2
0
mirror of https://github.com/koreader/koreader synced 2024-11-13 19:11:25 +00:00
koreader/frontend/languagesupport.lua

318 lines
13 KiB
Lua
Raw Normal View History

reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
--[[--
Language-specific handling module.
This module defines a somewhat generic system by which language-specific
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
plugins can improve KOReader's support for languages that are not close enough
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
to European languages to "just work".
This was originally designed to improve KoReader's Japanese support through the
Japanese plugin (plugins/japanese.koplugin) but it should be generic enough for
other language plugins to build off this framework. Examples of languages which
may require such a plugin include highly inflected/agglutinative languages
(Japanese and Korean) and languages where spaces are not used as a delimiting
character between words (Japanese and Chinese) and languages that use a
character set where KoReader's dependencies define a "word" as being a single
character.
This module works by providing a mechanism to define a series of callbacks (not
unlike UI Events) which are called during operations where language-specific
knowledge may be necessary (such as during text selection and dictionary lookup
of a text fragment).
]]
local WidgetContainer = require("ui/widget/container/widgetcontainer")
local dbg = require("dbg")
local logger = require("logger")
local _ = require("gettext")
-- Shared among all LanguageSupport instances to make sure we don't lose
-- plugins when reloading different viewers.
local PluginListSingleton = {}
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 LanguageSupport = WidgetContainer:extend{
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
name = "language_support",
plugins = PluginListSingleton,
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
}
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
--[[--
Registers a new language-specific plugin with given language_code.
If a plugin using the same language code already exists, the loading is
skipped. The language code is used to prioritise which language plugin should
be called first (if that plugin fails to handle the text, there is no such
plugin, or the document has no language defined then all of the plugins are
called one-by-one until one succeeds).
The follow handlers are defined (if you wish to support the handler create the
corresponding onHandler method):
- WordSelection (onWordSelection) is called when a highlight is
first created and can be used to modify the default "word boundary" word
selection to match user expectations.
Note that "XPointer" is used here but code written should take it as an
opaque object and only use the given callbacks since in future this feature
may work with PDFs and all of the XPointers will instead be PDF equivalents.
Called with {
pos0 = [XPointer], pos1 = [XPointer],
callbacks = {
-- Equivalent to document:getPrevVisibleChar(pos).
get_prev_char_pos(pos: XPointer) -> XPointer,
-- Equivalent to document:getNextVisibleChar(pos).
get_next_char_pos(pos: XPointer) -> XPointer,
-- Equivalent to document:getTextFromXPointers(pos0, pos1).
get_text_in_range(pos0: XPointer, pos1: XPointer) -> string,
}
} table as the only argument.
Must return the new (pos0, pos1) XPointers or nil if the word couldn't be
expanded.
- WordLookup (onWordLookup) is called when a dictionary lookup is triggered on
some text and can be used to adjust the word such that it is in the
dictionary form and can be found in the dictionary. This is primarily useful
for languages where StarDict "fuzzy searching" is not usable.
Called with
{ text = [string] }
table as the only argument.
Must return an array of candidate words (in decreasing order of preference)
which will be looked up or nil if no candidate words could be generated.
Note that if more than one candidate is found in the dictionary they will
all be displayed to the user. It is not necessary to include the original
word in the candidate list -- it will always be given highest priority.
@param Plugin to register (plugin.name is used as the internal name and must not be nil or "").
@treturn bool Whether the plugin was successfully registered.
]]
function LanguageSupport:registerPlugin(plugin)
if not dbg.dassert(plugin.name ~= nil and plugin.name ~= "", "plugin name must be non-empty") then
logger.warn("language support: ignoring attempted registration of plugin with empty name")
return false
end
if self.plugins[plugin.name] ~= nil then
logger.dbg("language support: overriding existing", plugin.name, "plugin")
else
logger.dbg("language support: registering", plugin.name, "plugin")
end
self.plugins[plugin.name] = plugin
return true
end
--- Returns whether there are any language plugins currently enabled.
-- @treturn bool whether there are any registered plugins
function LanguageSupport:hasActiveLanguagePlugins()
return next(self.plugins) ~= nil
end
local function callPlugin(plugin, handler_name, ...)
local handler = plugin["on"..handler_name]
if handler == nil then
logger.dbg("language plugin", plugin, "missing handler", handler)
return
end
-- Handler could return any number of values, collect them all.
local ret = {pcall(handler, plugin, ...)}
local ok = table.remove(ret, 1)
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
if not ok then
logger.err("language plugin", plugin, "crashed during", handler_name, "handler:", unpack(ret))
return
end
logger.dbg("language plugin", handler_name, "returned", ret)
return ret
end
function LanguageSupport:_findAndCallPlugin(language_code, handler_name, ...)
-- First try any plugin that supports the language code specified.
for name, plugin in pairs(self.plugins) do
if plugin:supportsLanguage(language_code) then
logger.dbg("language support: trying", name, "plugin's", handler_name)
local ret = callPlugin(plugin, handler_name, ...)
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
if ret ~= nil then
return unpack(ret)
end
end
end
-- Fallback path. Try every remaining plugin in case the document had the
-- wrong language defined (or no language defined) or had the correct
-- language defined but contained text not in the document language.
for name, plugin in pairs(self.plugins) do
if not plugin:supportsLanguage(language_code) then
logger.dbg("language support (fallback): trying", name, "plugin's", handler_name)
local ret = callPlugin(plugin, handler_name, ...)
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
if ret ~= nil then
return unpack(ret)
end
end
end
end
-- @fixme This was copied from readerhighlight, but it should probably live in
-- util or some textutil module.
local function cleanupSelectedText(text)
-- Trim spaces and new lines at start and end
text = text:gsub("^[\n%s]*", "")
text = text:gsub("[\n%s]*$", "")
-- Trim spaces around newlines
text = text:gsub("%s*\n%s*", "\n")
-- Trim consecutive spaces (that would probably have collapsed
-- in rendered CreDocuments)
text = text:gsub("%s%s+", " ")
return text
end
local function createDocumentCallbacks(document)
if not document or document.info.has_pages then
-- We need document:get{Prev,Next}VisibleChar at a minimum and there
-- isn't an alternative for PDFs at the moment (not to mention for
-- quite a few CJK PDFs, MuPDF seems to be unable to create selections
-- at a character level even using hold-and-pan).
logger.dbg("language support currently cannot expand document selections in non-EPUB formats")
return
end
return {
get_prev_char_pos = function(pos) return document:getPrevVisibleChar(pos) end,
get_next_char_pos = function(pos) return document:getNextVisibleChar(pos) end,
get_text_in_range = function(pos0, pos1) return document:getTextFromXPointers(pos0, pos1) end,
}
end
--- Called from ReaderHighlight:onHold after the document-specific handler has
-- successfully grabbed a "word" from the document. If the selection is to
-- updated, improveWordSelection will also update the document's internal
-- selection state (for crengine) to correctly display the right selection.
-- @param selection Text selection table to improve if possible.
-- @return New updated selected_text table which should be used (or nil).
function LanguageSupport:improveWordSelection(selection)
if not self:hasActiveLanguagePlugins() then return end -- nothing to do
if not self.document then
logger.dbg("language support: cannot improve word selection outside document")
return
end
local language_code = self.ui.doc_props.language or "unknown"
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
logger.dbg("language support: improving", language_code, "selection", selection)
-- Rather than requiring each language plugin to use document: methods
-- correctly, return a set of callbacks that are document-agnostic (and
-- have the document handle as an upvalue of the closure) and could be used
-- for non-EPUB formats in the future.
local callbacks = createDocumentCallbacks(self.document)
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
if not callbacks then
return
end
local new_pos0, new_pos1 = unpack(self:_findAndCallPlugin(
language_code, "WordSelection",
{ text = selection.text, pos0 = selection.pos0, pos1 = selection.pos1, callbacks = callbacks }
) or {})
-- If no plugin could update the selection (or after "expansion" the
-- selection is the same) then we can safely skip all of the subsequent
-- re-selection work.
if not new_pos0 or not new_pos1 or
(new_pos0 == selection.pos0 and new_pos1 == selection.pos1) then
logger.dbg("language support: no plugin could improve the selection")
return
end
logger.dbg("language support: updating selection\nfrom",
selection.pos0, ":", selection.pos1, "\nto", new_pos0, ":", new_pos1)
-- We want to use native crengine text selection here, but we cannot use
-- getTextFromPositions because the conversion to and from screen
-- co-ordinates leads to issues with text selection of <ruby> text. In
-- addition, using getTextFromXPointers means we can select text not on the
-- screen. But this means we need to manually create the text selection
-- object returned by getTextFromPositions (though this is not a big deal
-- because we'd have to generate the sboxes anyway).
local new_text = self.document:getTextFromXPointers(new_pos0, new_pos1, true)
if not new_text then
-- This really shouldn't happen since we started with some text.
logger.warn("language support: no text found in selection", new_pos0, ":", new_pos1)
return
end
return {
text = cleanupSelectedText(new_text),
pos0 = new_pos0,
pos1 = new_pos1,
sboxes = self.document:getScreenBoxesFromPositions(new_pos0, new_pos1, true),
}
end
--- Called from ReaderHighlight:startSdcv after the selected has text has been
-- OCR'd, cleaned, and otherwise made ready for sdcv.
-- @tparam string text Original text being searched by the user.
-- @treturn {string,...} Extra dictionary form candidates to search (or nil).
function LanguageSupport:extraDictionaryFormCandidates(text)
if not self:hasActiveLanguagePlugins() then return end -- nothing to do
local language_code = (self.ui.doc_props and self.ui.doc_props.language) or "unknown"
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
logger.dbg("language support: convert", text, "to dictionary form (marked as", language_code..")")
return self:_findAndCallPlugin(
language_code, "WordLookup",
{ text = text }
)
end
function LanguageSupport:addToMainMenu(menu_items)
if not self:hasActiveLanguagePlugins() then return end -- nothing to do
-- Sort the plugin keys so we have consistent ordering in the menu.
local plugin_names = {}
for name in pairs(self.plugins) do
table.insert(plugin_names, name)
end
table.sort(plugin_names)
-- Link up each plugin's submenu.
local sub_table = {}
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
for _, name in ipairs(plugin_names) do
local plugin = self.plugins[name]
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
if plugin.genMenuItem ~= nil then
local menuItem = plugin:genMenuItem()
-- Set help_text in case the plugin hasn't.
if not menuItem.help_text and not menuItem.help_text_func then
menuItem.help_text = plugin.description
end
table.insert(sub_table, menuItem)
else
-- Plugin didn't have a menu defined so use a basic fallback menu,
-- showing a description of the plugin when held for help.
reader: implement language-support plugin system This creates a new plugin system which hooks into a handful of reader operations in order to allow plugins to add language-specific support where the default reader falls short. The two hooks added are: * During hold-without-pan taps, language plugins can modify the selection in order to better match what users expect koreader to highlight when selecting a single word. The vast majority of CJK language words are more than one character, but KOReader treats all CJK characters as a single word by default, so adding this hook means that readers no longer need to manually select the whole word every time they need to look something. * During dictionary lookup, language plugins can propose alternative candidate words to look up if the selected word could not be found in the dictionary. This is pretty necessary for Japanese and Korean, both of which are highly agglutinative languages and the fuzzy searching system of StarDict is simply not usable because often the inflection of the word is so much longer than the dictionary form that sdcv decides to chop off the actual word and search for the inflection (which yields useless results). This system is of particular interest for readers of CJK languages (without this, looking up words using KOReader was fairly painful) but this system is designed to be minimal and language-agnostic enough that other languages could make use of it by creating their own plugins if the default "whole word" highlight and fuzzy-search system doesn't match their needs. Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
2021-10-23 10:13:09 +00:00
table.insert(sub_table, {
text = plugin.pretty_name or plugin.fullname or name,
help_text = plugin.description,
keep_menu_open = true
})
end
end
-- Only show the menu item if there are some plugins enabled.
if #sub_table ~= 0 then
menu_items.language_support = {
text = _("Language support plugins"),
sorting_hint = "document",
help_text = _([[
This menu lets you manage KOReader's language support plugins and their associated settings.
These plugins provide language-specific helpers to KOReader, to improve the reading experience with languages that require some extra handling when compared to most European languages.
In order to disable a language plugin, you need to disable it from the Plugin Management menu.]]),
sub_item_table = sub_table,
}
end
end
function LanguageSupport:init()
self.document = self.document or self.ui and self.ui.document
if self.ui and self.ui.menu then
self.ui.menu:registerToMainMenu(self)
end
end
return LanguageSupport