there were a number of issues with preloading:
- it could result in duplicate IDs in the document
- embedded styles failed to apply, because a preloaded item was
overriding them
- the type answer code needed rewriting to support it
so we're back to something closer to the old approach - a single node
that we fade out, update, and then fade in again. this means there's a
longer delay when revealing mathjax, but should minimize the breakage of
existing notes
this fixes a bug where navigating to the next/previous card using
shortcut keys resulted in the first field being clobbered
- get rid of the stealFocus option in favour of explicitly passing
focusTo to setNote()
- setFields() is no longer responsible for setting focus
- add focusTo var to the browser so that the row changed hook can
restore focus when navigating to next/previous card
- fix the row changed hook being called twice
- the blur event now includes the field number instead of relying on the
editor to have the correct currentField
- the current field is set to null on blur
- use deferred js and a callback rather than keeping track of when we
were loaded
- add shift+tab shortcut to go to previous field
- to minimize pop-in of images and mathjax, fade the previous card
out and fade the next card in
- render the answer at the same time as the question so it can be
shown immediately
- move reviewer css to separate file
- remove image preloading code that should no longer be necessary
- js code that was previously bundled in .py files is now in the
web folder
- add helpers to create links to bundled files, and update
stdHtml() to accept a list of javascript files to include
instead of text
- render MathJax in card layout and preview screens - these should be
updated in the future to update the document dynamically like the
reviewer does
- start media server earlier so it can be used to serve content for
the toolbar, etc
- work around a bug in WebEngine on Windows that could cause the
media server to hang
- add option to wrap html in implicit pre-wrap environment during
editing and review - defaults to off
- update paste filter to convert divs/Ps to newlines and non-breaking
spaces to normal ones
- catch enter key and write \n instead of creating a new div
also:
- remove extra caretToEnd() call that is no longer required
- add dd/dt/dl to allowed tags
we can't use an event filter on the top level webview, because it
ignores the return value of the filter and leads to Anki thinking
keys have been pressed twice
and if we use an event filter on the focusProxy(), the
keypress/release events are sent even when a text field is currently
focused, leading to shortcuts being triggered when typing in the answer
to solve this, we move away from handling the key press events
directly, and instead install shortcuts for the events we want to
trigger. in addition to the global shortcuts, each state can install
its own shortcuts, which we remove when transitioning to a new state
also remove the unused canFocus argument to ankiwebview, and accept a parent
argument as required by the code in forms/
- key presses while a webview is focused no longer make it to the
main window's keyPressEvent() routine, so AnkiWebView now uses its
event filter to pass the key events to the main window
- move the shared key handling out of keyPressEvent into
globalKeyHandler()
- make sure all key handling routines return true or false to
indicate if an event was handled or not
- remove focus when esc hit in the main window, to retain old
behaviour of allowing esc to clear focus from the type answer box
onClose() was calling closeAllWindows(), which sends a close signal
to dialogs without force=True, causing them to fire an asynchronous
save event that calls back after the collection has been closed
instead we just rely on the fact that unloadCollection() closes all
collection windows
saveNow() must run asynchronously, but unloadCollection() was not
designed to. fix the issue by skipping the save, as it's unlikely
the user can move the cursor from the editor to the
main window and start a sync before the timer has automatically saved
the contents.
Anyone else who's contributed anything to Anki - however small - is most
welcome to be listed in the about screen. Sending me a pull request that
adds your name or alias would be easiest, but if that is difficult
please let us know on the support site and we'll do it for you.
onedrive seems to frequently screw up the permissions on the documents
folder. since we're stuck with more difficult to find folders on
mac/win, we may as well migrate linux as well, to a more
standards-compliant path.
each time we send or receive a chunk of data we check to see if the
user wants to cancel sync
in the case of a hung connection, it will still take a minute to time
out
we do the backup compression/writing in a separate thread so it
doesn't slow down profile switching/syncing, and remove the option
to write uncompressed backups
the strip html option is no longer used, so remove it from preferences
- wrap request in AnkiRequestsClient so we can keep track of
upload/download bytes without having to monkey patch anything
- force a 64kB buffer size instead of the default 8kB
- show one decimal point in up/down so small requests still give
visual feedback
- update add-on downloading and update check to use requests
- remove the update throttling in aqt/sync.py, as it's not really
necessary anymore
The html of each anki button is stored in a list that is joined in a string at the end. This make it easier to run the filter.
Drawback: Only buttons on the right side of the editor can be added using the filter.
- insert resulting html directly into fields instead of relying on
fragile clipboard rewriting
- catch paste events in js and send them back to python so we cover
all possible ways of pasting
- use checksum in pasted image filenames instead of .cacheKey(),
which was not unique across runs
- try each _processX() in turn instead of having special cases inside
them
- rewrite the HTML filtering:
- img rewriting and stripping of certain tags handled in bs4; other
processing handled in js for easier sharing with other clients
- use a whitelist of HTML tags and properties instead of focusing
mainly on webkit cruft
- don't run filter when focus lost, as it no longer seems to
be necessary
- the "strip html" option is currently ignored
Bs4 will raise a warning if the markup (field input) starts with '/'. Suppressing the warning is probably the easiest solution, as Bs4 will still process it (no disadvantages?).
the current code was freezing when clicking on 'cards' in the
browser - it looks like like the javascript callback was never
being called despite calling processEvents(). so we need to
refactor the code to call saveNow() with a callback that does the
subsequent processing.
a lot of the browser code was implicitly calling saveNow() via
beginReset(), so we've had to change all that code to save
immediately before it begins any processing. found a probable bug in
the process - it doesn't look like onRowChange() was saving before
overwriting the note, so theoretically edits could be lost if the
user switched to another card very quickly after typing something.
onSearch() has been split into a GUI-activated onSearchActivated()
that takes care of saving, and a lower level search() that refreshes
the current search. it keeps track of the last search via an instance
variable so that it refreshes properly if a user accidentally adds
some characters to their search without activating the search, then
does something like reverse the sort order.
start the progress dialog immediately so that askUser()
becomes part of the progress dialog's modality
the problem with the previous approach was that beginReset() attempts
to save the note again, leading to schema mismatches
