00:00
local ffiUtil = require ( " ffi/util " )
local util = require ( " util " )
2017-06-23 17:04:11 +00:00
local _ = require ( " gettext " )
2017-04-09 13:35:01 +00:00
local Input = Device.input
local Screen = Device.screen
2012-04-22 19:29:48 +00:00
2019-11-26 14:03:05 +00:00
local DEFAULT_FULL_REFRESH_COUNT = 6
2015-12-26 00:48:40 +00:00
2012-04-22 19:29:48 +00:00
-- there is only one instance of this
2013-10-18 20:38:07 +00:00
local UIManager = {
2014-03-13 13:52:43 +00:00
-- trigger a full refresh when counter reaches FULL_REFRESH_COUNT
2016-02-25 08:54:41 +00:00
FULL_REFRESH_COUNT =
2020-07-17 06:54:21 +00:00
G_reader_settings : isTrue ( " night_mode " ) and G_reader_settings : readSetting ( " night_full_refresh_count " ) or G_reader_settings : readSetting ( " full_refresh_count " ) or DEFAULT_FULL_REFRESH_COUNT ,
2014-03-13 13:52:43 +00:00
refresh_count = 0 ,
2021-05-19 20:57:54 +00:00
currently_scrolling = false ,
2013-03-12 17:18:53 +00:00
2018-01-01 14:40:28 +00:00
-- How long to wait between ZMQ wakeups: 50ms.
ZMQ_TIMEOUT = 50 * 1000 ,
2014-06-26 17:37:33 +00:00
event_handlers = nil ,
2014-03-13 13:52:43 +00:00
_running = true ,
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
_now = TimeVal : now ( ) ,
2014-03-13 13:52:43 +00:00
_window_stack = { } ,
2015-12-26 00:48:40 +00:00
_task_queue = { } ,
_task_queue_dirty = false ,
2014-06-23 14:33:11 +00:00
_dirty = { } ,
_zeromqs = { } ,
2014-11-28 20:12:54 +00:00
_refresh_stack = { } ,
_refresh_func_stack = { } ,
2016-11-16 10:49:34 +00:00
_entered_poweroff_stage = false ,
2017-05-16 09:11:11 +00:00
_exit_code = nil ,
2020-08-30 13:01:56 +00:00
_prevent_standby_count = 0 ,
_prev_prevent_standby_count = 0 ,
2021-01-01 13:34:55 +00:00
_discard_events_till = nil ,
2017-06-23 17:04:11 +00:00
event_hook = require ( " ui/hook_container " ) : new ( )
2012-04-22 19:29:48 +00:00
}
2014-06-26 17:37:33 +00:00
function UIManager : init ( )
self.event_handlers = {
__default__ = function ( input_event )
self : sendEvent ( input_event )
end ,
SaveState = function ( )
2017-05-16 09:11:11 +00:00
self : flushSettings ( )
2014-08-26 10:10:26 +00:00
end ,
Power = function ( input_event )
Device : onPowerEvent ( input_event )
end ,
2014-06-26 17:37:33 +00:00
}
2016-11-16 10:49:34 +00:00
self.poweroff_action = function ( )
2021-01-12 21:31:38 +00:00
self._entered_poweroff_stage = true
Device.orig_rotation_mode = Device.screen : getRotationMode ( )
2020-07-09 17:11:44 +00:00
Screen : setRotationMode ( Screen.ORIENTATION_PORTRAIT )
2021-03-06 21:44:18 +00:00
local Screensaver = require ( " ui/screensaver " )
Screensaver : setup ( " poweroff " , _ ( " Powered off " ) )
2021-04-17 17:30:23 +00:00
if Device : hasEinkScreen ( ) and Screensaver : modeIsImage ( ) then
if Screensaver : withBackground ( ) then
Screen : clear ( )
end
Screen : refreshFull ( )
end
2021-03-06 21:44:18 +00:00
Screensaver : show ( )
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
if Device : needsScreenRefreshAfterResume ( ) then
Screen : refreshFull ( )
end
2016-11-16 10:49:34 +00:00
UIManager : nextTick ( function ( )
2017-09-04 19:05:05 +00:00
Device : saveSettings ( )
2020-11-14 22:39:19 +00:00
if Device : isKobo ( ) then
self._exit_code = 88
end
2016-11-16 10:49:34 +00:00
self : broadcastEvent ( Event : new ( " Close " ) )
Device : powerOff ( )
end )
end
2017-05-14 16:43:08 +00:00
self.reboot_action = function ( )
2021-01-12 21:31:38 +00:00
self._entered_poweroff_stage = true
Device.orig_rotation_mode = Device.screen : getRotationMode ( )
2020-07-09 17:11:44 +00:00
Screen : setRotationMode ( Screen.ORIENTATION_PORTRAIT )
2021-03-06 21:44:18 +00:00
local Screensaver = require ( " ui/screensaver " )
Screensaver : setup ( " reboot " , _ ( " Rebooting… " ) )
2021-04-17 17:30:23 +00:00
if Device : hasEinkScreen ( ) and Screensaver : modeIsImage ( ) then
if Screensaver : withBackground ( ) then
Screen : clear ( )
end
Screen : refreshFull ( )
end
2021-03-06 21:44:18 +00:00
Screensaver : show ( )
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
if Device : needsScreenRefreshAfterResume ( ) then
Screen : refreshFull ( )
end
2017-05-14 16:43:08 +00:00
UIManager : nextTick ( function ( )
2017-09-04 19:05:05 +00:00
Device : saveSettings ( )
2020-11-14 22:39:19 +00:00
if Device : isKobo ( ) then
self._exit_code = 88
end
2017-05-14 16:43:08 +00:00
self : broadcastEvent ( Event : new ( " Close " ) )
Device : reboot ( )
end )
end
2018-01-06 12:14:26 +00:00
if Device : isPocketBook ( ) then
2020-08-30 13:01:56 +00:00
-- Only fg/bg state plugin notifiers, not real power event.
2018-01-06 12:14:26 +00:00
self.event_handlers [ " Suspend " ] = function ( )
self : _beforeSuspend ( )
end
self.event_handlers [ " Resume " ] = function ( )
self : _afterResume ( )
end
end
2014-06-26 17:37:33 +00:00
if Device : isKobo ( ) then
2016-03-18 23:33:14 +00:00
-- We do not want auto suspend procedure to waste battery during
-- suspend. So let's unschedule it when suspending, and restart it after
2019-03-26 21:28:27 +00:00
-- resume. Done via the plugin's onSuspend/onResume handlers.
2016-05-02 07:51:58 +00:00
self.event_handlers [ " Suspend " ] = function ( )
2019-03-26 21:28:27 +00:00
-- Ignore the accelerometer (if that's not already the case) while we're alseep
if G_reader_settings : nilOrFalse ( " input_ignore_gsensor " ) then
2019-07-01 15:12:24 +00:00
Device : toggleGSensor ( false )
2019-03-26 21:28:27 +00:00
end
2017-05-16 09:11:11 +00:00
self : _beforeSuspend ( )
2016-05-02 07:51:58 +00:00
Device : onPowerEvent ( " Suspend " )
2014-08-26 10:10:26 +00:00
end
2016-05-02 07:51:58 +00:00
self.event_handlers [ " Resume " ] = function ( )
Device : onPowerEvent ( " Resume " )
2017-05-16 09:11:11 +00:00
self : _afterResume ( )
2019-03-26 21:28:27 +00:00
-- Stop ignoring the accelerometer (unless requested) when we wakeup
if G_reader_settings : nilOrFalse ( " input_ignore_gsensor " ) then
2019-07-01 15:12:24 +00:00
Device : toggleGSensor ( true )
2019-03-26 21:28:27 +00:00
end
2014-06-26 17:37:33 +00:00
end
2016-05-02 07:51:58 +00:00
self.event_handlers [ " PowerPress " ] = function ( )
2019-08-29 20:02:08 +00:00
-- Always schedule power off.
-- Press the power button for 2+ seconds to shutdown directly from suspend.
2016-11-16 10:49:34 +00:00
UIManager : scheduleIn ( 2 , self.poweroff_action )
2016-04-26 22:30:52 +00:00
end
2016-05-02 07:51:58 +00:00
self.event_handlers [ " PowerRelease " ] = function ( )
2016-11-16 10:49:34 +00:00
if not self._entered_poweroff_stage then
UIManager : unschedule ( self.poweroff_action )
2019-08-29 20:02:08 +00:00
-- resume if we were suspended
if Device.screen_saver_mode then
self : resume ( )
else
self : suspend ( )
end
2016-04-26 22:30:52 +00:00
end
end
2019-06-28 20:19:25 +00:00
-- Sleep Cover handling
2021-03-06 21:44:18 +00:00
if G_reader_settings : isTrue ( " ignore_power_sleepcover " ) then
2019-06-28 20:19:25 +00:00
-- NOTE: The hardware event itself will wake the kernel up if it's in suspend (:/).
-- Let the unexpected wakeup guard handle that.
self.event_handlers [ " SleepCoverClosed " ] = nil
self.event_handlers [ " SleepCoverOpened " ] = nil
2021-03-06 21:44:18 +00:00
elseif G_reader_settings : isTrue ( " ignore_open_sleepcover " ) then
2019-06-28 20:19:25 +00:00
-- Just ignore wakeup events, and do NOT set is_cover_closed,
-- so device/generic/device will let us use the power button to wake ;).
2016-09-14 05:16:31 +00:00
self.event_handlers [ " SleepCoverClosed " ] = function ( )
2017-05-16 09:11:11 +00:00
self : suspend ( )
2016-09-14 05:16:31 +00:00
end
self.event_handlers [ " SleepCoverOpened " ] = function ( )
Device.is_cover_closed = false
end
2016-09-12 04:38:06 +00:00
else
self.event_handlers [ " SleepCoverClosed " ] = function ( )
2019-06-28 20:19:25 +00:00
Device.is_cover_closed = true
self : suspend ( )
end
self.event_handlers [ " SleepCoverOpened " ] = function ( )
Device.is_cover_closed = false
self : resume ( )
2016-09-12 04:38:06 +00:00
end
end
2014-06-26 17:37:33 +00:00
self.event_handlers [ " Light " ] = function ( )
Device : getPowerDevice ( ) : toggleFrontlight ( )
end
2016-10-22 07:04:37 +00:00
self.event_handlers [ " Charging " ] = function ( )
2020-09-11 21:23:57 +00:00
self : _beforeCharging ( )
-- NOTE: Plug/unplug events will wake the device up, which is why we put it back to sleep.
if Device.screen_saver_mode then
self : suspend ( )
end
end
self.event_handlers [ " NotCharging " ] = function ( )
-- We need to put the device into suspension, other things need to be done before it.
self : _afterNotCharging ( )
if Device.screen_saver_mode then
self : suspend ( )
end
end
self.event_handlers [ " UsbPlugIn " ] = function ( )
2017-05-16 09:11:11 +00:00
self : _beforeCharging ( )
2020-08-25 00:00:59 +00:00
-- NOTE: Plug/unplug events will wake the device up, which is why we put it back to sleep.
2017-03-24 06:36:15 +00:00
if Device.screen_saver_mode then
2017-05-16 09:11:11 +00:00
self : suspend ( )
2020-08-25 00:00:59 +00:00
else
-- Potentially start an USBMS session
local MassStorage = require ( " ui/elements/mass_storage " )
2020-08-26 23:06:41 +00:00
MassStorage : start ( )
2017-03-24 06:36:15 +00:00
end
end
2020-09-11 21:23:57 +00:00
self.event_handlers [ " UsbPlugOut " ] = function ( )
2017-05-16 09:11:11 +00:00
-- We need to put the device into suspension, other things need to be done before it.
self : _afterNotCharging ( )
2016-09-20 07:14:14 +00:00
if Device.screen_saver_mode then
2017-05-16 09:11:11 +00:00
self : suspend ( )
2021-08-24 19:03:40 +00:00
else
-- Potentially dismiss the USBMS ConfirmBox
local MassStorage = require ( " ui/elements/mass_storage " )
MassStorage : dismiss ( )
2016-09-20 07:14:14 +00:00
end
end
2014-06-26 18:28:57 +00:00
self.event_handlers [ " __default__ " ] = function ( input_event )
2018-03-28 10:57:55 +00:00
-- Suspension in Kobo can be interrupted by screen updates. We ignore user touch input
-- in screen_saver_mode so screen updates won't be triggered in suspend mode.
-- We should not call self:suspend() in screen_saver_mode lest we stay on forever
-- trying to reschedule suspend. Other systems take care of unintended wake-up.
if not Device.screen_saver_mode then
2014-06-26 18:28:57 +00:00
self : sendEvent ( input_event )
end
end
2014-06-26 17:37:33 +00:00
elseif Device : isKindle ( ) then
self.event_handlers [ " IntoSS " ] = function ( )
2017-05-16 09:11:11 +00:00
self : _beforeSuspend ( )
2014-06-26 17:37:33 +00:00
Device : intoScreenSaver ( )
end
self.event_handlers [ " OutOfSS " ] = function ( )
Device : outofScreenSaver ( )
2017-05-16 09:11:11 +00:00
self : _afterResume ( ) ;
2014-06-26 17:37:33 +00:00
end
self.event_handlers [ " Charging " ] = function ( )
2017-05-16 09:11:11 +00:00
self : _beforeCharging ( )
2014-06-26 17:37:33 +00:00
Device : usbPlugIn ( )
end
self.event_handlers [ " NotCharging " ] = function ( )
Device : usbPlugOut ( )
2017-05-16 09:11:11 +00:00
self : _afterNotCharging ( )
2014-06-21 16:24:37 +00:00
end
2020-02-08 00:58:10 +00:00
elseif Device : isRemarkable ( ) then
2020-10-09 17:23:57 +00:00
self.event_handlers [ " Suspend " ] = function ( )
self : _beforeSuspend ( )
Device : onPowerEvent ( " Suspend " )
end
self.event_handlers [ " Resume " ] = function ( )
Device : onPowerEvent ( " Resume " )
self : _afterResume ( )
end
2020-02-08 00:58:10 +00:00
self.event_handlers [ " PowerPress " ] = function ( )
UIManager : scheduleIn ( 2 , self.poweroff_action )
end
self.event_handlers [ " PowerRelease " ] = function ( )
if not self._entered_poweroff_stage then
UIManager : unschedule ( self.poweroff_action )
-- resume if we were suspended
if Device.screen_saver_mode then
self : resume ( )
else
self : suspend ( )
end
end
end
self.event_handlers [ " __default__ " ] = function ( input_event )
-- Same as in Kobo: we want to ignore keys during suspension
if not Device.screen_saver_mode then
self : sendEvent ( input_event )
end
end
2018-09-07 23:37:04 +00:00
elseif Device : isSonyPRSTUX ( ) then
self.event_handlers [ " PowerPress " ] = function ( )
UIManager : scheduleIn ( 2 , self.poweroff_action )
end
self.event_handlers [ " PowerRelease " ] = function ( )
if not self._entered_poweroff_stage then
UIManager : unschedule ( self.poweroff_action )
2019-09-01 19:37:08 +00:00
-- resume if we were suspended
if Device.screen_saver_mode then
self : resume ( )
else
self : suspend ( )
end
2018-09-07 23:37:04 +00:00
end
end
2019-09-01 19:37:08 +00:00
self.event_handlers [ " Suspend " ] = function ( )
self : _beforeSuspend ( )
Device : intoScreenSaver ( )
Device : suspend ( )
end
2018-09-07 23:37:04 +00:00
self.event_handlers [ " Resume " ] = function ( )
Device : resume ( )
Device : outofScreenSaver ( )
self : _afterResume ( )
end
2019-02-08 22:14:23 +00:00
self.event_handlers [ " Charging " ] = function ( )
self : _beforeCharging ( )
end
self.event_handlers [ " NotCharging " ] = function ( )
self : _afterNotCharging ( )
end
self.event_handlers [ " UsbPlugIn " ] = function ( )
if Device.screen_saver_mode then
Device : resume ( )
Device : outofScreenSaver ( )
self : _afterResume ( )
end
Device : usbPlugIn ( )
end
self.event_handlers [ " UsbPlugOut " ] = function ( )
Device : usbPlugOut ( )
end
2018-09-07 23:37:04 +00:00
self.event_handlers [ " __default__ " ] = function ( input_event )
-- Same as in Kobo: we want to ignore keys during suspension
if not Device.screen_saver_mode then
self : sendEvent ( input_event )
end
end
2018-10-31 22:48:36 +00:00
elseif Device : isCervantes ( ) then
self.event_handlers [ " Suspend " ] = function ( )
self : _beforeSuspend ( )
Device : onPowerEvent ( " Suspend " )
end
self.event_handlers [ " Resume " ] = function ( )
Device : onPowerEvent ( " Resume " )
self : _afterResume ( )
end
self.event_handlers [ " PowerPress " ] = function ( )
UIManager : scheduleIn ( 2 , self.poweroff_action )
end
self.event_handlers [ " PowerRelease " ] = function ( )
if not self._entered_poweroff_stage then
UIManager : unschedule ( self.poweroff_action )
2019-09-01 19:37:08 +00:00
-- resume if we were suspended
if Device.screen_saver_mode then
self : resume ( )
else
self : suspend ( )
end
2018-10-31 22:48:36 +00:00
end
end
self.event_handlers [ " Charging " ] = function ( )
2018-11-10 12:16:03 +00:00
self : _beforeCharging ( )
2018-10-31 22:48:36 +00:00
if Device.screen_saver_mode then
self : suspend ( )
2018-11-10 12:16:03 +00:00
end
2018-10-31 22:48:36 +00:00
end
self.event_handlers [ " NotCharging " ] = function ( )
2018-11-10 12:16:03 +00:00
self : _afterNotCharging ( )
2018-10-31 22:48:36 +00:00
if Device.screen_saver_mode then
self : suspend ( )
2018-11-10 12:16:03 +00:00
end
2018-10-31 22:48:36 +00:00
end
self.event_handlers [ " UsbPlugIn " ] = function ( )
2018-11-10 12:16:03 +00:00
self : _beforeCharging ( )
2018-10-31 22:48:36 +00:00
if Device.screen_saver_mode then
self : suspend ( )
2020-08-26 23:06:41 +00:00
else
-- Potentially start an USBMS session
local MassStorage = require ( " ui/elements/mass_storage " )
MassStorage : start ( )
2018-11-10 12:16:03 +00:00
end
2018-10-31 22:48:36 +00:00
end
2020-09-11 21:23:57 +00:00
self.event_handlers [ " UsbPlugOut " ] = function ( )
2018-11-10 12:16:03 +00:00
self : _afterNotCharging ( )
2018-10-31 22:48:36 +00:00
if Device.screen_saver_mode then
self : suspend ( )
2021-08-24 19:03:40 +00:00
else
-- Potentially dismiss the USBMS ConfirmBox
local MassStorage = require ( " ui/elements/mass_storage " )
MassStorage : dismiss ( )
2018-11-10 12:16:03 +00:00
end
2018-10-31 22:48:36 +00:00
end
self.event_handlers [ " __default__ " ] = function ( input_event )
2018-11-10 12:16:03 +00:00
-- Same as in Kobo: we want to ignore keys during suspension
2018-10-31 22:48:36 +00:00
if not Device.screen_saver_mode then
self : sendEvent ( input_event )
end
end
2017-06-23 17:04:11 +00:00
elseif Device : isSDL ( ) then
self.event_handlers [ " Suspend " ] = function ( )
self : _beforeSuspend ( )
Device : simulateSuspend ( )
end
self.event_handlers [ " Resume " ] = function ( )
Device : simulateResume ( )
self : _afterResume ( )
end
2014-06-21 16:24:37 +00:00
end
end
2017-04-09 13:35:01 +00:00
--[[--
Registers and shows a widget .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
Widgets are registered in a stack , from bottom to top in registration order ,
with a few tweaks to handle modals & toasts :
toast widgets are stacked together on top ,
then modal widgets are stacked together , and finally come standard widgets .
If you think about how painting will be handled ( also bottom to top ) , this makes perfect sense ; ) .
For more details about refreshtype , refreshregion & refreshdither see the description of ` setDirty ` .
If refreshtype is omitted , no refresh will be enqueued at this time .
@ param widget a @ { ui.widget . widget | widget } object
@ string refreshtype ` " full " ` , ` " flashpartial " ` , ` " flashui " ` , ` " partial " ` , ` " ui " ` , ` " fast " ` ( optional )
@ param refreshregion a rectangle @ { ui.geometry . Geom | Geom } object ( optional , requires refreshtype to be set )
@ int x horizontal screen offset ( optional , ` 0 ` if omitted )
@ int y vertical screen offset ( optional , ` 0 ` if omitted )
@ bool refreshdither ` true ` if widget requires dithering ( optional , requires refreshtype to be set )
@ see setDirty
2017-04-09 13:35:01 +00:00
] ]
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
function UIManager : show ( widget , refreshtype , refreshregion , x , y , refreshdither )
2016-08-12 06:05:18 +00:00
if not widget then
2016-12-29 08:10:38 +00:00
logger.dbg ( " widget not exist to be shown " )
2016-08-12 06:05:18 +00:00
return
end
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
logger.dbg ( " show widget: " , widget.id or widget.name or tostring ( widget ) )
2016-08-12 06:05:18 +00:00
2014-09-03 03:13:46 +00:00
self._running = true
2014-10-30 08:01:01 +00:00
local window = { x = x or 0 , y = y or 0 , widget = widget }
2021-02-07 17:29:32 +00:00
-- put this window on top of the topmost non-modal window
2014-10-30 08:01:01 +00:00
for i = # self._window_stack , 0 , - 1 do
local top_window = self._window_stack [ i ]
2021-01-11 17:14:12 +00:00
-- toasts are stacked on top of other toasts,
-- then come modals, and then other widgets
if top_window and top_window.widget . toast then
if widget.toast then
table.insert ( self._window_stack , i + 1 , window )
break
end
elseif widget.modal or not top_window or not top_window.widget . modal then
2014-10-30 08:01:01 +00:00
table.insert ( self._window_stack , i + 1 , window )
break
end
end
2014-03-13 13:52:43 +00:00
-- and schedule it to be painted
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
self : setDirty ( widget , refreshtype , refreshregion , refreshdither )
2014-03-13 13:52:43 +00:00
-- tell the widget that it is shown now
widget : handleEvent ( Event : new ( " Show " ) )
-- check if this widget disables double tap gesture
2016-06-27 15:00:49 +00:00
if widget.disable_double_tap == false then
Input.disable_double_tap = false
else
2014-03-13 13:52:43 +00:00
Input.disable_double_tap = true
end
2020-10-28 21:47:42 +00:00
-- a widget may override tap interval (when it doesn't, nil restores the default)
Input.tap_interval_override = widget.tap_interval_override
2012-04-22 19:29:48 +00:00
end
2017-04-09 13:35:01 +00:00
--[[--
Unregisters a widget .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
It will be removed from the stack .
Will flag uncovered widgets as dirty .
For more details about refreshtype , refreshregion & refreshdither see the description of ` setDirty ` .
If refreshtype is omitted , no extra refresh will be enqueued at this time , leaving only those from the uncovered widgets .
@ param widget a @ { ui.widget . widget | widget } object
@ string refreshtype ` " full " ` , ` " flashpartial " ` , ` " flashui " ` , ` " partial " ` , ` " ui " ` , ` " fast " ` ( optional )
@ param refreshregion a rectangle @ { ui.geometry . Geom | Geom } object ( optional , requires refreshtype to be set )
@ bool refreshdither ` true ` if the refresh requires dithering ( optional , requires refreshtype to be set )
@ see setDirty
2017-04-09 13:35:01 +00:00
] ]
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
function UIManager : close ( widget , refreshtype , refreshregion , refreshdither )
2014-11-06 06:00:47 +00:00
if not widget then
2018-06-02 16:10:55 +00:00
logger.dbg ( " widget to be closed does not exist " )
2014-11-06 06:00:47 +00:00
return
end
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
logger.dbg ( " close widget: " , widget.name or widget.id or tostring ( widget ) )
2014-03-13 13:52:43 +00:00
local dirty = false
2021-05-31 20:19:24 +00:00
-- First notify the closed widget to save its settings...
2017-05-16 09:11:11 +00:00
widget : handleEvent ( Event : new ( " FlushSettings " ) )
2021-05-31 20:19:24 +00:00
-- ...and notify it that it ought to be gone now.
2016-06-21 08:50:01 +00:00
widget : handleEvent ( Event : new ( " CloseWidget " ) )
2021-05-31 20:19:24 +00:00
-- Make sure it's disabled by default and check if there are any widgets that want it disabled or enabled.
2016-07-14 00:53:29 +00:00
Input.disable_double_tap = true
2019-03-02 15:50:07 +00:00
local requested_disable_double_tap = nil
2020-10-26 11:28:00 +00:00
local is_covered = false
local start_idx = 1
2021-05-31 20:19:24 +00:00
-- Then remove all references to that widget on stack and refresh.
2014-03-13 13:52:43 +00:00
for i = # self._window_stack , 1 , - 1 do
if self._window_stack [ i ] . widget == widget then
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
self._dirty [ self._window_stack [ i ] . widget ] = nil
2014-03-13 13:52:43 +00:00
table.remove ( self._window_stack , i )
dirty = true
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
else
2020-10-26 11:28:00 +00:00
-- If anything else on the stack not already hidden by (i.e., below) a fullscreen widget was dithered, honor the hint
if self._window_stack [ i ] . widget.dithered and not is_covered then
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
refreshdither = true
logger.dbg ( " Lower widget " , self._window_stack [ i ] . widget.name or self._window_stack [ i ] . widget.id or tostring ( self._window_stack [ i ] . widget ) , " was dithered, honoring the dithering hint " )
end
2020-10-26 11:28:00 +00:00
-- Remember the uppermost widget that covers the full screen, so we don't bother calling setDirty on hidden (i.e., lower) widgets in the following dirty loop.
-- _repaint already does that later on to skip the actual paintTo calls, so this ensures we limit the refresh queue to stuff that will actually get painted.
if not is_covered and self._window_stack [ i ] . widget.covers_fullscreen then
is_covered = true
start_idx = i
logger.dbg ( " Lower widget " , self._window_stack [ i ] . widget.name or self._window_stack [ i ] . widget.id or tostring ( self._window_stack [ i ] . widget ) , " covers the full screen " )
if i > 1 then
logger.dbg ( " not refreshing " , i - 1 , " covered widget(s) " )
end
end
2019-03-02 15:50:07 +00:00
-- Set double tap to how the topmost specifying widget wants it
if requested_disable_double_tap == nil and self._window_stack [ i ] . widget.disable_double_tap ~= nil then
requested_disable_double_tap = self._window_stack [ i ] . widget.disable_double_tap
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
end
2014-03-13 13:52:43 +00:00
end
end
2019-03-02 15:50:07 +00:00
if requested_disable_double_tap ~= nil then
Input.disable_double_tap = requested_disable_double_tap
end
2020-10-28 21:47:42 +00:00
if # self._window_stack > 0 then
-- set tap interval override to what the topmost widget specifies (when it doesn't, nil restores the default)
Input.tap_interval_override = self._window_stack [ # self._window_stack ] . widget.tap_interval_override
end
2018-01-16 11:30:23 +00:00
if dirty and not widget.invisible then
2020-10-26 11:28:00 +00:00
-- schedule the remaining visible (i.e., uncovered) widgets to be painted
for i = start_idx , # self._window_stack do
2014-11-30 22:01:13 +00:00
self : setDirty ( self._window_stack [ i ] . widget )
2014-03-13 13:52:43 +00:00
end
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
self : _refresh ( refreshtype , refreshregion , refreshdither )
2014-03-13 13:52:43 +00:00
end
2012-04-22 19:29:48 +00:00
end
2015-12-26 00:48:40 +00:00
-- schedule an execution task, task queue is in ascendant order
2019-09-12 12:15:08 +00:00
function UIManager : schedule ( time , action , ... )
2015-12-26 00:48:40 +00:00
local p , s , e = 1 , 1 , # self._task_queue
if e ~= 0 then
-- do a binary insert
repeat
p = math.floor ( s + ( e - s ) / 2 )
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
local p_time = self._task_queue [ p ] . time
if time > p_time then
2015-12-26 00:48:40 +00:00
if s == e then
p = e + 1
break
elseif s + 1 == e then
s = e
else
s = p
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
elseif time < p_time then
2015-12-26 00:48:40 +00:00
e = p
if s == e then
break
end
else
2016-04-24 04:16:37 +00:00
-- for fairness, it's better to make p+1 is strictly less than
-- p might want to revisit here in the future
2015-12-26 00:48:40 +00:00
break
end
until e < s
end
2019-09-12 12:15:08 +00:00
table.insert ( self._task_queue , p , {
time = time ,
action = action ,
2021-02-22 01:09:44 +00:00
argc = select ( ' # ' , ... ) ,
2019-09-12 12:15:08 +00:00
args = { ... } ,
} )
2015-12-26 00:48:40 +00:00
self._task_queue_dirty = true
2012-04-22 19:29:48 +00:00
end
2016-03-29 06:37:15 +00:00
dbg : guard ( UIManager , ' schedule ' ,
function ( self , time , action )
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
assert ( time.sec >= 0 , " Only positive time allowed " )
2021-09-25 09:02:10 +00:00
assert ( action ~= nil , " No action " )
2016-03-29 06:37:15 +00:00
end )
2012-04-22 19:29:48 +00:00
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Schedules a task to be run a certain amount of seconds from now .
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
@ number seconds scheduling delay in seconds ( supports decimal values , 1 ms resolution ) .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
@ func action reference to the task to be scheduled ( may be anonymous )
@ param ... optional arguments passed to action
@ see unschedule
] ]
2019-09-12 12:15:08 +00:00
function UIManager : scheduleIn ( seconds , action , ... )
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
-- We might run significantly late inside an UI frame, so we can't use the cached value here.
-- It would also cause some bad interactions with the way nextTick & co behave.
local when = TimeVal : now ( ) + TimeVal : fromnumber ( seconds )
2019-09-12 12:15:08 +00:00
self : schedule ( when , action , ... )
2012-04-22 19:29:48 +00:00
end
2016-03-29 06:37:15 +00:00
dbg : guard ( UIManager , ' scheduleIn ' ,
function ( self , seconds , action )
assert ( seconds >= 0 , " Only positive seconds allowed " )
end )
2012-04-22 19:29:48 +00:00
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Schedules a task for the next UI tick .
@ func action reference to the task to be scheduled ( may be anonymous )
@ param ... optional arguments passed to action
@ see scheduleIn
] ]
function UIManager : nextTick ( action , ... )
return self : scheduleIn ( 0 , action , ... )
2016-02-17 06:36:40 +00:00
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Schedules a task to be run two UI ticks from now .
Useful to run UI callbacks ASAP without skipping repaints .
@ func action reference to the task to be scheduled ( may be anonymous )
@ param ... optional arguments passed to action
@ see nextTick
] ]
function UIManager : tickAfterNext ( action , ... )
-- Storing varargs is a bit iffy as we don't build LuaJIT w/ 5.2 compat, so we don't have access to table.pack...
-- c.f., http://lua-users.org/wiki/VarargTheSecondClassCitizen
local n = select ( ' # ' , ... )
local va = { ... }
return self : nextTick ( function ( ) self : nextTick ( action , unpack ( va , 1 , n ) ) end )
2018-06-02 16:10:55 +00:00
end
--[[
-- NOTE: This appears to work *nearly* just as well, but does sometimes go too fast (might depend on kernel HZ & NO_HZ settings?)
function UIManager : tickAfterNext ( action )
return self : scheduleIn ( 0.001 , action )
end
--]]
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Unschedules a previously scheduled task .
2017-04-09 13:35:01 +00:00
In order to unschedule anonymous functions , store a reference .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
@ func action
@ see scheduleIn
2017-04-09 13:35:01 +00:00
@ usage
self.anonymousFunction = function ( ) self : regularFunction ( ) end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
UIManager : scheduleIn ( 10.5 , self.anonymousFunction )
2017-04-09 13:35:01 +00:00
UIManager : unschedule ( self.anonymousFunction )
] ]
2014-11-06 11:59:00 +00:00
function UIManager : unschedule ( action )
2021-01-29 17:39:08 +00:00
local removed = false
2015-12-26 00:48:40 +00:00
for i = # self._task_queue , 1 , - 1 do
2016-03-08 06:52:52 +00:00
if self._task_queue [ i ] . action == action then
2015-12-26 00:48:40 +00:00
table.remove ( self._task_queue , i )
2021-01-29 17:39:08 +00:00
removed = true
2014-11-06 11:59:00 +00:00
end
end
2021-01-29 17:39:08 +00:00
return removed
2014-11-06 11:59:00 +00:00
end
2016-03-29 06:37:15 +00:00
dbg : guard ( UIManager , ' unschedule ' ,
function ( self , action ) assert ( action ~= nil ) end )
2014-11-06 11:59:00 +00:00
2017-04-09 13:35:01 +00:00
--[[--
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
Mark a window - level widget as dirty , enqueuing a repaint & refresh request for that widget , to be processed on the next UI tick .
2014-11-28 20:12:54 +00:00
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
The second parameter ( refreshtype ) can either specify a refreshtype
2021-02-07 17:29:32 +00:00
( optionally in combination with a refreshregion - which is suggested ,
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
and an even more optional refreshdither flag if the content requires dithering ) ;
or a function that returns a refreshtype , refreshregion tuple ( or a refreshtype , refreshregion , refreshdither triple ) ,
which will be called * after * painting the widget .
Tame some ButtonTable users into re-using Buttontable instances if possible (#7166)
* QuickDictLookup, ImageViewer, NumberPicker: Smarter `update` that will re-use most of the widget's layout instead of re-instantiating all the things.
* SpinWidget/DoubleSpinWidget: The NumberPicker change above renders a hack to preserve alpha on these widgets almost unnecessary. Also fixed said hack to also apply to the center, value button.
* Button: Don't re-instantiate the frame in setText/setIcon when unnecessary (e.g., no change at all, or no layout change).
* Button: Add a refresh method that repaints and refreshes a *specific* Button (provided it's been painted once) all on its lonesome.
* ConfigDialog: Free everything that's going to be re-instatiated on update
* A few more post #7118 fixes:
* SkimTo: Always flag the chapter nav buttons as vsync
* Button: Fix the highlight on rounded buttons when vsync is enabled (e.g., it's now entirely visible, instead of showing a weird inverted corner glitch).
* Some more heuristic tweaks in Menu/TouchMenu/Button/IconButton
* ButtonTable: fix the annoying rounding issue I'd noticed in #7054 ;).
* Enable dithering in TextBoxWidget (e.g., in the Wikipedia full view). This involved moving the HW dithering align fixup to base, where it always ought to have been ;).
* Switch a few widgets that were using "partial" on close to "ui", or, more rarely, "flashui". The intent being to limit "partial" purely to the Reader, because it has a latency cost when mixed with other refreshes, which happens often enough in UI ;).
* Minor documentation tweaks around UIManager's `setDirty` to reflect that change.
* ReaderFooter: Force a footer repaint on resume if it is visible (otherwise, just update it).
* ReaderBookmark: In the same vein, don't repaint an invisible footer on bookmark count changes.
2021-01-28 23:20:15 +00:00
This is an interesting distinction , because a widget ' s geometry,
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
usually stored in a field named ` dimen ` , is ( generally ) only computed at painting time ( e.g . , during ` paintTo ` ) .
Tame some ButtonTable users into re-using Buttontable instances if possible (#7166)
* QuickDictLookup, ImageViewer, NumberPicker: Smarter `update` that will re-use most of the widget's layout instead of re-instantiating all the things.
* SpinWidget/DoubleSpinWidget: The NumberPicker change above renders a hack to preserve alpha on these widgets almost unnecessary. Also fixed said hack to also apply to the center, value button.
* Button: Don't re-instantiate the frame in setText/setIcon when unnecessary (e.g., no change at all, or no layout change).
* Button: Add a refresh method that repaints and refreshes a *specific* Button (provided it's been painted once) all on its lonesome.
* ConfigDialog: Free everything that's going to be re-instatiated on update
* A few more post #7118 fixes:
* SkimTo: Always flag the chapter nav buttons as vsync
* Button: Fix the highlight on rounded buttons when vsync is enabled (e.g., it's now entirely visible, instead of showing a weird inverted corner glitch).
* Some more heuristic tweaks in Menu/TouchMenu/Button/IconButton
* ButtonTable: fix the annoying rounding issue I'd noticed in #7054 ;).
* Enable dithering in TextBoxWidget (e.g., in the Wikipedia full view). This involved moving the HW dithering align fixup to base, where it always ought to have been ;).
* Switch a few widgets that were using "partial" on close to "ui", or, more rarely, "flashui". The intent being to limit "partial" purely to the Reader, because it has a latency cost when mixed with other refreshes, which happens often enough in UI ;).
* Minor documentation tweaks around UIManager's `setDirty` to reflect that change.
* ReaderFooter: Force a footer repaint on resume if it is visible (otherwise, just update it).
* ReaderBookmark: In the same vein, don't repaint an invisible footer on bookmark count changes.
2021-01-28 23:20:15 +00:00
The TL ; DR being : if you already know the region , you can pass everything by value directly ,
( it ' ll make for slightly more readable debug logs),
but if the region will only be known after the widget has been painted , pass a function .
Note that , technically , it means that stuff passed by value will be enqueued earlier in the refresh stack .
In practice , since the stack of ( both types of ) refreshes is optimized into as few actual refresh ioctls as possible ,
and that during the next ` _repaint ` tick ( which is when ` paintTo ` for dirty widgets happens ) ,
this shouldn ' t change much in the grand scheme of things, but it ought to be noted ;).
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
See ` _repaint ` for more details about how the repaint & refresh queues are processed ,
and ` handleInput ` for more details about when those queues are actually drained .
What you should essentially remember is that ` setDirty ` doesn ' t actually "do" anything visible on its own.
It doesn ' t block, and when it returns, nothing new has actually been painted or refreshed.
It just appends stuff to the paint and / or refresh queues .
2018-06-02 16:10:55 +00:00
Here ' s a quick rundown of what each refreshtype should be used for:
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
* ` full ` : high - fidelity flashing refresh ( e.g . , large images ) .
Highest quality , but highest latency .
Don ' t abuse if you only want a flash (in this case, prefer `flashui` or `flashpartial`).
* ` partial ` : medium fidelity refresh ( e.g . , text on a white background ) .
Can be promoted to flashing after ` FULL_REFRESH_COUNT ` refreshes .
Don ' t abuse to avoid spurious flashes.
In practice , this means this should mostly always be limited to ReaderUI .
* ` ui ` : medium fidelity refresh ( e.g . , mixed content ) .
Should apply to most UI elements .
When in doubt , use this .
* ` fast ` : low fidelity refresh ( e.g . , monochrome content ) .
Should apply to most highlighting effects achieved through inversion .
Note that if your highlighted element contains text ,
you might want to keep the unhighlight refresh as ` " ui " ` instead , for crisper text .
( Or optimize that refresh away entirely , if you can get away with it ) .
* ` flashui ` : like ` ui ` , but flashing .
Can be used when showing a UI element for the first time , or when closing one , to avoid ghosting .
* ` flashpartial ` : like ` partial ` , but flashing ( and not counting towards flashing promotions ) .
Can be used when closing an UI element ( usually over ReaderUI ) , to avoid ghosting .
You can even drop the region in these cases , to ensure a fullscreen flash .
NOTE : On REAGL devices , ` flashpartial ` will NOT actually flash ( by design ) .
As such , even onCloseWidget , you might prefer ` flashui ` in most instances .
NOTE : You ' ll notice a trend on UI elements that are usually shown *over* some kind of text (generally ReaderUI)
of using ` " ui " ` onShow & onUpdate , but ` " partial " ` onCloseWidget .
This is by design : ` " partial " ` is what the reader ( ReaderUI ) uses , as it ' s tailor-made for pure text
over a white background , so this ensures we resume the usual flow of the reader .
The same dynamic is true for their flashing counterparts , in the rare instances we enforce flashes .
Any kind of ` " partial " ` refresh * will * count towards a flashing promotion after ` FULL_REFRESH_COUNT ` refreshes ,
so making sure your stuff only applies to the proper region is key to avoiding spurious large black flashes .
That said , depending on your use case , using ` " ui " ` onCloseWidget can be a perfectly valid decision ,
and will ensure never seeing a flash because of that widget .
Remember that the FM uses ` " ui " ` , so , if said widgets are shown over the FM ,
prefer using ` " ui " ` or ` " flashui " ` onCloseWidget .
2018-06-02 16:10:55 +00:00
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
The final parameter ( refreshdither ) is an optional hint for devices with hardware dithering support that this repaint
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
could benefit from dithering ( e.g . , because it contains an image ) .
2014-11-28 20:12:54 +00:00
Tame some ButtonTable users into re-using Buttontable instances if possible (#7166)
* QuickDictLookup, ImageViewer, NumberPicker: Smarter `update` that will re-use most of the widget's layout instead of re-instantiating all the things.
* SpinWidget/DoubleSpinWidget: The NumberPicker change above renders a hack to preserve alpha on these widgets almost unnecessary. Also fixed said hack to also apply to the center, value button.
* Button: Don't re-instantiate the frame in setText/setIcon when unnecessary (e.g., no change at all, or no layout change).
* Button: Add a refresh method that repaints and refreshes a *specific* Button (provided it's been painted once) all on its lonesome.
* ConfigDialog: Free everything that's going to be re-instatiated on update
* A few more post #7118 fixes:
* SkimTo: Always flag the chapter nav buttons as vsync
* Button: Fix the highlight on rounded buttons when vsync is enabled (e.g., it's now entirely visible, instead of showing a weird inverted corner glitch).
* Some more heuristic tweaks in Menu/TouchMenu/Button/IconButton
* ButtonTable: fix the annoying rounding issue I'd noticed in #7054 ;).
* Enable dithering in TextBoxWidget (e.g., in the Wikipedia full view). This involved moving the HW dithering align fixup to base, where it always ought to have been ;).
* Switch a few widgets that were using "partial" on close to "ui", or, more rarely, "flashui". The intent being to limit "partial" purely to the Reader, because it has a latency cost when mixed with other refreshes, which happens often enough in UI ;).
* Minor documentation tweaks around UIManager's `setDirty` to reflect that change.
* ReaderFooter: Force a footer repaint on resume if it is visible (otherwise, just update it).
* ReaderBookmark: In the same vein, don't repaint an invisible footer on bookmark count changes.
2021-01-28 23:20:15 +00:00
As far as the actual lifecycle of a widget goes , the rules are :
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
Tame some ButtonTable users into re-using Buttontable instances if possible (#7166)
* QuickDictLookup, ImageViewer, NumberPicker: Smarter `update` that will re-use most of the widget's layout instead of re-instantiating all the things.
* SpinWidget/DoubleSpinWidget: The NumberPicker change above renders a hack to preserve alpha on these widgets almost unnecessary. Also fixed said hack to also apply to the center, value button.
* Button: Don't re-instantiate the frame in setText/setIcon when unnecessary (e.g., no change at all, or no layout change).
* Button: Add a refresh method that repaints and refreshes a *specific* Button (provided it's been painted once) all on its lonesome.
* ConfigDialog: Free everything that's going to be re-instatiated on update
* A few more post #7118 fixes:
* SkimTo: Always flag the chapter nav buttons as vsync
* Button: Fix the highlight on rounded buttons when vsync is enabled (e.g., it's now entirely visible, instead of showing a weird inverted corner glitch).
* Some more heuristic tweaks in Menu/TouchMenu/Button/IconButton
* ButtonTable: fix the annoying rounding issue I'd noticed in #7054 ;).
* Enable dithering in TextBoxWidget (e.g., in the Wikipedia full view). This involved moving the HW dithering align fixup to base, where it always ought to have been ;).
* Switch a few widgets that were using "partial" on close to "ui", or, more rarely, "flashui". The intent being to limit "partial" purely to the Reader, because it has a latency cost when mixed with other refreshes, which happens often enough in UI ;).
* Minor documentation tweaks around UIManager's `setDirty` to reflect that change.
* ReaderFooter: Force a footer repaint on resume if it is visible (otherwise, just update it).
* ReaderBookmark: In the same vein, don't repaint an invisible footer on bookmark count changes.
2021-01-28 23:20:15 +00:00
* What you ` show ` , you ` close ` .
* If you know the dimensions of the widget ( or simply of the region you want to refresh ) , you can pass it directly :
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
* to ` show ` ( as ` show ` calls ` setDirty ` ) ,
* to ` close ` ( as ` close ` will also call ` setDirty ` on the remaining dirty and visible widgets ,
and will also enqueue a refresh based on that if there are dirty widgets ) .
* Otherwise , you can use , respectively , a widget ' s `Show` & `CloseWidget` handlers for that via `setDirty` calls.
Tame some ButtonTable users into re-using Buttontable instances if possible (#7166)
* QuickDictLookup, ImageViewer, NumberPicker: Smarter `update` that will re-use most of the widget's layout instead of re-instantiating all the things.
* SpinWidget/DoubleSpinWidget: The NumberPicker change above renders a hack to preserve alpha on these widgets almost unnecessary. Also fixed said hack to also apply to the center, value button.
* Button: Don't re-instantiate the frame in setText/setIcon when unnecessary (e.g., no change at all, or no layout change).
* Button: Add a refresh method that repaints and refreshes a *specific* Button (provided it's been painted once) all on its lonesome.
* ConfigDialog: Free everything that's going to be re-instatiated on update
* A few more post #7118 fixes:
* SkimTo: Always flag the chapter nav buttons as vsync
* Button: Fix the highlight on rounded buttons when vsync is enabled (e.g., it's now entirely visible, instead of showing a weird inverted corner glitch).
* Some more heuristic tweaks in Menu/TouchMenu/Button/IconButton
* ButtonTable: fix the annoying rounding issue I'd noticed in #7054 ;).
* Enable dithering in TextBoxWidget (e.g., in the Wikipedia full view). This involved moving the HW dithering align fixup to base, where it always ought to have been ;).
* Switch a few widgets that were using "partial" on close to "ui", or, more rarely, "flashui". The intent being to limit "partial" purely to the Reader, because it has a latency cost when mixed with other refreshes, which happens often enough in UI ;).
* Minor documentation tweaks around UIManager's `setDirty` to reflect that change.
* ReaderFooter: Force a footer repaint on resume if it is visible (otherwise, just update it).
* ReaderBookmark: In the same vein, don't repaint an invisible footer on bookmark count changes.
2021-01-28 23:20:15 +00:00
This can also be useful if * child * widgets have specific needs ( e.g . , flashing , dithering ) that they want to inject in the refresh queue .
* Remember that events propagate children first ( in array order , starting at the top ) , and that if * any * event handler returns true ,
the propagation of that specific event for this widget tree stops * immediately * .
( This generally means that , unless you know what you ' re doing (e.g., a widget that will *always* be used as a parent),
you generally * don ' t* want to return true in `Show` or `CloseWidget` handlers).
* If any widget requires freeing non - Lua resources ( e.g . , FFI / C ) , having a ` free ` method called from its ` CloseWidget ` handler is ideal :
this ' ll ensure that *any* widget including it will be sure that resources are freed when it (or its parent) are closed.
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
* Note that there * is * a ` Close ` event , but it has very specific use - cases , generally involving * programmatically * ` close ` ing a ` show ` n widget :
* It is broadcast ( e.g . , sent to every widget in the window stack ; the same rules about propagation apply , but only per * window - level widget * )
at poweroff / reboot .
* It can also be used as a keypress handler by @ { ui.widget . container.inputcontainer | InputContainer } , generally bound to the Back key .
Please refrain from implementing custom ` onClose ` methods if that ' s not their intended purpose ;).
Tame some ButtonTable users into re-using Buttontable instances if possible (#7166)
* QuickDictLookup, ImageViewer, NumberPicker: Smarter `update` that will re-use most of the widget's layout instead of re-instantiating all the things.
* SpinWidget/DoubleSpinWidget: The NumberPicker change above renders a hack to preserve alpha on these widgets almost unnecessary. Also fixed said hack to also apply to the center, value button.
* Button: Don't re-instantiate the frame in setText/setIcon when unnecessary (e.g., no change at all, or no layout change).
* Button: Add a refresh method that repaints and refreshes a *specific* Button (provided it's been painted once) all on its lonesome.
* ConfigDialog: Free everything that's going to be re-instatiated on update
* A few more post #7118 fixes:
* SkimTo: Always flag the chapter nav buttons as vsync
* Button: Fix the highlight on rounded buttons when vsync is enabled (e.g., it's now entirely visible, instead of showing a weird inverted corner glitch).
* Some more heuristic tweaks in Menu/TouchMenu/Button/IconButton
* ButtonTable: fix the annoying rounding issue I'd noticed in #7054 ;).
* Enable dithering in TextBoxWidget (e.g., in the Wikipedia full view). This involved moving the HW dithering align fixup to base, where it always ought to have been ;).
* Switch a few widgets that were using "partial" on close to "ui", or, more rarely, "flashui". The intent being to limit "partial" purely to the Reader, because it has a latency cost when mixed with other refreshes, which happens often enough in UI ;).
* Minor documentation tweaks around UIManager's `setDirty` to reflect that change.
* ReaderFooter: Force a footer repaint on resume if it is visible (otherwise, just update it).
* ReaderBookmark: In the same vein, don't repaint an invisible footer on bookmark count changes.
2021-01-28 23:20:15 +00:00
On the subject of widgets and child widgets ,
you might have noticed an unspoken convention across the codebase of widgets having a field called ` show_parent ` .
Since handling this is entirely at the programmer ' s behest, here ' s how we usually use it :
Basically , we cascade a field named ` show_parent ` to every child widget that matter
( e.g . , those that serve an UI purpose , as opposed to , say , a container ) .
This ensures that every subwidget can reference its actual parent
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
( ideally , all the way to the window - level widget it belongs to , i.e . , the one that was passed to ` show ` , hence the name ; ) ) ,
to , among other things , flag the right widget for repaint via ` setDirty ` ( c.f . , those pesky debug warnings when that ' s done wrong ;p) when they want to request a repaint.
This is why you often see stuff doing , when instantiating a new widget , ` FancyWidget : new { show_parent = self.show_parent or self } ` ;
Tame some ButtonTable users into re-using Buttontable instances if possible (#7166)
* QuickDictLookup, ImageViewer, NumberPicker: Smarter `update` that will re-use most of the widget's layout instead of re-instantiating all the things.
* SpinWidget/DoubleSpinWidget: The NumberPicker change above renders a hack to preserve alpha on these widgets almost unnecessary. Also fixed said hack to also apply to the center, value button.
* Button: Don't re-instantiate the frame in setText/setIcon when unnecessary (e.g., no change at all, or no layout change).
* Button: Add a refresh method that repaints and refreshes a *specific* Button (provided it's been painted once) all on its lonesome.
* ConfigDialog: Free everything that's going to be re-instatiated on update
* A few more post #7118 fixes:
* SkimTo: Always flag the chapter nav buttons as vsync
* Button: Fix the highlight on rounded buttons when vsync is enabled (e.g., it's now entirely visible, instead of showing a weird inverted corner glitch).
* Some more heuristic tweaks in Menu/TouchMenu/Button/IconButton
* ButtonTable: fix the annoying rounding issue I'd noticed in #7054 ;).
* Enable dithering in TextBoxWidget (e.g., in the Wikipedia full view). This involved moving the HW dithering align fixup to base, where it always ought to have been ;).
* Switch a few widgets that were using "partial" on close to "ui", or, more rarely, "flashui". The intent being to limit "partial" purely to the Reader, because it has a latency cost when mixed with other refreshes, which happens often enough in UI ;).
* Minor documentation tweaks around UIManager's `setDirty` to reflect that change.
* ReaderFooter: Force a footer repaint on resume if it is visible (otherwise, just update it).
* ReaderBookmark: In the same vein, don't repaint an invisible footer on bookmark count changes.
2021-01-28 23:20:15 +00:00
meaning , if I ' m already a subwidget, cascade my parent, otherwise, it means I ' m a window - level widget , so cascade myself as that widget ' s parent ;).
2021-02-02 03:27:14 +00:00
Another convention ( that a few things rely on ) is naming a ( persistent ) MovableContainer wrapping a full widget ` movable ` , accessible as an instance field .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
This is useful when it ' s used for transparency purposes, which, e.g., `setDirty` and @{ui.widget.button|Button} rely on to handle updating translucent widgets properly,
2021-02-02 03:27:14 +00:00
by checking if self.show_parent . movable exists and is currently translucent ; ) .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
When I mentioned passing the * right * widget to ` setDirty ` earlier , what I meant is that ` setDirty ` will only actually flag a widget for repaint
2021-02-07 17:29:32 +00:00
* if * that widget is a window - level widget ( that is , a widget that was passed to ` show ` earlier and hasn ' t been `close` ' d yet ) ,
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
hence the ` self.show_parent ` convention detailed above to get at the proper widget from within a subwidget ; ) .
2021-02-07 17:29:32 +00:00
Otherwise , you ' ll notice in debug mode that a debug guard will shout at you if that contract is broken,
and what happens in practice is the same thing as if an explicit ` nil ` were passed : no widgets will actually be flagged for repaint ,
and only the * refresh * matching the requested region * will * be enqueued .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
This is why you ' ll find a number of valid use-cases for passing a `nil` here, when you *just* want a screen refresh without a repaint :).
The string ` " all " ` is also accepted in place of a widget , and will do the obvious thing : flag the * full * window stack , bottom to top , for repaint ,
2021-02-07 17:29:32 +00:00
while still honoring the refresh region ( e.g . , this doesn ' t enforce a full-screen refresh).
2017-04-09 13:35:01 +00:00
@ usage
2014-11-28 20:12:54 +00:00
UIManager : setDirty ( self.widget , " partial " )
UIManager : setDirty ( self.widget , " partial " , Geom : new { x = 10 , y = 10 , w = 100 , h = 50 } )
UIManager : setDirty ( self.widget , function ( ) return " ui " , self.someelement . dimen end )
2017-04-09 13:35:01 +00:00
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
@ param widget a window - level widget object , ` " all " ` , or ` nil `
@ param refreshtype ` " full " ` , ` " flashpartial " ` , ` " flashui " ` , ` " partial " ` , ` " ui " ` , ` " fast " ` ( or a lambda , see description above )
@ param refreshregion a rectangle @ { ui.geometry . Geom | Geom } object ( optional , omitting it means the region will cover the full screen )
@ bool refreshdither ` true ` if widget requires dithering ( optional )
] ]
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
function UIManager : setDirty ( widget , refreshtype , refreshregion , refreshdither )
2014-08-20 06:45:11 +00:00
if widget then
2014-11-28 20:12:54 +00:00
if widget == " all " then
-- special case: set all top-level widgets as being "dirty".
for i = 1 , # self._window_stack do
self._dirty [ self._window_stack [ i ] . widget ] = true
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
-- If any of 'em were dithered, honor their dithering hint
if self._window_stack [ i ] . widget.dithered then
-- NOTE: That works when refreshtype is NOT a function,
-- which is why _repaint does another pass of this check ;).
2020-02-09 21:50:06 +00:00
logger.dbg ( " setDirty on all widgets: found a dithered widget, infecting the refresh queue " )
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
refreshdither = true
end
2014-11-28 20:12:54 +00:00
end
2018-01-16 11:30:23 +00:00
elseif not widget.invisible then
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- NOTE: If our widget is translucent, or belongs to a translucent MovableContainer,
-- we'll want to flag everything below it as dirty, too,
-- because doing transparency right requires having an up to date background against which to blend.
-- (The typecheck is because some widgets use an alpha boolean trap for internal alpha handling (e.g., ImageWidget)).
local handle_alpha = false
-- NOTE: We only ever check the dirty flag on top-level widgets, so only set it there!
-- Enable verbose debug to catch misbehaving widgets via our post-guard.
for i = # self._window_stack , 1 , - 1 do
if handle_alpha then
self._dirty [ self._window_stack [ i ] . widget ] = true
logger.dbg ( " setDirty: Marking as dirty widget: " , self._window_stack [ i ] . widget.name or self._window_stack [ i ] . widget.id or tostring ( self._window_stack [ i ] . widget ) , " because it's below translucent widget: " , widget.name or widget.id or tostring ( widget ) )
-- Stop flagging widgets at the uppermost one that covers the full screen
if self._window_stack [ i ] . widget.covers_fullscreen then
break
end
end
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
if self._window_stack [ i ] . widget == widget then
self._dirty [ widget ] = true
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- We've got a match, now check if it's translucent...
handle_alpha = ( widget.alpha and type ( widget.alpha ) == " number " and widget.alpha < 1 and widget.alpha > 0 )
or ( widget.movable and widget.movable . alpha and widget.movable . alpha < 1 and widget.movable . alpha > 0 )
-- We shouldn't be seeing the same widget at two different spots in the stack, so, we're done,
-- except when we need to keep looping to flag widgets below us in order to handle a translucent widget...
if not handle_alpha then
break
end
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
end
end
-- Again, if it's flagged as dithered, honor that
if widget.dithered then
refreshdither = true
end
2014-11-28 20:12:54 +00:00
end
2020-02-09 21:50:06 +00:00
else
-- Another special case: if we did NOT specify a widget, but requested a full refresh nonetheless (i.e., a diagonal swipe),
-- we'll want to check the window stack in order to honor dithering...
if refreshtype == " full " then
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
for i = # self._window_stack , 1 , - 1 do
2020-02-09 21:50:06 +00:00
-- If any of 'em were dithered, honor their dithering hint
if self._window_stack [ i ] . widget.dithered then
logger.dbg ( " setDirty full on no specific widget: found a dithered widget, infecting the refresh queue " )
refreshdither = true
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- One is enough ;)
break
2020-02-09 21:50:06 +00:00
end
end
end
2014-11-28 20:12:54 +00:00
end
-- handle refresh information
if type ( refreshtype ) == " function " then
-- callback, will be issued after painting
table.insert ( self._refresh_func_stack , refreshtype )
2018-06-02 16:10:55 +00:00
if dbg.is_on then
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- NOTE: It's too early to tell what the function will return (especially the region), because the widget hasn't been painted yet.
-- Consuming the lambda now also appears have nasty side-effects that render it useless later, subtly breaking a whole lot of things...
-- Thankfully, we can track them in _refresh()'s logging very soon after that...
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
logger.dbg ( " setDirty via a func from widget " , widget and ( widget.name or widget.id or tostring ( widget ) ) or " nil " )
2018-06-02 16:10:55 +00:00
end
2014-11-28 20:12:54 +00:00
else
-- otherwise, enqueue refresh
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
self : _refresh ( refreshtype , refreshregion , refreshdither )
2018-06-02 16:10:55 +00:00
if dbg.is_on then
if refreshregion then
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
logger.dbg ( " setDirty " , refreshtype and refreshtype or " nil " , " from widget " , widget and ( widget.name or widget.id or tostring ( widget ) ) or " nil " , " w/ region " , refreshregion.x , refreshregion.y , refreshregion.w , refreshregion.h , refreshdither and " AND w/ HW dithering " or " " )
2018-06-02 16:10:55 +00:00
else
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
logger.dbg ( " setDirty " , refreshtype and refreshtype or " nil " , " from widget " , widget and ( widget.name or widget.id or tostring ( widget ) ) or " nil " , " w/ NO region " , refreshdither and " AND w/ HW dithering " or " " )
2018-06-02 16:10:55 +00:00
end
end
2014-08-20 06:45:11 +00:00
end
2012-04-22 19:29:48 +00:00
end
2016-03-29 06:37:15 +00:00
dbg : guard ( UIManager , ' setDirty ' ,
nil ,
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
function ( self , widget , refreshtype , refreshregion , refreshdither )
2016-06-21 08:50:01 +00:00
if not widget or widget == " all " then return end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- when debugging, we check if we were handed a valid window-level widget,
-- which would be a widget that was previously passed to `show`.
2016-03-29 06:37:15 +00:00
local found = false
for i = 1 , # self._window_stack do
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
if self._window_stack [ i ] . widget == widget then
found = true
break
end
2016-03-29 06:37:15 +00:00
end
if not found then
2016-08-12 06:05:18 +00:00
dbg : v ( " INFO: invalid widget for setDirty() " , debug.traceback ( ) )
2016-03-29 06:37:15 +00:00
end
end )
2012-04-22 19:29:48 +00:00
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Clear the full repaint & refresh queues .
NOTE : Beware ! This doesn ' t take any prisonners!
You shouldn ' t have to resort to this unless in very specific circumstances!
plugins / coverbrowser.koplugin / covermenu.lua building a franken - menu out of buttondialogtitle & buttondialog
and wanting to avoid inheriting their original paint / refresh cycle being a prime example .
--]]
2018-06-02 16:10:55 +00:00
function UIManager : clearRenderStack ( )
logger.dbg ( " clearRenderStack: Clearing the full render stack! " )
self._dirty = { }
self._refresh_func_stack = { }
self._refresh_stack = { }
end
2014-06-23 14:33:11 +00:00
function UIManager : insertZMQ ( zeromq )
table.insert ( self._zeromqs , zeromq )
return zeromq
end
function UIManager : removeZMQ ( zeromq )
for i = # self._zeromqs , 1 , - 1 do
if self._zeromqs [ i ] == zeromq then
table.remove ( self._zeromqs , i )
end
end
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Sets the full refresh rate for e - ink screens ( ` FULL_REFRESH_COUNT ` ) .
This is the amount of ` " partial " ` refreshes before the next one gets promoted to ` " full " ` .
Also makes the refresh rate persistent in global reader settings .
@ see setDirty
--]]
2020-07-17 06:54:21 +00:00
function UIManager : setRefreshRate ( rate , night_rate )
2020-08-21 12:33:08 +00:00
logger.dbg ( " set screen full refresh rate " , rate , night_rate )
if G_reader_settings : isTrue ( " night_mode " ) then
if night_rate then
self.FULL_REFRESH_COUNT = night_rate
end
else
if rate then
self.FULL_REFRESH_COUNT = rate
end
end
if rate then
G_reader_settings : saveSetting ( " full_refresh_count " , rate )
end
if night_rate then
G_reader_settings : saveSetting ( " night_full_refresh_count " , night_rate )
end
2014-07-26 14:22:52 +00:00
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--- Returns the full refresh rate for e-ink screens (`FULL_REFRESH_COUNT`).
2020-07-17 06:54:21 +00:00
function UIManager : getRefreshRate ( )
2020-07-20 06:48:07 +00:00
return G_reader_settings : readSetting ( " full_refresh_count " ) or DEFAULT_FULL_REFRESH_COUNT , G_reader_settings : readSetting ( " night_full_refresh_count " ) or G_reader_settings : readSetting ( " full_refresh_count " ) or DEFAULT_FULL_REFRESH_COUNT
2020-07-17 06:54:21 +00:00
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--- Toggles Night Mode (i.e., inverted rendering).
2020-07-17 06:54:21 +00:00
function UIManager : ToggleNightMode ( night_mode )
if night_mode then
2020-07-20 06:48:07 +00:00
self.FULL_REFRESH_COUNT = G_reader_settings : readSetting ( " night_full_refresh_count " ) or G_reader_settings : readSetting ( " full_refresh_count " ) or DEFAULT_FULL_REFRESH_COUNT
2020-07-17 06:54:21 +00:00
else
2020-07-20 06:48:07 +00:00
self.FULL_REFRESH_COUNT = G_reader_settings : readSetting ( " full_refresh_count " ) or DEFAULT_FULL_REFRESH_COUNT
2020-07-17 06:54:21 +00:00
end
2014-07-26 14:22:52 +00:00
end
Revamp "flash_ui" handling (#7118)
* Wherever possible, do an actual dumb invert on the Screen BB instead of repainting the widget, *then* inverting it (which is what the "invert" flag does).
* Instead of playing with nextTick/tickAfterNext delays, explicitly fence stuff with forceRePaint
* And, in the few cases where said Mk. 7 quirk kicks in, make the fences more marked by using a well-placed WAIT_FOR_UPDATE_COMPLETE
* Fix an issue in Button where show/hide & enable/disable where actually all toggles, which meant that duplicate calls or timing issues would do the wrong thing. (This broke dimming some icons, and mistakenly dropped the background from FM chevrons, for example).
* Speaking of, fix Button's hide/show to actually restore the background properly (there was a stupid typo in the variable name)
* Still in Button, fix the insanity of the double repaint on rounded buttons. Turns out it made sense, after all (and was related to said missing background, and bad interaction with invert & text with no background).
* KeyValuePage suffered from a similar issue with broken highlights (all black) because of missing backgrounds.
* In ConfigDialog, only instanciate IconButtons once (because every tab switch causes a full instantiation; and the initial display implies a full instanciation and an initial tab switch). Otherwise, both instances linger, and catch taps, and as such, double highlights.
* ConfigDialog: Restore the "don't repaint ReaderUI" when switching between similarly sized tabs (re #6131). I never could reproduce that on eInk, and I can't now on the emulator, so I'm assuming @poire-z fixed it during the swap to SVG icons.
* KeyValuePage: Only instanciate Buttons once (again, this is a widget that goes through a full init every page). Again, caused highlight/dimming issues because buttons were stacked.
* Menu: Ditto.
* TouchMenu: Now home of the gnarliest unhilight heuristics, because of the sheer amount of different things that can happen (and/or thanks to stuff not flagged covers_fullscreen properly ;p).
* Bump base
https://github.com/koreader/koreader-base/pull/1280
https://github.com/koreader/koreader-base/pull/1282
https://github.com/koreader/koreader-base/pull/1283
https://github.com/koreader/koreader-base/pull/1284
* Bump android-luajit-launcher
https://github.com/koreader/android-luajit-launcher/pull/284
https://github.com/koreader/android-luajit-launcher/pull/285
https://github.com/koreader/android-luajit-launcher/pull/286
https://github.com/koreader/android-luajit-launcher/pull/287
2021-01-10 00:51:09 +00:00
--- Get top widget (name if possible, ref otherwise).
2019-09-01 19:30:19 +00:00
function UIManager : getTopWidget ( )
Revamp "flash_ui" handling (#7118)
* Wherever possible, do an actual dumb invert on the Screen BB instead of repainting the widget, *then* inverting it (which is what the "invert" flag does).
* Instead of playing with nextTick/tickAfterNext delays, explicitly fence stuff with forceRePaint
* And, in the few cases where said Mk. 7 quirk kicks in, make the fences more marked by using a well-placed WAIT_FOR_UPDATE_COMPLETE
* Fix an issue in Button where show/hide & enable/disable where actually all toggles, which meant that duplicate calls or timing issues would do the wrong thing. (This broke dimming some icons, and mistakenly dropped the background from FM chevrons, for example).
* Speaking of, fix Button's hide/show to actually restore the background properly (there was a stupid typo in the variable name)
* Still in Button, fix the insanity of the double repaint on rounded buttons. Turns out it made sense, after all (and was related to said missing background, and bad interaction with invert & text with no background).
* KeyValuePage suffered from a similar issue with broken highlights (all black) because of missing backgrounds.
* In ConfigDialog, only instanciate IconButtons once (because every tab switch causes a full instantiation; and the initial display implies a full instanciation and an initial tab switch). Otherwise, both instances linger, and catch taps, and as such, double highlights.
* ConfigDialog: Restore the "don't repaint ReaderUI" when switching between similarly sized tabs (re #6131). I never could reproduce that on eInk, and I can't now on the emulator, so I'm assuming @poire-z fixed it during the swap to SVG icons.
* KeyValuePage: Only instanciate Buttons once (again, this is a widget that goes through a full init every page). Again, caused highlight/dimming issues because buttons were stacked.
* Menu: Ditto.
* TouchMenu: Now home of the gnarliest unhilight heuristics, because of the sheer amount of different things that can happen (and/or thanks to stuff not flagged covers_fullscreen properly ;p).
* Bump base
https://github.com/koreader/koreader-base/pull/1280
https://github.com/koreader/koreader-base/pull/1282
https://github.com/koreader/koreader-base/pull/1283
https://github.com/koreader/koreader-base/pull/1284
* Bump android-luajit-launcher
https://github.com/koreader/android-luajit-launcher/pull/284
https://github.com/koreader/android-luajit-launcher/pull/285
https://github.com/koreader/android-luajit-launcher/pull/286
https://github.com/koreader/android-luajit-launcher/pull/287
2021-01-10 00:51:09 +00:00
local top = self._window_stack [ # self._window_stack ]
if not top or not top.widget then
return nil
end
if top.widget . name then
return top.widget . name
end
return top.widget
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Get the * second * topmost widget , if there is one ( name if possible , ref otherwise ) .
Useful when VirtualKeyboard is involved , as it * always * steals the top spot ; ) .
NOTE : Will skip over VirtualKeyboard instances , plural , in case there are multiple ( because , apparently , we can do that .. ugh ) .
--]]
2021-02-07 17:29:32 +00:00
function UIManager : getSecondTopmostWidget ( )
if # self._window_stack <= 1 then
-- Not enough widgets in the stack, bye!
return nil
end
-- Because everything is terrible, you can actually instantiate multiple VirtualKeyboards,
-- and they'll stack at the top, so, loop until we get something that *isn't* VK...
for i = # self._window_stack - 1 , 1 , - 1 do
local sec = self._window_stack [ i ]
if not sec or not sec.widget then
return nil
end
if sec.widget . name then
if sec.widget . name ~= " VirtualKeyboard " then
return sec.widget . name
end
-- Meaning if name is set, and is set to VK => continue, as we want the *next* widget.
-- I *really* miss the continue keyword, Lua :/.
else
return sec.widget
end
end
return nil
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--- Check if a widget is still in the window stack, or is a subwidget of a widget still in the window stack.
Revamp "flash_ui" handling (#7118)
* Wherever possible, do an actual dumb invert on the Screen BB instead of repainting the widget, *then* inverting it (which is what the "invert" flag does).
* Instead of playing with nextTick/tickAfterNext delays, explicitly fence stuff with forceRePaint
* And, in the few cases where said Mk. 7 quirk kicks in, make the fences more marked by using a well-placed WAIT_FOR_UPDATE_COMPLETE
* Fix an issue in Button where show/hide & enable/disable where actually all toggles, which meant that duplicate calls or timing issues would do the wrong thing. (This broke dimming some icons, and mistakenly dropped the background from FM chevrons, for example).
* Speaking of, fix Button's hide/show to actually restore the background properly (there was a stupid typo in the variable name)
* Still in Button, fix the insanity of the double repaint on rounded buttons. Turns out it made sense, after all (and was related to said missing background, and bad interaction with invert & text with no background).
* KeyValuePage suffered from a similar issue with broken highlights (all black) because of missing backgrounds.
* In ConfigDialog, only instanciate IconButtons once (because every tab switch causes a full instantiation; and the initial display implies a full instanciation and an initial tab switch). Otherwise, both instances linger, and catch taps, and as such, double highlights.
* ConfigDialog: Restore the "don't repaint ReaderUI" when switching between similarly sized tabs (re #6131). I never could reproduce that on eInk, and I can't now on the emulator, so I'm assuming @poire-z fixed it during the swap to SVG icons.
* KeyValuePage: Only instanciate Buttons once (again, this is a widget that goes through a full init every page). Again, caused highlight/dimming issues because buttons were stacked.
* Menu: Ditto.
* TouchMenu: Now home of the gnarliest unhilight heuristics, because of the sheer amount of different things that can happen (and/or thanks to stuff not flagged covers_fullscreen properly ;p).
* Bump base
https://github.com/koreader/koreader-base/pull/1280
https://github.com/koreader/koreader-base/pull/1282
https://github.com/koreader/koreader-base/pull/1283
https://github.com/koreader/koreader-base/pull/1284
* Bump android-luajit-launcher
https://github.com/koreader/android-luajit-launcher/pull/284
https://github.com/koreader/android-luajit-launcher/pull/285
https://github.com/koreader/android-luajit-launcher/pull/286
https://github.com/koreader/android-luajit-launcher/pull/287
2021-01-10 00:51:09 +00:00
function UIManager : isSubwidgetShown ( widget , max_depth )
for i = # self._window_stack , 1 , - 1 do
local matched , depth = util.arrayReferences ( self._window_stack [ i ] . widget , widget , max_depth )
if matched then
return matched , depth , self._window_stack [ i ] . widget
end
end
return false
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--- Same as `isSubwidgetShown`, but only check window-level widgets (e.g., what's directly registered in the window stack), don't recurse.
Revamp "flash_ui" handling (#7118)
* Wherever possible, do an actual dumb invert on the Screen BB instead of repainting the widget, *then* inverting it (which is what the "invert" flag does).
* Instead of playing with nextTick/tickAfterNext delays, explicitly fence stuff with forceRePaint
* And, in the few cases where said Mk. 7 quirk kicks in, make the fences more marked by using a well-placed WAIT_FOR_UPDATE_COMPLETE
* Fix an issue in Button where show/hide & enable/disable where actually all toggles, which meant that duplicate calls or timing issues would do the wrong thing. (This broke dimming some icons, and mistakenly dropped the background from FM chevrons, for example).
* Speaking of, fix Button's hide/show to actually restore the background properly (there was a stupid typo in the variable name)
* Still in Button, fix the insanity of the double repaint on rounded buttons. Turns out it made sense, after all (and was related to said missing background, and bad interaction with invert & text with no background).
* KeyValuePage suffered from a similar issue with broken highlights (all black) because of missing backgrounds.
* In ConfigDialog, only instanciate IconButtons once (because every tab switch causes a full instantiation; and the initial display implies a full instanciation and an initial tab switch). Otherwise, both instances linger, and catch taps, and as such, double highlights.
* ConfigDialog: Restore the "don't repaint ReaderUI" when switching between similarly sized tabs (re #6131). I never could reproduce that on eInk, and I can't now on the emulator, so I'm assuming @poire-z fixed it during the swap to SVG icons.
* KeyValuePage: Only instanciate Buttons once (again, this is a widget that goes through a full init every page). Again, caused highlight/dimming issues because buttons were stacked.
* Menu: Ditto.
* TouchMenu: Now home of the gnarliest unhilight heuristics, because of the sheer amount of different things that can happen (and/or thanks to stuff not flagged covers_fullscreen properly ;p).
* Bump base
https://github.com/koreader/koreader-base/pull/1280
https://github.com/koreader/koreader-base/pull/1282
https://github.com/koreader/koreader-base/pull/1283
https://github.com/koreader/koreader-base/pull/1284
* Bump android-luajit-launcher
https://github.com/koreader/android-luajit-launcher/pull/284
https://github.com/koreader/android-luajit-launcher/pull/285
https://github.com/koreader/android-luajit-launcher/pull/286
https://github.com/koreader/android-luajit-launcher/pull/287
2021-01-10 00:51:09 +00:00
function UIManager : isWidgetShown ( widget )
for i = # self._window_stack , 1 , - 1 do
if self._window_stack [ i ] . widget == widget then
return true
end
end
return false
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Returns the region of the previous refresh .
@ return a rectangle @ { ui.geometry . Geom | Geom } object
] ]
Revamp "flash_ui" handling (#7118)
* Wherever possible, do an actual dumb invert on the Screen BB instead of repainting the widget, *then* inverting it (which is what the "invert" flag does).
* Instead of playing with nextTick/tickAfterNext delays, explicitly fence stuff with forceRePaint
* And, in the few cases where said Mk. 7 quirk kicks in, make the fences more marked by using a well-placed WAIT_FOR_UPDATE_COMPLETE
* Fix an issue in Button where show/hide & enable/disable where actually all toggles, which meant that duplicate calls or timing issues would do the wrong thing. (This broke dimming some icons, and mistakenly dropped the background from FM chevrons, for example).
* Speaking of, fix Button's hide/show to actually restore the background properly (there was a stupid typo in the variable name)
* Still in Button, fix the insanity of the double repaint on rounded buttons. Turns out it made sense, after all (and was related to said missing background, and bad interaction with invert & text with no background).
* KeyValuePage suffered from a similar issue with broken highlights (all black) because of missing backgrounds.
* In ConfigDialog, only instanciate IconButtons once (because every tab switch causes a full instantiation; and the initial display implies a full instanciation and an initial tab switch). Otherwise, both instances linger, and catch taps, and as such, double highlights.
* ConfigDialog: Restore the "don't repaint ReaderUI" when switching between similarly sized tabs (re #6131). I never could reproduce that on eInk, and I can't now on the emulator, so I'm assuming @poire-z fixed it during the swap to SVG icons.
* KeyValuePage: Only instanciate Buttons once (again, this is a widget that goes through a full init every page). Again, caused highlight/dimming issues because buttons were stacked.
* Menu: Ditto.
* TouchMenu: Now home of the gnarliest unhilight heuristics, because of the sheer amount of different things that can happen (and/or thanks to stuff not flagged covers_fullscreen properly ;p).
* Bump base
https://github.com/koreader/koreader-base/pull/1280
https://github.com/koreader/koreader-base/pull/1282
https://github.com/koreader/koreader-base/pull/1283
https://github.com/koreader/koreader-base/pull/1284
* Bump android-luajit-launcher
https://github.com/koreader/android-luajit-launcher/pull/284
https://github.com/koreader/android-luajit-launcher/pull/285
https://github.com/koreader/android-luajit-launcher/pull/286
https://github.com/koreader/android-luajit-launcher/pull/287
2021-01-10 00:51:09 +00:00
function UIManager : getPreviousRefreshRegion ( )
return self._last_refresh_region
2019-09-01 19:30:19 +00:00
end
2017-04-09 13:35:01 +00:00
--- Signals to quit.
2012-04-22 19:29:48 +00:00
function UIManager : quit ( )
2017-05-16 09:11:11 +00:00
if not self._running then return end
2017-04-09 13:35:01 +00:00
logger.info ( " quitting uimanager " )
2015-12-26 00:48:40 +00:00
self._task_queue_dirty = false
2014-03-13 13:52:43 +00:00
self._running = false
2015-03-02 09:15:26 +00:00
self._run_forever = nil
2014-09-03 03:13:46 +00:00
for i = # self._window_stack , 1 , - 1 do
table.remove ( self._window_stack , i )
end
2015-12-26 00:48:40 +00:00
for i = # self._task_queue , 1 , - 1 do
table.remove ( self._task_queue , i )
2014-09-03 03:13:46 +00:00
end
2014-06-23 14:33:11 +00:00
for i = # self._zeromqs , 1 , - 1 do
self._zeromqs [ i ] : stop ( )
table.remove ( self._zeromqs , i )
end
2015-03-02 09:15:26 +00:00
if self.looper then
self.looper : close ( )
self.looper = nil
end
2012-04-22 19:29:48 +00:00
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Request all @ { ui.event . Event | Event } s to be ignored for some duration .
@ param set_or_seconds either ` true ` , in which case a platform - specific delay is chosen , or a duration in seconds ( *** int *** ) .
] ]
2021-01-01 13:34:55 +00:00
function UIManager : discardEvents ( set_or_seconds )
if not set_or_seconds then -- remove any previously set
self._discard_events_till = nil
return
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
local delay
2021-01-01 13:34:55 +00:00
if set_or_seconds == true then
-- Use an adequate delay to account for device refresh duration
-- so any events happening in this delay (ie. before a widget
-- is really painted on screen) are discarded.
if Device : hasEinkScreen ( ) then
-- A screen refresh can take a few 100ms,
2021-01-07 17:05:18 +00:00
-- sometimes > 500ms on some devices/temperatures.
-- So, block for 400ms (to have it displayed) + 400ms
-- for user reaction to it
2021-04-14 23:00:28 +00:00
delay = TimeVal : new { sec = 0 , usec = 800000 }
2021-01-01 13:34:55 +00:00
else
2021-01-07 17:05:18 +00:00
-- On non-eInk screen, display is usually instantaneous
2021-04-14 23:00:28 +00:00
delay = TimeVal : new { sec = 0 , usec = 400000 }
2021-01-01 13:34:55 +00:00
end
else -- we expect a number
2021-04-14 23:00:28 +00:00
delay = TimeVal : new { sec = set_or_seconds , usec = 0 }
2021-01-01 13:34:55 +00:00
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
self._discard_events_till = self._now + delay
2021-01-01 13:34:55 +00:00
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
ReaderUI: Saner FM/RD lifecycle
* Ensure that going from one to the other tears down the former and
its plugins before instantiating the latter and its plugins.
UIManager: Unify Event sending & broadcasting
* Make the two behave the same way (walk the widget stack from top to
bottom), and properly handle the window stack shrinking shrinking
*and* growing.
Previously, broadcasting happened bottom-to-top and didn't really
handle the list shrinking/growing, while sending only handled the list
shrinking by a single element, and hopefully that element being the one
the event was just sent to.
These two items combined allowed us to optimize suboptimal
refresh behavior with Menu and other Menu classes when
opening/closing a document.
e.g., the "opening document" Notification is now properly regional,
and the "open last doc" option no longer flashes like a crazy person
anymore.
Plugins: Allow optimizing Menu refresh with custom menus, too.
Requires moving Menu's close_callback *after* onMenuSelect, which, eh,
probably makes sense, and is probably harmless in the grand scheme of
things.
2021-05-01 16:53:04 +00:00
Transmits an @ { ui.event . Event | Event } to active widgets , top to bottom .
Stops at the first handler that returns ` true ` .
Note that most complex widgets are based on @ { ui.widget . container.WidgetContainer | WidgetContainer } ,
which itself will take care of propagating an event to its members .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
@ param event an @ { ui.event . Event | Event } object
] ]
2012-04-22 19:29:48 +00:00
function UIManager : sendEvent ( event )
2014-06-07 05:36:58 +00:00
if # self._window_stack == 0 then return end
2016-05-30 04:57:27 +00:00
2021-01-01 13:34:55 +00:00
-- Ensure discardEvents
if self._discard_events_till then
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
if TimeVal : now ( ) < self._discard_events_till then
2021-01-01 13:34:55 +00:00
return
else
self._discard_events_till = nil
end
end
2021-01-11 17:14:12 +00:00
-- The top widget gets to be the first to get the event
2016-05-30 04:57:27 +00:00
local top_widget = self._window_stack [ # self._window_stack ]
2021-01-11 17:14:12 +00:00
-- A toast widget gets closed by any event, and
-- lets the event be handled by a lower widget
-- (Notification is our single widget with toast=true)
while top_widget.widget . toast do -- close them all
self : close ( top_widget.widget )
if # self._window_stack == 0 then return end
top_widget = self._window_stack [ # self._window_stack ]
end
2016-05-30 04:57:27 +00:00
if top_widget.widget : handleEvent ( event ) then
2014-03-13 13:52:43 +00:00
return
end
2016-05-30 04:57:27 +00:00
if top_widget.widget . active_widgets then
for _ , active_widget in ipairs ( top_widget.widget . active_widgets ) do
if active_widget : handleEvent ( event ) then return end
2014-03-13 13:52:43 +00:00
end
2016-05-30 04:57:27 +00:00
end
ReaderUI: Saner FM/RD lifecycle
* Ensure that going from one to the other tears down the former and
its plugins before instantiating the latter and its plugins.
UIManager: Unify Event sending & broadcasting
* Make the two behave the same way (walk the widget stack from top to
bottom), and properly handle the window stack shrinking shrinking
*and* growing.
Previously, broadcasting happened bottom-to-top and didn't really
handle the list shrinking/growing, while sending only handled the list
shrinking by a single element, and hopefully that element being the one
the event was just sent to.
These two items combined allowed us to optimize suboptimal
refresh behavior with Menu and other Menu classes when
opening/closing a document.
e.g., the "opening document" Notification is now properly regional,
and the "open last doc" option no longer flashes like a crazy person
anymore.
Plugins: Allow optimizing Menu refresh with custom menus, too.
Requires moving Menu's close_callback *after* onMenuSelect, which, eh,
probably makes sense, and is probably harmless in the grand scheme of
things.
2021-05-01 16:53:04 +00:00
-- If the event was not consumed (no handler returned true), active widgets (from top to bottom) can access it.
-- NOTE: _window_stack can shrink/grow when widgets are closed (CloseWidget & Close events) or opened.
-- Simply looping in reverse would only cover the list shrinking, and that only by a *single* element,
-- something we can't really guarantee, hence the more dogged iterator below,
-- which relies on a hash check of already processed widgets (LuaJIT actually hashes the table's GC reference),
-- rather than a simple loop counter, and will in fact iterate *at least* #items ^ 2 times.
-- Thankfully, that list should be very small, so the overhead should be minimal.
2016-05-30 04:57:27 +00:00
local checked_widgets = { top_widget }
ReaderUI: Saner FM/RD lifecycle
* Ensure that going from one to the other tears down the former and
its plugins before instantiating the latter and its plugins.
UIManager: Unify Event sending & broadcasting
* Make the two behave the same way (walk the widget stack from top to
bottom), and properly handle the window stack shrinking shrinking
*and* growing.
Previously, broadcasting happened bottom-to-top and didn't really
handle the list shrinking/growing, while sending only handled the list
shrinking by a single element, and hopefully that element being the one
the event was just sent to.
These two items combined allowed us to optimize suboptimal
refresh behavior with Menu and other Menu classes when
opening/closing a document.
e.g., the "opening document" Notification is now properly regional,
and the "open last doc" option no longer flashes like a crazy person
anymore.
Plugins: Allow optimizing Menu refresh with custom menus, too.
Requires moving Menu's close_callback *after* onMenuSelect, which, eh,
probably makes sense, and is probably harmless in the grand scheme of
things.
2021-05-01 16:53:04 +00:00
local i = # self._window_stack
while i > 0 do
2016-05-30 04:57:27 +00:00
local widget = self._window_stack [ i ]
if checked_widgets [ widget ] == nil then
ReaderUI: Saner FM/RD lifecycle
* Ensure that going from one to the other tears down the former and
its plugins before instantiating the latter and its plugins.
UIManager: Unify Event sending & broadcasting
* Make the two behave the same way (walk the widget stack from top to
bottom), and properly handle the window stack shrinking shrinking
*and* growing.
Previously, broadcasting happened bottom-to-top and didn't really
handle the list shrinking/growing, while sending only handled the list
shrinking by a single element, and hopefully that element being the one
the event was just sent to.
These two items combined allowed us to optimize suboptimal
refresh behavior with Menu and other Menu classes when
opening/closing a document.
e.g., the "opening document" Notification is now properly regional,
and the "open last doc" option no longer flashes like a crazy person
anymore.
Plugins: Allow optimizing Menu refresh with custom menus, too.
Requires moving Menu's close_callback *after* onMenuSelect, which, eh,
probably makes sense, and is probably harmless in the grand scheme of
things.
2021-05-01 16:53:04 +00:00
checked_widgets [ widget ] = true
-- Widget's active widgets have precedence to handle this event
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- NOTE: While FileManager only has a single (screenshotter), ReaderUI has many active_widgets (each ReaderUI module gets added to the list).
2016-05-30 04:57:27 +00:00
if widget.widget . active_widgets then
for _ , active_widget in ipairs ( widget.widget . active_widgets ) do
if active_widget : handleEvent ( event ) then return end
end
2014-03-13 13:52:43 +00:00
end
2019-11-06 20:02:27 +00:00
if widget.widget . is_always_active then
ReaderUI: Saner FM/RD lifecycle
* Ensure that going from one to the other tears down the former and
its plugins before instantiating the latter and its plugins.
UIManager: Unify Event sending & broadcasting
* Make the two behave the same way (walk the widget stack from top to
bottom), and properly handle the window stack shrinking shrinking
*and* growing.
Previously, broadcasting happened bottom-to-top and didn't really
handle the list shrinking/growing, while sending only handled the list
shrinking by a single element, and hopefully that element being the one
the event was just sent to.
These two items combined allowed us to optimize suboptimal
refresh behavior with Menu and other Menu classes when
opening/closing a document.
e.g., the "opening document" Notification is now properly regional,
and the "open last doc" option no longer flashes like a crazy person
anymore.
Plugins: Allow optimizing Menu refresh with custom menus, too.
Requires moving Menu's close_callback *after* onMenuSelect, which, eh,
probably makes sense, and is probably harmless in the grand scheme of
things.
2021-05-01 16:53:04 +00:00
-- Widget itself is flagged always active, let it handle the event
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- NOTE: is_always_active widgets currently are widgets that want to show a VirtualKeyboard or listen to Dispatcher events
2016-08-18 15:47:51 +00:00
if widget.widget : handleEvent ( event ) then return end
end
ReaderUI: Saner FM/RD lifecycle
* Ensure that going from one to the other tears down the former and
its plugins before instantiating the latter and its plugins.
UIManager: Unify Event sending & broadcasting
* Make the two behave the same way (walk the widget stack from top to
bottom), and properly handle the window stack shrinking shrinking
*and* growing.
Previously, broadcasting happened bottom-to-top and didn't really
handle the list shrinking/growing, while sending only handled the list
shrinking by a single element, and hopefully that element being the one
the event was just sent to.
These two items combined allowed us to optimize suboptimal
refresh behavior with Menu and other Menu classes when
opening/closing a document.
e.g., the "opening document" Notification is now properly regional,
and the "open last doc" option no longer flashes like a crazy person
anymore.
Plugins: Allow optimizing Menu refresh with custom menus, too.
Requires moving Menu's close_callback *after* onMenuSelect, which, eh,
probably makes sense, and is probably harmless in the grand scheme of
things.
2021-05-01 16:53:04 +00:00
i = # self._window_stack
else
i = i - 1
2014-03-13 13:52:43 +00:00
end
end
2012-04-22 19:29:48 +00:00
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Transmits an @ { ui.event . Event | Event } to all registered widgets .
@ param event an @ { ui.event . Event | Event } object
] ]
2016-05-01 12:46:17 +00:00
function UIManager : broadcastEvent ( event )
ReaderUI: Saner FM/RD lifecycle
* Ensure that going from one to the other tears down the former and
its plugins before instantiating the latter and its plugins.
UIManager: Unify Event sending & broadcasting
* Make the two behave the same way (walk the widget stack from top to
bottom), and properly handle the window stack shrinking shrinking
*and* growing.
Previously, broadcasting happened bottom-to-top and didn't really
handle the list shrinking/growing, while sending only handled the list
shrinking by a single element, and hopefully that element being the one
the event was just sent to.
These two items combined allowed us to optimize suboptimal
refresh behavior with Menu and other Menu classes when
opening/closing a document.
e.g., the "opening document" Notification is now properly regional,
and the "open last doc" option no longer flashes like a crazy person
anymore.
Plugins: Allow optimizing Menu refresh with custom menus, too.
Requires moving Menu's close_callback *after* onMenuSelect, which, eh,
probably makes sense, and is probably harmless in the grand scheme of
things.
2021-05-01 16:53:04 +00:00
-- Unlike sendEvent, we send the event to *all* (window-level) widgets (i.e., we don't stop, even if a handler returns true).
-- NOTE: Same defensive approach to _window_stack changing from under our feet as above.
local checked_widgets = { }
local i = # self._window_stack
while i > 0 do
local widget = self._window_stack [ i ]
if checked_widgets [ widget ] == nil then
checked_widgets [ widget ] = true
widget.widget : handleEvent ( event )
i = # self._window_stack
else
i = i - 1
2016-05-01 12:46:17 +00:00
end
end
end
2014-11-28 20:12:54 +00:00
function UIManager : _checkTasks ( )
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
self._now = TimeVal : now ( )
2014-03-13 13:52:43 +00:00
local wait_until = nil
2015-12-26 00:48:40 +00:00
2016-03-18 23:33:14 +00:00
-- task.action may schedule other events
self._task_queue_dirty = false
2015-12-26 00:48:40 +00:00
while true do
2021-08-19 16:14:19 +00:00
if # self._task_queue == 0 then
-- Nothing to do!
2015-12-26 00:48:40 +00:00
break
end
2021-08-19 16:14:19 +00:00
local next_task = self._task_queue [ 1 ]
local task_tv = next_task.time or TimeVal.zero
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
if task_tv <= self._now then
2015-12-26 00:48:40 +00:00
-- remove from table
2021-08-19 16:14:19 +00:00
local task = table.remove ( self._task_queue , 1 )
2015-12-26 00:48:40 +00:00
-- task is pending to be executed right now. do it.
-- NOTE: be careful that task.action() might modify
-- _task_queue here. So need to avoid race condition
2021-02-22 01:09:44 +00:00
task.action ( unpack ( task.args , 1 , task.argc ) )
2015-12-26 00:48:40 +00:00
else
-- queue is sorted in ascendant order, safe to assume all items
-- are future tasks for now
2021-08-19 16:14:19 +00:00
wait_until = next_task.time
2015-12-26 00:48:40 +00:00
break
end
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
return wait_until , self._now
end
--[[--
Returns a TimeVal object corresponding to the last UI tick .
This is essentially a cached TimeVal : now ( ) , computed at the top of every iteration of the main UI loop ,
( right before checking / running scheduled tasks ) .
This is mainly useful to compute / schedule stuff in the same time scale as the UI loop ( i.e . , MONOTONIC ) ,
without having to resort to a syscall .
It should never be significantly stale ( i.e . , it should be precise enough ) ,
unless you ' re blocking the UI for a significant amount of time in the same UI tick.
Prefer the appropriate TimeVal method for your needs if you require perfect accuracy
( e.g . , when you ' re actually working on the event loop *itself* (UIManager, Input, GestureDetector)).
This is * NOT * wall clock time ( REALTIME ) .
] ]
function UIManager : getTime ( )
return self._now
2013-01-05 14:28:14 +00:00
end
2014-11-28 20:12:54 +00:00
-- precedence of refresh modes:
2018-06-02 16:10:55 +00:00
local refresh_modes = { fast = 1 , ui = 2 , partial = 3 , flashui = 4 , flashpartial = 5 , full = 6 }
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
-- NOTE: We might want to introduce a "force_fast" that points to fast, but has the highest priority,
-- for the few cases where we might *really* want to enforce fast (for stuff like panning or skimming?).
2014-11-28 20:12:54 +00:00
-- refresh methods in framebuffer implementation
local refresh_methods = {
fast = " refreshFast " ,
ui = " refreshUI " ,
partial = " refreshPartial " ,
2018-06-02 16:10:55 +00:00
flashui = " refreshFlashUI " ,
flashpartial = " refreshFlashPartial " ,
2014-11-28 20:12:54 +00:00
full = " refreshFull " ,
}
--[[
2017-04-09 13:35:01 +00:00
Compares refresh mode .
2014-11-28 20:12:54 +00:00
2017-04-09 13:35:01 +00:00
Will return the mode that takes precedence .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
] ]
2014-11-28 20:12:54 +00:00
local function update_mode ( mode1 , mode2 )
if refresh_modes [ mode1 ] > refresh_modes [ mode2 ] then
2018-06-02 16:10:55 +00:00
logger.dbg ( " update_mode: Update refresh mode " , mode2 , " to " , mode1 )
2014-11-28 20:12:54 +00:00
return mode1
else
return mode2
end
end
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
--[[
Compares dither hints .
Dither always wins .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
] ]
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
local function update_dither ( dither1 , dither2 )
if dither1 and not dither2 then
logger.dbg ( " update_dither: Update dither hint " , dither2 , " to " , dither1 )
return dither1
else
return dither2
end
end
2017-04-09 13:35:01 +00:00
--[[--
Enqueues a refresh .
2014-11-28 20:12:54 +00:00
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
Widgets call this in their ` paintTo ( ) ` method in order to notify
2014-11-28 20:12:54 +00:00
UIManager that a certain part of the screen is to be refreshed .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
@ string mode
refresh mode ( ` " full " ` , ` " flashpartial " ` , ` " flashui " ` , ` " partial " ` , ` " ui " ` , ` " fast " ` )
2017-04-09 13:35:01 +00:00
@ param region
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
A rectangle @ { ui.geometry . Geom | Geom } object that specifies the region to be updated .
Optional , update will affect whole screen if not specified .
2017-04-09 13:35:01 +00:00
Note that this should be the exception .
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
@ bool dither
A hint to request hardware dithering ( if supported ) .
Optional , no dithering requested if not specified or not supported .
@ local Not to be used outside of UIManager !
] ]
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
function UIManager : _refresh ( mode , region , dither )
if not mode then
-- If we're trying to float a dither hint up from a lower widget after a close, mode might be nil...
-- So use the lowest priority refresh mode (short of fast, because that'd do half-toning).
if dither then
mode = " ui "
else
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- Otherwise, this is most likely from a `show` or `close` that wasn't passed specific refresh details,
-- (which is the vast majority of them), in which case we drop it to avoid enqueuing a useless full-screen refresh.
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
return
end
end
2021-05-19 20:57:54 +00:00
-- Downgrade all refreshes to "fast" when ReaderPaging or ReaderScrolling have set this flag
if self.currently_scrolling then
mode = " fast "
end
2016-12-04 14:17:04 +00:00
if not region and mode == " full " then
self.refresh_count = 0 -- reset counter on explicit full refresh
end
2018-06-02 16:10:55 +00:00
-- Handle downgrading flashing modes to non-flashing modes, according to user settings.
-- NOTE: Do it before "full" promotion and collision checks/update_mode.
if G_reader_settings : isTrue ( " avoid_flashing_ui " ) then
if mode == " flashui " then
mode = " ui "
logger.dbg ( " _refresh: downgraded flashui refresh to " , mode )
elseif mode == " flashpartial " then
mode = " partial "
logger.dbg ( " _refresh: downgraded flashpartial refresh to " , mode )
elseif mode == " partial " and region then
mode = " ui "
logger.dbg ( " _refresh: downgraded regional partial refresh to " , mode )
end
end
-- special case: "partial" refreshes
-- will get promoted every self.FULL_REFRESH_COUNT refreshes
2014-12-04 07:46:47 +00:00
-- since _refresh can be called mutiple times via setDirty called in
2018-06-02 16:10:55 +00:00
-- different widgets before a real screen repaint, we should make sure
2014-12-04 07:46:47 +00:00
-- refresh_count is incremented by only once at most for each repaint
2018-06-02 16:10:55 +00:00
-- NOTE: Ideally, we'd only check for "partial"" w/ no region set (that neatly narrows it down to just the reader).
-- In practice, we also want to promote refreshes in a few other places, except purely text-poor UI elements.
-- (Putting "ui" in that list is problematic with a number of UI elements, most notably, ReaderHighlight,
-- because it is implemented as "ui" over the full viewport, since we can't devise a proper bounding box).
-- So we settle for only "partial", but treating full-screen ones slightly differently.
if mode == " partial " and not self.refresh_counted then
2014-11-28 20:12:54 +00:00
self.refresh_count = ( self.refresh_count + 1 ) % self.FULL_REFRESH_COUNT
if self.refresh_count == self.FULL_REFRESH_COUNT - 1 then
2018-06-02 16:10:55 +00:00
-- NOTE: Promote to "full" if no region (reader), to "flashui" otherwise (UI)
if region then
mode = " flashui "
else
mode = " full "
end
logger.dbg ( " _refresh: promote refresh to " , mode )
2014-11-28 20:12:54 +00:00
end
2014-12-04 07:46:47 +00:00
self.refresh_counted = true
2014-11-28 20:12:54 +00:00
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- if no region is specified, use the screen's dimensions
2014-11-28 20:12:54 +00:00
region = region or Geom : new { w = Screen : getWidth ( ) , h = Screen : getHeight ( ) }
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
-- if no dithering hint was specified, don't request dithering
dither = dither or false
2018-06-02 16:10:55 +00:00
-- NOTE: While, ideally, we shouldn't merge refreshes w/ different waveform modes,
-- this allows us to optimize away a number of quirks of our rendering stack
2020-02-03 19:08:18 +00:00
-- (e.g., multiple setDirty calls queued when showing/closing a widget because of update mechanisms),
2018-06-02 16:10:55 +00:00
-- as well as a few actually effective merges
2020-02-03 19:08:18 +00:00
-- (e.g., the disappearance of a selection HL with the following menu update).
2014-11-28 20:12:54 +00:00
for i = 1 , # self._refresh_stack do
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- Check for collision with refreshes that are already enqueued
-- NOTE: intersect *means* intersect: we won't merge edge-to-edge regions (but the EPDC probably will).
2014-11-28 20:12:54 +00:00
if region : intersectWith ( self._refresh_stack [ i ] . region ) then
-- combine both refreshes' regions
local combined = region : combine ( self._refresh_stack [ i ] . region )
-- update the mode, if needed
2016-02-16 07:10:07 +00:00
mode = update_mode ( mode , self._refresh_stack [ i ] . mode )
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
-- dithering hints are viral, one is enough to infect the whole queue
dither = update_dither ( dither , self._refresh_stack [ i ] . dither )
2018-06-02 16:10:55 +00:00
-- remove colliding refresh
2014-11-28 20:12:54 +00:00
table.remove ( self._refresh_stack , i )
-- and try again with combined data
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
return self : _refresh ( mode , combined , dither )
2014-11-28 20:12:54 +00:00
end
end
2018-06-02 16:10:55 +00:00
-- if we've stopped hitting collisions, enqueue the refresh
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
logger.dbg ( " _refresh: Enqueued " , mode , " update for region " , region.x , region.y , region.w , region.h , dither and " w/ HW dithering " or " " )
table.insert ( self._refresh_stack , { mode = mode , region = region , dither = dither } )
2014-11-28 20:12:54 +00:00
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Repaints dirty widgets .
This will also drain the refresh queue , effectively refreshing the screen region ( s ) matching those freshly repainted widgets .
There may be refreshes enqueued without any widgets needing to be repainted ( c.f . , ` setDirty ` ' s behavior when passed a `nil` widget),
in which case , nothing is repainted , but the refreshes are still drained and executed .
2020-06-14 00:21:41 +00:00
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
@ local Not to be used outside of UIManager !
--]]
2014-11-28 20:12:54 +00:00
function UIManager : _repaint ( )
2014-11-20 19:24:02 +00:00
-- flag in which we will record if we did any repaints at all
-- will trigger a refresh if set.
2014-11-14 08:09:50 +00:00
local dirty = false
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
-- remember if any of our repaints were dithered
local dithered = false
2014-11-20 19:24:02 +00:00
2018-03-17 22:02:32 +00:00
-- We don't need to call paintTo() on widgets that are under
-- a widget that covers the full screen
local start_idx = 1
for i = # self._window_stack , 1 , - 1 do
if self._window_stack [ i ] . widget.covers_fullscreen then
start_idx = i
if i > 1 then
logger.dbg ( " not painting " , i - 1 , " covered widget(s) " )
end
break
end
end
2020-09-12 20:57:53 +00:00
-- Show IDs of covered widgets when debugging
--[[
if start_idx > 1 then
for i = 1 , start_idx - 1 do
local widget = self._window_stack [ i ]
logger.dbg ( " NOT painting widget: " , widget.widget . name or widget.widget . id or tostring ( widget ) )
end
end
--]]
2018-03-17 22:02:32 +00:00
for i = start_idx , # self._window_stack do
local widget = self._window_stack [ i ]
2014-11-28 20:12:54 +00:00
-- paint if current widget or any widget underneath is dirty
if dirty or self._dirty [ widget.widget ] then
-- pass hint to widget that we got when setting widget dirty
-- the widget can use this to decide which parts should be refreshed
2018-03-17 22:02:32 +00:00
logger.dbg ( " painting widget: " , widget.widget . name or widget.widget . id or tostring ( widget ) )
2020-09-22 20:04:37 +00:00
Screen : beforePaint ( )
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- NOTE: Nothing actually seems to use the final argument?
-- Could be used by widgets to know whether they're being repainted because they're actually dirty (it's true),
-- or because something below them was (it's nil).
2014-11-28 20:12:54 +00:00
widget.widget : paintTo ( Screen.bb , widget.x , widget.y , self._dirty [ widget.widget ] )
2014-11-20 19:24:02 +00:00
2014-11-14 08:09:50 +00:00
-- and remove from list after painting
self._dirty [ widget.widget ] = nil
2014-11-20 19:24:02 +00:00
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- trigger a repaint for every widget above us, too
2014-11-14 08:09:50 +00:00
dirty = true
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
-- if any of 'em were dithered, we'll want to dither the final refresh
if widget.widget . dithered then
logger.dbg ( " _repaint: it was dithered, infecting the refresh queue " )
dithered = true
end
2014-11-14 08:09:50 +00:00
end
end
2014-11-28 20:12:54 +00:00
-- execute pending refresh functions
for _ , refreshfunc in ipairs ( self._refresh_func_stack ) do
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
local refreshtype , region , dither = refreshfunc ( )
-- honor dithering hints from *anywhere* in the dirty stack
dither = update_dither ( dither , dithered )
if refreshtype then self : _refresh ( refreshtype , region , dither ) end
2014-11-28 20:12:54 +00:00
end
self._refresh_func_stack = { }
2016-06-21 08:50:01 +00:00
-- we should have at least one refresh if we did repaint. If we don't, we
-- add one now and log a warning if we are debugging
2014-11-28 20:12:54 +00:00
if dirty and # self._refresh_stack == 0 then
2017-09-08 13:30:29 +00:00
logger.dbg ( " no refresh got enqueued. Will do a partial full screen refresh, which might be inefficient " )
2014-11-28 20:12:54 +00:00
self : _refresh ( " partial " )
end
2014-11-20 19:24:02 +00:00
2014-11-28 20:12:54 +00:00
-- execute refreshes:
for _ , refresh in ipairs ( self._refresh_stack ) do
2019-04-18 21:26:53 +00:00
-- Honor dithering hints from *anywhere* in the dirty stack
refresh.dither = update_dither ( refresh.dither , dithered )
-- If HW dithering is disabled, unconditionally drop the dither flag
if not Screen.hw_dithering then
refresh.dither = nil
end
2016-08-12 09:47:35 +00:00
dbg : v ( " triggering refresh " , refresh )
Revamp "flash_ui" handling (#7118)
* Wherever possible, do an actual dumb invert on the Screen BB instead of repainting the widget, *then* inverting it (which is what the "invert" flag does).
* Instead of playing with nextTick/tickAfterNext delays, explicitly fence stuff with forceRePaint
* And, in the few cases where said Mk. 7 quirk kicks in, make the fences more marked by using a well-placed WAIT_FOR_UPDATE_COMPLETE
* Fix an issue in Button where show/hide & enable/disable where actually all toggles, which meant that duplicate calls or timing issues would do the wrong thing. (This broke dimming some icons, and mistakenly dropped the background from FM chevrons, for example).
* Speaking of, fix Button's hide/show to actually restore the background properly (there was a stupid typo in the variable name)
* Still in Button, fix the insanity of the double repaint on rounded buttons. Turns out it made sense, after all (and was related to said missing background, and bad interaction with invert & text with no background).
* KeyValuePage suffered from a similar issue with broken highlights (all black) because of missing backgrounds.
* In ConfigDialog, only instanciate IconButtons once (because every tab switch causes a full instantiation; and the initial display implies a full instanciation and an initial tab switch). Otherwise, both instances linger, and catch taps, and as such, double highlights.
* ConfigDialog: Restore the "don't repaint ReaderUI" when switching between similarly sized tabs (re #6131). I never could reproduce that on eInk, and I can't now on the emulator, so I'm assuming @poire-z fixed it during the swap to SVG icons.
* KeyValuePage: Only instanciate Buttons once (again, this is a widget that goes through a full init every page). Again, caused highlight/dimming issues because buttons were stacked.
* Menu: Ditto.
* TouchMenu: Now home of the gnarliest unhilight heuristics, because of the sheer amount of different things that can happen (and/or thanks to stuff not flagged covers_fullscreen properly ;p).
* Bump base
https://github.com/koreader/koreader-base/pull/1280
https://github.com/koreader/koreader-base/pull/1282
https://github.com/koreader/koreader-base/pull/1283
https://github.com/koreader/koreader-base/pull/1284
* Bump android-luajit-launcher
https://github.com/koreader/android-luajit-launcher/pull/284
https://github.com/koreader/android-luajit-launcher/pull/285
https://github.com/koreader/android-luajit-launcher/pull/286
https://github.com/koreader/android-luajit-launcher/pull/287
2021-01-10 00:51:09 +00:00
-- Remember the refresh region
self._last_refresh_region = refresh.region
2014-11-28 20:12:54 +00:00
Screen [ refresh_methods [ refresh.mode ] ] ( Screen ,
2020-06-14 00:21:41 +00:00
refresh.region . x , refresh.region . y ,
refresh.region . w , refresh.region . h ,
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
refresh.dither )
2014-11-14 08:09:50 +00:00
end
2020-10-18 18:38:17 +00:00
-- Don't trigger afterPaint if we did not, in fact, paint anything
if dirty then
Screen : afterPaint ( )
end
2014-11-28 20:12:54 +00:00
self._refresh_stack = { }
2014-12-04 07:46:47 +00:00
self.refresh_counted = false
2014-11-14 08:09:50 +00:00
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--- Explicitly drain the paint & refresh queues *now*, instead of waiting for the next UI tick.
2016-04-24 04:09:50 +00:00
function UIManager : forceRePaint ( )
self : _repaint ( )
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Ask the EPDC to * block * until our previous refresh ioctl has completed .
This interacts sanely with the existing low - level handling of this in ` framebuffer_mxcfb `
( i.e . , it doesn ' t even try to wait for a marker that fb has already waited for, and vice-versa).
Will return immediately if it has already completed .
If the device isn ' t a Linux + MXCFB device, this is a NOP.
] ]
Revamp "flash_ui" handling (#7118)
* Wherever possible, do an actual dumb invert on the Screen BB instead of repainting the widget, *then* inverting it (which is what the "invert" flag does).
* Instead of playing with nextTick/tickAfterNext delays, explicitly fence stuff with forceRePaint
* And, in the few cases where said Mk. 7 quirk kicks in, make the fences more marked by using a well-placed WAIT_FOR_UPDATE_COMPLETE
* Fix an issue in Button where show/hide & enable/disable where actually all toggles, which meant that duplicate calls or timing issues would do the wrong thing. (This broke dimming some icons, and mistakenly dropped the background from FM chevrons, for example).
* Speaking of, fix Button's hide/show to actually restore the background properly (there was a stupid typo in the variable name)
* Still in Button, fix the insanity of the double repaint on rounded buttons. Turns out it made sense, after all (and was related to said missing background, and bad interaction with invert & text with no background).
* KeyValuePage suffered from a similar issue with broken highlights (all black) because of missing backgrounds.
* In ConfigDialog, only instanciate IconButtons once (because every tab switch causes a full instantiation; and the initial display implies a full instanciation and an initial tab switch). Otherwise, both instances linger, and catch taps, and as such, double highlights.
* ConfigDialog: Restore the "don't repaint ReaderUI" when switching between similarly sized tabs (re #6131). I never could reproduce that on eInk, and I can't now on the emulator, so I'm assuming @poire-z fixed it during the swap to SVG icons.
* KeyValuePage: Only instanciate Buttons once (again, this is a widget that goes through a full init every page). Again, caused highlight/dimming issues because buttons were stacked.
* Menu: Ditto.
* TouchMenu: Now home of the gnarliest unhilight heuristics, because of the sheer amount of different things that can happen (and/or thanks to stuff not flagged covers_fullscreen properly ;p).
* Bump base
https://github.com/koreader/koreader-base/pull/1280
https://github.com/koreader/koreader-base/pull/1282
https://github.com/koreader/koreader-base/pull/1283
https://github.com/koreader/koreader-base/pull/1284
* Bump android-luajit-launcher
https://github.com/koreader/android-luajit-launcher/pull/284
https://github.com/koreader/android-luajit-launcher/pull/285
https://github.com/koreader/android-luajit-launcher/pull/286
https://github.com/koreader/android-luajit-launcher/pull/287
2021-01-10 00:51:09 +00:00
function UIManager : waitForVSync ( )
Screen : refreshWaitForLast ( )
end
2021-02-22 01:09:44 +00:00
--[[--
Yield to the EPDC .
This is a dumb workaround for potential races with the EPDC when we request a refresh on a specific region ,
and then proceed to * write * to the framebuffer , in the same region , very , very , very soon after that .
This basically just puts ourselves to sleep for a very short amount of time , to let the kernel do its thing in peace .
@ int sleep_us Amount of time to sleep for ( in µ s ) . ( Optional , defaults to 1 ms ) .
] ]
function UIManager : yieldToEPDC ( sleep_us )
if Device : hasEinkScreen ( ) then
-- NOTE: Early empiric evidence suggests that going as low as 1ms is enough to do the trick.
-- Consider jumping to the jiffy resolution (100Hz/10ms) if it turns out it isn't ;).
ffiUtil.usleep ( sleep_us or 1000 )
end
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Used to repaint a specific sub - widget that isn ' t on the `_window_stack` itself.
Useful to avoid repainting a complex widget when we just want to invert an icon , for instance .
No safety checks on x & y * by design * . I want this to blow up if used wrong .
This is an explicit repaint * now * : it bypasses and ignores the paint queue ( unlike ` setDirty ` ) .
@ param widget a @ { ui.widget . widget | widget } object
@ int x left origin of widget ( in the Screen buffer , e.g . , ` widget.dimen . x ` )
@ int y top origin of widget ( in the Screen buffer , e.g . , ` widget.dimen . y ` )
] ]
Enable HW dithering in a few key places (#4541)
* Enable HW dithering on supported devices (Clara HD, Forma; Oasis 2, PW4)
* FileManager and co. (where appropriate, i.e., when covers are shown)
* Book Status
* Reader, where appropriate:
* CRe: on pages whith image content (for over 7.5% of the screen area, should hopefully leave stuff like bullet points or small scene breaks alone).
* Other engines: on user-request (in the gear tab of the bottom menu), via the new "Dithering" knob (will only appear on supported devices).
* ScreenSaver
* ImageViewer
* Minimize repaints when flash_ui is enabled (by, almost everywhere, only repainting the flashing element, and not the toplevel window which hosts it).
(The first pass of this involved fixing a few Button instances whose show_parent was wrong, in particular, chevrons in the FM & TopMenu).
* Hunted down a few redundant repaints (unneeded setDirty("all") calls),
either by switching the widget to nil when only a refresh was needed, and not a repaint,
or by passing the appropritate widget to setDirty.
(Note to self: Enable *verbose* debugging to catch broken setDirty calls via its post guard).
There were also a few instances of 'em right behind a widget close.
* Don't repaint the underlying widget when initially showing TopMenu & ConfigDialog.
We unfortunately do need to do it when switching tabs, because of their variable heights.
* On Kobo, disabled the extra and completely useless full refresh before suspend/reboot/poweroff, as well as on resume. No more double refreshes!
* Fix another debug guard in Kobo sysfs_light
* Switch ImageWidget & ImageViewer mostly to "ui" updates, which will be better suited to image content pretty much everywhere, REAGL or not.
PS: (Almost :100: commits! :D)
2019-02-07 00:14:37 +00:00
function UIManager : widgetRepaint ( widget , x , y )
if not widget then return end
logger.dbg ( " Explicit widgetRepaint: " , widget.name or widget.id or tostring ( widget ) , " @ ( " , x , " , " , y , " ) " )
widget : paintTo ( Screen.bb , x , y )
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Same idea as ` widgetRepaint ` , but does a simple ` bb : invertRect ` on the Screen buffer , without actually going through the widget ' s `paintTo` method.
@ param widget a @ { ui.widget . widget | widget } object
@ int x left origin of the rectangle to invert ( in the Screen buffer , e.g . , ` widget.dimen . x ` )
@ int y top origin of the rectangle ( in the Screen buffer , e.g . , ` widget.dimen . y ` )
@ int w width of the rectangle ( optional , will use ` widget.dimen . w ` like ` paintTo ` would if omitted )
@ int h height of the rectangle ( optional , will use ` widget.dimen . h ` like ` paintTo ` would if omitted )
@ see widgetRepaint
--]]
Revamp "flash_ui" handling (#7118)
* Wherever possible, do an actual dumb invert on the Screen BB instead of repainting the widget, *then* inverting it (which is what the "invert" flag does).
* Instead of playing with nextTick/tickAfterNext delays, explicitly fence stuff with forceRePaint
* And, in the few cases where said Mk. 7 quirk kicks in, make the fences more marked by using a well-placed WAIT_FOR_UPDATE_COMPLETE
* Fix an issue in Button where show/hide & enable/disable where actually all toggles, which meant that duplicate calls or timing issues would do the wrong thing. (This broke dimming some icons, and mistakenly dropped the background from FM chevrons, for example).
* Speaking of, fix Button's hide/show to actually restore the background properly (there was a stupid typo in the variable name)
* Still in Button, fix the insanity of the double repaint on rounded buttons. Turns out it made sense, after all (and was related to said missing background, and bad interaction with invert & text with no background).
* KeyValuePage suffered from a similar issue with broken highlights (all black) because of missing backgrounds.
* In ConfigDialog, only instanciate IconButtons once (because every tab switch causes a full instantiation; and the initial display implies a full instanciation and an initial tab switch). Otherwise, both instances linger, and catch taps, and as such, double highlights.
* ConfigDialog: Restore the "don't repaint ReaderUI" when switching between similarly sized tabs (re #6131). I never could reproduce that on eInk, and I can't now on the emulator, so I'm assuming @poire-z fixed it during the swap to SVG icons.
* KeyValuePage: Only instanciate Buttons once (again, this is a widget that goes through a full init every page). Again, caused highlight/dimming issues because buttons were stacked.
* Menu: Ditto.
* TouchMenu: Now home of the gnarliest unhilight heuristics, because of the sheer amount of different things that can happen (and/or thanks to stuff not flagged covers_fullscreen properly ;p).
* Bump base
https://github.com/koreader/koreader-base/pull/1280
https://github.com/koreader/koreader-base/pull/1282
https://github.com/koreader/koreader-base/pull/1283
https://github.com/koreader/koreader-base/pull/1284
* Bump android-luajit-launcher
https://github.com/koreader/android-luajit-launcher/pull/284
https://github.com/koreader/android-luajit-launcher/pull/285
https://github.com/koreader/android-luajit-launcher/pull/286
https://github.com/koreader/android-luajit-launcher/pull/287
2021-01-10 00:51:09 +00:00
function UIManager : widgetInvert ( widget , x , y , w , h )
if not widget then return end
logger.dbg ( " Explicit widgetInvert: " , widget.name or widget.id or tostring ( widget ) , " @ ( " , x , " , " , y , " ) " )
Screen.bb : invertRect ( x , y , w or widget.dimen . w , h or widget.dimen . h )
end
2015-03-21 05:18:34 +00:00
function UIManager : setInputTimeout ( timeout )
self.INPUT_TIMEOUT = timeout or 200 * 1000
end
function UIManager : resetInputTimeout ( )
self.INPUT_TIMEOUT = nil
end
2018-01-01 14:40:28 +00:00
function UIManager : handleInputEvent ( input_event )
if input_event.handler ~= " onInputError " then
self.event_hook : execute ( " InputEvent " , input_event )
end
local handler = self.event_handlers [ input_event ]
if handler then
handler ( input_event )
else
self.event_handlers [ " __default__ " ] ( input_event )
end
end
-- Process all pending events on all registered ZMQs.
function UIManager : processZMQs ( )
for _ , zeromq in ipairs ( self._zeromqs ) do
2021-04-02 23:48:35 +00:00
for input_event in zeromq.waitEvent , zeromq do
2018-01-01 14:40:28 +00:00
self : handleInputEvent ( input_event )
end
end
end
2015-03-02 09:15:26 +00:00
function UIManager : handleInput ( )
local wait_until , now
-- run this in a loop, so that paints can trigger events
-- that will be honored when calculating the time to wait
-- for input events:
repeat
wait_until , now = self : _checkTasks ( )
2016-03-29 06:37:15 +00:00
--dbg("---------------------------------------------------")
2020-09-12 20:57:53 +00:00
--dbg("wait_until", wait_until)
--dbg("now", now)
2016-03-29 06:37:15 +00:00
--dbg("exec stack", self._task_queue)
--dbg("window stack", self._window_stack)
--dbg("dirty stack", self._dirty)
--dbg("---------------------------------------------------")
2015-03-02 09:15:26 +00:00
-- stop when we have no window to show
if # self._window_stack == 0 and not self._run_forever then
2016-12-29 08:10:38 +00:00
logger.info ( " no dialog left to show " )
2015-03-02 09:15:26 +00:00
self : quit ( )
return nil
end
2012-04-22 19:29:48 +00:00
2015-03-02 09:15:26 +00:00
self : _repaint ( )
2015-12-26 00:48:40 +00:00
until not self._task_queue_dirty
2015-03-02 09:15:26 +00:00
Calibre: Minor QoL fixes (#7528)
* CalibreMetadata: Get rid of the now useless NULL-hunt: here, this was basically looking for `rapidjson.null` to replace them with... `rapidjson.null` :?. IIRC, that's a remnant of a quirk of the previous JSON parser (possibly even the previous, *previous* JSON parser ^^).
* CalibreSearch: Update the actually relevant NULL-hunt to make it explicit: replace JSON NULLs with Lua nils, instead of relying on an implementation detail of Lua-RapidJSON, because that detail just changed data type ;).
* UIManager: Make sure tasks scheduled during the final ZMQ callback are honored. e.g., the Calibre "Disconnect" handler. This happened to mostly work purely by chance before the event loop rework.
* Calibre: Restore a proper receiveCallback handler after receiving a book, in order not to break the "Disconnect" handler's state (and, well, get a working Disconnect handler, period ^^).
* Calibre: Unbreak metadata cache when it's initialized by a search (regression since #7159).
* Calibre: Handle UTC <-> local time conversions when checking the cache's timestamp against the Calibre metadata timestamp.
* Bump base (Unbreak CRe on Android, update RapidJSON)
2021-04-12 00:31:53 +00:00
-- NOTE: Compute deadline *before* processing ZMQs, in order to be able to catch tasks scheduled *during*
-- the final ZMQ callback.
-- This ensures that we get to honor a single ZMQ_TIMEOUT *after* the final ZMQ callback,
-- which gives us a chance for another iteration, meaning going through _checkTasks to catch said scheduled tasks.
2018-01-01 14:40:28 +00:00
-- Figure out how long to wait.
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
-- Ultimately, that'll be the earliest of INPUT_TIMEOUT, ZMQ_TIMEOUT or the next earliest scheduled task.
local deadline
2018-01-01 14:40:28 +00:00
-- Default to INPUT_TIMEOUT (which may be nil, i.e. block until an event happens).
local wait_us = self.INPUT_TIMEOUT
-- If we have any ZMQs registered, ZMQ_TIMEOUT is another upper bound.
if # self._zeromqs > 0 then
wait_us = math.min ( wait_us or math.huge , self.ZMQ_TIMEOUT )
2015-03-02 09:15:26 +00:00
end
2012-04-22 19:29:48 +00:00
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
-- We pass that on as an absolute deadline, not a relative wait time.
if wait_us then
deadline = now + TimeVal : new { usec = wait_us }
end
-- If there's a scheduled task pending, that puts an upper bound on how long to wait.
if wait_until and ( not deadline or wait_until < deadline ) then
-- ^ We don't have a TIMEOUT induced deadline, making the choice easy.
-- ^ We have a task scheduled for *before* our TIMEOUT induced deadline.
deadline = wait_until
end
Calibre: Minor QoL fixes (#7528)
* CalibreMetadata: Get rid of the now useless NULL-hunt: here, this was basically looking for `rapidjson.null` to replace them with... `rapidjson.null` :?. IIRC, that's a remnant of a quirk of the previous JSON parser (possibly even the previous, *previous* JSON parser ^^).
* CalibreSearch: Update the actually relevant NULL-hunt to make it explicit: replace JSON NULLs with Lua nils, instead of relying on an implementation detail of Lua-RapidJSON, because that detail just changed data type ;).
* UIManager: Make sure tasks scheduled during the final ZMQ callback are honored. e.g., the Calibre "Disconnect" handler. This happened to mostly work purely by chance before the event loop rework.
* Calibre: Restore a proper receiveCallback handler after receiving a book, in order not to break the "Disconnect" handler's state (and, well, get a working Disconnect handler, period ^^).
* Calibre: Unbreak metadata cache when it's initialized by a search (regression since #7159).
* Calibre: Handle UTC <-> local time conversions when checking the cache's timestamp against the Calibre metadata timestamp.
* Bump base (Unbreak CRe on Android, update RapidJSON)
2021-04-12 00:31:53 +00:00
-- Run ZMQs if any
self : processZMQs ( )
2020-08-30 13:01:56 +00:00
-- If allowed, entering standby (from which we can wake by input) must trigger in response to event
-- this function emits (plugin), or within waitEvent() right after (hardware).
-- Anywhere else breaks preventStandby/allowStandby invariants used by background jobs while UI is left running.
self : _standbyTransition ( )
2021-04-02 23:48:35 +00:00
-- wait for next batch of events
local input_events = Input : waitEvent ( now , deadline )
2018-01-01 14:40:28 +00:00
2021-04-02 23:48:35 +00:00
-- delegate each input event to handler
if input_events then
-- Handle the full batch of events
for __ , ev in ipairs ( input_events ) do
self : handleInputEvent ( ev )
end
2014-03-13 13:52:43 +00:00
end
2015-03-02 09:15:26 +00:00
if self.looper then
2016-12-29 08:10:38 +00:00
logger.info ( " handle input in turbo I/O looper " )
2015-03-21 05:18:34 +00:00
self.looper : add_callback ( function ( )
2019-08-23 17:53:53 +00:00
--- @fixme Force close looper when there is unhandled error,
2015-03-21 05:18:34 +00:00
-- otherwise the looper will hang. Any better solution?
xpcall ( function ( ) self : handleInput ( ) end , function ( err )
io.stderr : write ( err .. " \n " )
io.stderr : write ( debug.traceback ( ) .. " \n " )
io.stderr : flush ( )
self.looper : close ( )
2020-12-25 15:19:16 +00:00
os.exit ( 1 , true )
2015-03-21 05:18:34 +00:00
end )
end )
2015-03-02 09:15:26 +00:00
end
end
2016-08-02 09:21:27 +00:00
function UIManager : onRotation ( )
2021-04-02 23:48:35 +00:00
self : setDirty ( " all " , " full " )
2016-08-02 09:21:27 +00:00
self : forceRePaint ( )
end
2015-03-02 09:15:26 +00:00
function UIManager : initLooper ( )
2015-06-02 05:07:17 +00:00
if DUSE_TURBO_LIB and not self.looper then
2016-02-16 17:58:24 +00:00
TURBO_SSL = true -- luacheck: ignore
__TURBO_USE_LUASOCKET__ = true -- luacheck: ignore
2015-03-02 09:15:26 +00:00
local turbo = require ( " turbo " )
self.looper = turbo.ioloop . instance ( )
end
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
This is the main loop of the UI controller .
It is intended to manage input events and delegate them to dialogs .
--]]
2015-03-02 09:15:26 +00:00
function UIManager : run ( )
self._running = true
2015-03-20 02:14:13 +00:00
self : initLooper ( )
2015-04-28 03:06:00 +00:00
-- currently there is no Turbo support for Windows
2015-03-20 02:14:13 +00:00
-- use our own main loop
if not self.looper then
while self._running do
self : handleInput ( )
end
2015-03-02 09:15:26 +00:00
else
2015-03-20 02:14:13 +00:00
self.looper : add_callback ( function ( ) self : handleInput ( ) end )
2015-03-02 09:15:26 +00:00
self.looper : start ( )
end
2017-05-16 09:11:11 +00:00
return self._exit_code
2015-03-02 09:15:26 +00:00
end
-- run uimanager forever for testing purpose
function UIManager : runForever ( )
self._run_forever = true
2017-05-16 09:11:11 +00:00
return self : run ( )
2012-04-22 19:29:48 +00:00
end
2013-10-18 20:38:07 +00:00
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- The common operations that should be performed before suspending the device.
2017-05-16 09:11:11 +00:00
function UIManager : _beforeSuspend ( )
self : flushSettings ( )
self : broadcastEvent ( Event : new ( " Suspend " ) )
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
-- Reset gesture detection state to a blank slate (anything power-management related emits KEY events, which don't need gesture detection).
Input : resetState ( )
2017-05-16 09:11:11 +00:00
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- The common operations that should be performed after resuming the device.
2017-05-16 09:11:11 +00:00
function UIManager : _afterResume ( )
self : broadcastEvent ( Event : new ( " Resume " ) )
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- The common operations that should be performed when the device is plugged to a power source.
2017-05-16 09:11:11 +00:00
function UIManager : _beforeCharging ( )
2020-10-21 23:17:34 +00:00
if G_reader_settings : nilOrTrue ( " enable_charging_led " ) then
Device : toggleChargingLED ( true )
end
2017-05-16 09:11:11 +00:00
self : broadcastEvent ( Event : new ( " Charging " ) )
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- The common operations that should be performed when the device is unplugged from a power source.
2017-05-16 09:11:11 +00:00
function UIManager : _afterNotCharging ( )
2020-10-21 23:17:34 +00:00
if G_reader_settings : nilOrTrue ( " enable_charging_led " ) then
Device : toggleChargingLED ( false )
end
2017-05-16 09:11:11 +00:00
self : broadcastEvent ( Event : new ( " NotCharging " ) )
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Executes all the operations of a suspension ( i.e . , sleep ) request .
This function usually puts the device into suspension .
] ]
2017-05-16 09:11:11 +00:00
function UIManager : suspend ( )
2020-02-08 00:58:10 +00:00
if Device : isCervantes ( ) or Device : isKobo ( ) or Device : isSDL ( ) or Device : isRemarkable ( ) or Device : isSonyPRSTUX ( ) then
2017-05-16 09:11:11 +00:00
self.event_handlers [ " Suspend " ] ( )
elseif Device : isKindle ( ) then
2019-03-21 10:52:58 +00:00
Device.powerd : toggleSuspend ( )
2020-10-17 10:59:24 +00:00
elseif Device.isPocketBook ( ) and Device.canSuspend ( ) then
Device : suspend ( )
2017-05-16 09:11:11 +00:00
end
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Executes all the operations of a resume ( i.e . , wakeup ) request .
This function usually wakes up the device .
] ]
2017-05-16 09:11:11 +00:00
function UIManager : resume ( )
2020-02-08 00:58:10 +00:00
if Device : isCervantes ( ) or Device : isKobo ( ) or Device : isSDL ( ) or Device : isRemarkable ( ) or Device : isSonyPRSTUX ( ) then
2017-05-16 09:11:11 +00:00
self.event_handlers [ " Resume " ] ( )
elseif Device : isKindle ( ) then
self.event_handlers [ " OutOfSS " ] ( )
end
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Release standby lock .
Called once we ' re done with whatever we were doing in the background.
Standby is re - enabled only after all issued prevents are paired with allowStandby for each one .
] ]
2020-08-30 13:01:56 +00:00
function UIManager : allowStandby ( )
assert ( self._prevent_standby_count > 0 , " allowing standby that isn't prevented; you have an allow/prevent mismatch somewhere " )
self._prevent_standby_count = self._prevent_standby_count - 1
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--[[--
Prevent standby .
i.e . , something is happening in background , yet UI may tick .
] ]
2020-08-30 13:01:56 +00:00
function UIManager : preventStandby ( )
self._prevent_standby_count = self._prevent_standby_count + 1
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
-- The allow/prevent calls above can interminently allow standbys, but we're not interested until
2020-08-30 13:01:56 +00:00
-- the state change crosses UI tick boundary, which is what self._prev_prevent_standby_count is tracking.
function UIManager : _standbyTransition ( )
if self._prevent_standby_count == 0 and self._prev_prevent_standby_count > 0 then
-- edge prevent->allow
logger.dbg ( " allow standby " )
Device : setAutoStandby ( true )
self : broadcastEvent ( Event : new ( " AllowStandby " ) )
elseif self._prevent_standby_count > 0 and self._prev_prevent_standby_count == 0 then
-- edge allow->prevent
logger.dbg ( " prevent standby " )
Device : setAutoStandby ( false )
self : broadcastEvent ( Event : new ( " PreventStandby " ) )
end
self._prev_prevent_standby_count = self._prevent_standby_count
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--- Broadcasts a `FlushSettings` Event to *all* widgets.
2017-05-16 09:11:11 +00:00
function UIManager : flushSettings ( )
self : broadcastEvent ( Event : new ( " FlushSettings " ) )
end
Revamp flash_ui handling, once more, with feeling ;) (#7262)
* Simplify flash_ui handling (by handling the unhighlight pre-callback, c.f., #7262 for more details).
* UIManager: Handle translucent window-level widgets (and those wrapped in a translucent MovableContainer) properly in setDirty directly, making sure what's *underneath* them gets repainted to avoid alpha layering glitches. (This was previously handled via localized hacks).
* Update UIManager's documentation, and format it properly for ldoc parsing, making the HTML docs more useful.
* ReaderView: Reinitialize the various page areas when opening a new document, to prevent poisoning from the previous document.
* Event: Handle nils in an event's arguments.
* CheckButton/RadioButton: Switch to simple inversion to handle highlighting
* CheckButton: Make the highlight span the inner frame's width, instead of just the text's width, if possible.
* AlphaContainer: Fix & simplify, given the UIManager alpha handling.
* MovableContainer: When translucent, cache the canvas bb used for composition.
* Avoid spurious refreshes in a few widgets using various dummy *TextWidgets in order to first compute a text height.
* KeyValuePage: Avoid floats in size computations.
2021-02-20 17:22:48 +00:00
--- Sanely restart KOReader (on supported platforms).
2017-05-16 09:11:11 +00:00
function UIManager : restartKOReader ( )
self : quit ( )
-- This is just a magic number to indicate the restart request for shell scripts.
self._exit_code = 85
end
The great Input/GestureDetector/TimeVal spring cleanup (a.k.a., a saner main loop) (#7415)
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
2021-03-30 00:57:59 +00:00
--- Sanely abort KOReader (e.g., exit sanely, but with a non-zero return code).
function UIManager : abort ( )
self : quit ( )
self._exit_code = 1
end
2014-06-26 17:37:33 +00:00
UIManager : init ( )
2013-10-18 20:38:07 +00:00
return UIManager