2
0
mirror of https://github.com/koreader/koreader synced 2024-11-18 03:25:46 +00:00
koreader/frontend/docsettings.lua

323 lines
13 KiB
Lua
Raw Normal View History

2017-04-14 04:11:10 +00:00
--[[--
This module is responsible for reading and writing `metadata.lua` files
in the so-called sidecar directory
([Wikipedia definition](https://en.wikipedia.org/wiki/Sidecar_file)).
]]
local DataStorage = require("datastorage")
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
local LuaSettings = require("luasettings")
local dump = require("dump")
local ffiutil = require("ffi/util")
2017-04-14 04:11:10 +00:00
local lfs = require("libs/libkoreader-lfs")
2017-08-17 01:38:58 +00:00
local logger = require("logger")
local util = require("util")
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
local DocSettings = LuaSettings:extend{}
local HISTORY_DIR = DataStorage:getHistoryDir()
local function buildCandidates(list)
local candidates = {}
local previous_entry_exists = false
for i, file_path in ipairs(list) do
-- Ignore missing files.
if file_path ~= "" and lfs.attributes(file_path, "mode") == "file" then
local mtime = lfs.attributes(file_path, "modification")
-- NOTE: Extra trickery: if we're inserting a "backup" file, and its primary buddy exists,
-- make sure it will *never* sort ahead of it by using the same mtime.
-- This aims to avoid weird UTC/localtime issues when USBMS is involved,
-- c.f., https://github.com/koreader/koreader/issues/9227#issuecomment-1345263324
if file_path:sub(-4) == ".old" and previous_entry_exists then
local primary_mtime = candidates[#candidates].mtime
-- Only proceed with the switcheroo when necessary, and warn about it.
if primary_mtime < mtime then
logger.warn("DocSettings: Backup", file_path, "is newer (", mtime, ") than its primary (", primary_mtime, "), fudging timestamps!")
-- Use the most recent timestamp for both (i.e., the backup's).
candidates[#candidates].mtime = mtime
end
end
table.insert(candidates, {
path = file_path,
mtime = mtime,
prio = i,
}
)
previous_entry_exists = true
else
previous_entry_exists = false
end
end
-- MRU sort, tie breaker is insertion order (higher priority locations were inserted first).
-- Iff a primary/backup pair of file both exist, of the two of them, the primary one *always* has priority,
-- regardless of mtime (c.f., NOTE above).
table.sort(candidates, function(l, r)
if l.mtime == r.mtime then
return l.prio < r.prio
else
return l.mtime > r.mtime
end
end)
return candidates
end
2017-04-14 04:11:10 +00:00
--- Returns path to sidecar directory (`filename.sdr`).
--
-- Sidecar directory is the file without _last_ suffix.
2017-04-14 04:11:10 +00:00
-- @string doc_path path to the document (e.g., `/foo/bar.pdf`)
-- @treturn string path to the sidecar directory (e.g., `/foo/bar.sdr`)
2016-01-06 05:28:26 +00:00
function DocSettings:getSidecarDir(doc_path)
if doc_path == nil or doc_path == '' then return '' end
local file_without_suffix = doc_path:match("(.*)%.")
if file_without_suffix then
return file_without_suffix..".sdr"
end
return doc_path..".sdr"
end
2017-04-14 04:11:10 +00:00
--- Returns path to `metadata.lua` file.
-- @string doc_path path to the document (e.g., `/foo/bar.pdf`)
-- @treturn string path to `/foo/bar.sdr/metadata.lua` file
function DocSettings:getSidecarFile(doc_path)
if doc_path == nil or doc_path == '' then return '' end
2017-02-08 18:46:55 +00:00
-- If the file does not have a suffix or we are working on a directory, we
-- should ignore the suffix part in metadata file path.
local suffix = doc_path:match(".*%.(.+)")
if suffix == nil then
suffix = ''
end
return self:getSidecarDir(doc_path) .. "/metadata." .. suffix .. ".lua"
end
2017-04-14 04:11:10 +00:00
--- Returns `true` if there is a `metadata.lua` file.
-- @string doc_path path to the document (e.g., `/foo/bar.pdf`)
-- @treturn bool
function DocSettings:hasSidecarFile(doc_path)
2017-02-08 18:46:55 +00:00
return lfs.attributes(self:getSidecarFile(doc_path), "mode") == "file"
2016-01-06 05:28:26 +00:00
end
function DocSettings:getHistoryPath(fullpath)
return HISTORY_DIR .. "/[" .. fullpath:gsub("(.*/)([^/]+)", "%1] %2"):gsub("/", "#") .. ".lua"
end
2013-03-12 04:51:00 +00:00
function DocSettings:getPathFromHistory(hist_name)
if hist_name == nil or hist_name == '' then return '' end
if hist_name:sub(-4) ~= ".lua" then return '' end -- ignore .lua.old backups
2014-03-13 13:52:43 +00:00
-- 1. select everything included in brackets
local s = string.match(hist_name,"%b[]")
if s == nil or s == '' then return '' end
2014-03-13 13:52:43 +00:00
-- 2. crop the bracket-sign from both sides
-- 3. and finally replace decorative signs '#' to dir-char '/'
return string.gsub(string.sub(s, 2, -3), "#", "/")
2013-03-12 04:51:00 +00:00
end
function DocSettings:getNameFromHistory(hist_name)
if hist_name == nil or hist_name == '' then return '' end
if hist_name:sub(-4) ~= ".lua" then return '' end -- ignore .lua.old backups
local s = string.match(hist_name, "%b[]")
if s == nil or s == '' then return '' end
2014-03-13 13:52:43 +00:00
-- at first, search for path length
-- and return the rest of string without 4 last characters (".lua")
return string.sub(hist_name, string.len(s)+2, -5)
2013-03-12 04:51:00 +00:00
end
function DocSettings:getLastSaveTime(doc_path)
local attr = lfs.attributes(self:getSidecarFile(doc_path))
if attr and attr.mode == "file" then
return attr.modification
end
end
function DocSettings:ensureSidecar(sidecar)
if lfs.attributes(sidecar, "mode") ~= "directory" then
lfs.mkdir(sidecar)
end
end
2017-04-14 04:11:10 +00:00
--- Opens a document's individual settings (font, margin, dictionary, etc.)
-- @string docfile path to the document (e.g., `/foo/bar.pdf`)
-- @treturn DocSettings object
function DocSettings:open(docfile)
--- @todo (zijiehe): Remove history_path, use only sidecar.
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
-- NOTE: Beware, our new instance is new, but self is still DocSettings!
local new = DocSettings:extend{}
new.history_file = new:getHistoryPath(docfile)
local sidecar = new:getSidecarDir(docfile)
new.sidecar = sidecar
DocSettings:ensureSidecar(sidecar)
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
-- If there is a file which has a same name as the sidecar directory,
-- or the file system is read-only, we should not waste time to read it.
if lfs.attributes(sidecar, "mode") == "directory" then
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
-- New sidecar file name is metadata.{file last suffix}.lua.
-- So we can handle two files with only different suffixes.
new.sidecar_file = new:getSidecarFile(docfile)
new.legacy_sidecar_file = sidecar.."/"..
ffiutil.basename(docfile)..".lua"
end
-- Candidates list, in order of priority:
local candidates_list = {
-- New sidecar file
new.sidecar_file or "",
-- Backup file of new sidecar file
new.sidecar_file and (new.sidecar_file .. ".old") or "",
-- Legacy sidecar file
new.legacy_sidecar_file or "",
-- Legacy history folder
new.history_file,
-- Backup file in legacy history folder
new.history_file .. ".old",
-- Legacy kpdfview setting
docfile..".kpdfview.lua",
}
-- We get back an array of tables for *existing* candidates, sorted MRU first (insertion order breaks ties).
local candidates = buildCandidates(candidates_list)
local ok, stored, filepath
for _, t in ipairs(candidates) do
local candidate_path = t.path
-- Ignore empty files
if lfs.attributes(candidate_path, "size") > 0 then
ok, stored = pcall(dofile, candidate_path)
-- Ignore empty tables
if ok and next(stored) ~= nil then
logger.dbg("DocSettings: data is read from", candidate_path)
filepath = candidate_path
break
end
end
logger.dbg("DocSettings:", candidate_path, "is invalid, removed.")
os.remove(candidate_path)
2014-03-13 13:52:43 +00:00
end
if ok and stored then
2014-03-13 13:52:43 +00:00
new.data = stored
new.candidates = candidates
new.filepath = filepath
else
new.data = {}
2014-03-13 13:52:43 +00:00
end
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
return new
end
2017-04-14 04:11:10 +00:00
--- Serializes settings and writes them to `metadata.lua`.
function DocSettings:flush()
-- write serialized version of the data table into one of
-- i) sidecar directory in the same directory of the document or
-- ii) history directory in root directory of KOReader
if not self.history_file and not self.sidecar_file then
2014-03-13 13:52:43 +00:00
return
end
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
-- If we can write to sidecar_file, we do not need to write to history_file anymore.
local serials = {}
if self.sidecar_file then
table.insert(serials, self.sidecar_file)
end
if self.history_file then
table.insert(serials, self.history_file)
end
self:ensureSidecar(self.sidecar)
local s_out = dump(self.data)
for _, f in ipairs(serials) do
local directory_updated = false
2017-08-17 01:38:58 +00:00
if lfs.attributes(f, "mode") == "file" then
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
-- As an additional safety measure (to the ffiutil.fsync* calls used below),
-- we only backup the file to .old when it has not been modified in the last 60 seconds.
-- This should ensure in the case the fsync calls are not supported
-- that the OS may have itself sync'ed that file content in the meantime.
local mtime = lfs.attributes(f, "modification")
if mtime < os.time() - 60 then
logger.dbg("DocSettings: Renamed", f, "to", f .. ".old")
os.rename(f, f .. ".old")
directory_updated = true -- fsync directory content too below
end
2017-08-17 01:38:58 +00:00
end
logger.dbg("DocSettings: Writing to", f)
local f_out = io.open(f, "w")
if f_out ~= nil then
f_out:write("-- we can read Lua syntax here!\nreturn ")
f_out:write(s_out)
f_out:write("\n")
ffiutil.fsyncOpenedFile(f_out) -- force flush to the storage device
f_out:close()
if self.candidates ~= nil
and G_reader_settings:nilOrFalse("preserve_legacy_docsetting") then
for _, t in ipairs(self.candidates) do
local candidate_path = t.path
if candidate_path ~= f and candidate_path ~= f .. ".old" then
logger.dbg("DocSettings: Removed legacy file", candidate_path)
os.remove(candidate_path)
-- We should not remove sidecar folder, as it may
-- contain Kindle history files.
end
end
end
if directory_updated then
-- Ensure the file renaming is flushed to storage device
ffiutil.fsyncDirectory(f)
end
break
end
2014-03-13 13:52:43 +00:00
end
end
function DocSettings:getFilePath()
return self.filepath
end
2017-04-14 04:11:10 +00:00
--- Purges (removes) sidecar directory.
function DocSettings:purge(full)
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
-- Remove any of the old ones we may consider as candidates in DocSettings:open()
if self.history_file then
os.remove(self.history_file)
os.remove(self.history_file .. ".old")
end
if self.legacy_sidecar_file then
os.remove(self.legacy_sidecar_file)
end
if lfs.attributes(self.sidecar, "mode") == "directory" then
if full then
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
-- Asked to remove all the content of this .sdr directory, whether it's ours or not
ffiutil.purgeDir(self.sidecar)
else
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
-- Only remove the files we know we may have created with our usual names.
for f in lfs.dir(self.sidecar) do
local fullpath = self.sidecar.."/"..f
local to_remove = false
if lfs.attributes(fullpath, "mode") == "file" then
-- Currently, we only create a single file in there,
-- named metadata.suffix.lua (ie. metadata.epub.lua),
-- with possibly backups named metadata.epub.lua.old and
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
-- metadata.epub.lua.old_dom20180528,
-- so all sharing the same base: self.sidecar_file
if util.stringStartsWith(fullpath, self.sidecar_file) then
to_remove = true
end
end
if to_remove then
os.remove(fullpath)
logger.dbg("DocSettings: purged:", fullpath)
end
end
-- If the sidecar folder ends up empty, os.remove() can delete it.
-- Otherwise, the following statement has no effect.
os.remove(self.sidecar)
end
end
Clarify our OOP semantics across the codebase (#9586) Basically: * Use `extend` for class definitions * Use `new` for object instantiations That includes some minor code cleanups along the way: * Updated `Widget`'s docs to make the semantics clearer. * Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283) * Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass). * Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events. * Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier. * Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references. * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak). * Terminal: Make sure the shell is killed on plugin teardown. * InputText: Fix Home/End/Del physical keys to behave sensibly. * InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...). * OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of. * ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed! * Kobo: Minor code cleanups.
2022-10-06 00:14:48 +00:00
-- We should have meet the candidate we used and remove it above.
-- But in case we didn't, remove it.
if self.filepath and lfs.attributes(self.filepath, "mode") == "file" then
os.remove(self.filepath)
end
self.data = {}
end
2013-10-18 20:38:07 +00:00
return DocSettings