Anki 24.10 Beta

[DerIshmaelite]

I have made a separate topic about this to keep this topic relevant to the current Beta.

FSRS 5: <1d Scheduling and Learning Steps

Further on his topic on Github: Let FSRS control short term schedule by L-M-Sherlock · Pull Request #3375 · ankitects/anki (github.com) [image] I see that a toggle is a good middle ground for this. You either: 1-Keep learning with normal learning steps as before 2-Keep learning with normal learning steps and toggle on further automatic scheduling 3-Leave learning steps empty and only learn with automatic scheduling Regarding the complexity of the UI, the only change would be the addition…

[Keks]

This does not apply to the current beta version.
When I manually increase the FSRS parameters (w2) in Ankidroid after optimization, the RMSE decreases. This happens with almost all presets.

[Kuuube]

Hi, a user reported a break in the Kanji Grid Kuuube addon from this update. The all_names_and_ids method appears to be broken. Calling it results in a crash without any logs from Anki.

Windows event viewer logs the following error:

Faulting application name: anki.exe, version: 0.0.0.0, time stamp: 0x66f6873c
Faulting module name: _message.cp39-win_amd64.pyd, version: 0.0.0.0, time stamp: 0x66cf5411
Exception code: 0xc0000005
Fault offset: 0x00000000000110f8
Faulting process id: 0x0xA83C
Faulting application start time: 0x0x1DB15B74B7B1534
Faulting application path: C:\Users\username_removed\AppData\Local\Programs\Anki\anki.exe
Faulting module path: C:\Users\username_removed\AppData\Local\Programs\Anki\lib\google\_upb\_message.cp39-win_amd64.pyd
Report Id: c10337f5-c99f-4cf3-a2e3-b5668f0e3221
Faulting package full name: 
Faulting package-relative application ID: 

This occurs on anki-24.10-windows-qt6. It does not occur on linux or the previous windows version,anki-24.06.3-windows-qt6.

The following code in the __init.py__ of an addon can replicate this issue (load up Anki and click the “Run the bug” menu button):

from aqt import mw
from aqt.qt import (QAction)

def setup(mw):
    def the_bug():
        mw.col.models.all_names_and_ids()

    menuAction = QAction("Run the bug", mw, triggered=the_bug)
    mw.form.menuTools.addSeparator()
    mw.form.menuTools.addAction(menuAction)

if __name__ != "__main__":
    setup(mw)

[dae]

Beta 2 is now available:

Releases · ankitects/anki

Replying to a few posts above:

@jcznk

When closing the Deck Options screen, the Discard changes? dialog is not shown if changes were made only by directly interacting with an input field (either using the keyboard or the mouse, but not through UI buttons).

Logged on Deck option changes not detected until focus is lost · Issue #3472 · ankitects/anki · GitHub

@jcznk

On the Preferences screen, under the Review section, is including the hours past midnight and mins strings inside the input boxes an intentional design choice?

Yes - I prefer the look of the former, but the latter is more accessible for screen readers.

@sorata

Can we make it such that addition of extra parameters to the end is ignored by current FSRS?

What happens when some future FSRS release changes the meaning of existing parameters in such a way that they no longer make sense to apply to the older algorithm? I don’t think it’s sufficient to just ignore trailing numbers, as you also need a way to prevent reviews on older clients if it no longer makes sense.

@Expertium

Btw, Dae, I trust that you won’t release a new version if there are a lot of issues, despite what the version name suggests.

As has been mentioned previously, the version number is not set in stone until release. If it needs to be delayed due to important unresolved bugs, it will be delayed.

@Expertium

Dae wasn’t opposed to adding them in a later beta

Yes, but that’s assuming they were almost ready to go. If they are not ready soon, they’ll need to skip this release.

[Expertium]

What about this?

@Expertium

• Update the yellow warning message to “Please ensure all of your Anki clients are Anki(Mobile) 24.10+ or AnkiDroid 2.20(?)+. FSRS will not work correctly if one of your clients is older.”
• When the user clicks “Optimize”, display a warning that says “FSRS-5 parameters cannot be used on mobile devices/AnkiWeb for now. Do not optimize if you plan to use Anki on other devices” or something like that, but shorter. This is a band-aid, we’ll remove it in the final release. David said they’re getting error reports from people who don’t read and are optimizing their parameters using this beta and then syncing to Ankidroid.

[cjdduarte]

This error is still present in this Beta 2 version.

[Expertium]

As I said, I can’t reproduce it. In fact, I don’t think it’s an “error”, it’s just that if the new parameters aren’t better than the old ones (which is possible), the old ones are kept.
You can ping LMSherlock, but he’s on vacation right now.
EDIT: nevermind, I can reproduce it, it happens with one of my presets, too. I still think that this isn’t a bug/error, more like undesirable consequences of how RMSE is compared.

[oktoberpaard]

So the current status in beta 2 is that FSRS can optimize the last two parameters, but still can’t do intraday scheduling, correct? So it’s best if I leave my leaning steps at the current value of 15m.

[Expertium]

So the current status in beta 2 is that FSRS can optimize the last two parameters, but still can’t do intraday scheduling, correct?

Yes.

[vaibhav]

I believe that this table can be made much prettier.

[Keks]

[cjdduarte]

BETA 2:

It seems that something in the layout of the new version of Anki is breaking the buttons, and both the <td align="end"> and <td align="start"> elements are causing issues. These elements are occasionally overlapping the “Good” and “Easy” buttons, preventing them from being clicked.

version 24.06

version 24.10

[nwwt]

The two cases are fixed again in Beta 2.

We couldn’t agree on {{type:text:FieldName}} being a good idea (and it would require some making internal field replacement processing more dynamic/forgiving), so it’s not implemented. If somebody is interested, please comment.

I had another idea about a {{css:Field}} feature to safely use fields (esp. Tags with hierarchical ones) as CSS identifiers, but it’s also too niche/problematic.

[salman]

Get the same error in beta 2

In the Editor and browser when I right-click an image

Anki 24.10 (85f034b1)
Python 3.9.18 Qt 6.6.2 PyQt 6.6.1
Platform: Windows-10-10.0.22631

Traceback (most recent call last):
File “aqt.editor”, line 1595, in contextMenuEvent
File “aqt.editor”, line 1612, in _add_image_menu
AttributeError: ‘NoneType’ object has no attribute ‘mediaUrl’

[vaibhav]

@salman

Get the same error in beta 2

In the Editor and browser when I right-click an image

Should be fixed in the next beta

[cjdduarte]

BETA2

[DerIshmaelite]

@Expertium Is the short term scheduler already been enabled

---

Also I only see a sum total of only 17 parameters in the FSRS parameter box after clicking Optimise All Presets (that is fixed now). Shouldn’t it be 19 though

[zxhd863943427]

After testing the new version, it can be closed successfully, but you need to press the close button twice… How can I get more information to help troubleshoot an issue?

[jcznk]

@dae

Yes - I prefer the look of the former, but the latter is more accessible for screen readers.

That makes sense, and it seems like a step in the right direction for Anki. The only thing that slightly bothers me, and that I think makes navigating with the keyboard and mouse a bit more difficult, is that the cursor can now be placed in one of two positions when clicking inside the input box (depending on how far left or right you click). Either it’s immediately after the numerical value to be edited (good), or after the unmodifiable string (less ideal). While the latter can be easily fixed by moving left with the keyboard—and doesn’t really matter if the user plans to change the numerical value using the keyboard’s up/down arrows—I still think it would make for a smoother user experience if this could somehow be prevented.

[voczi]

Well, it seems like you have already posted your debug info. It seems to close on a single click for me… so I feel like it’s something to do with Qt5, since that’s the only difference on my end?