Difficulty to know which plural form is selected

[CyrilleB79]

I am a blind translator of NVDA screen reader.

As a blind user, when translating plural forms it's difficult to know from the GUI which form is being translated.

Steps to reproduce with NVDA

1. Open a .po file containing a string to translate with a plural form, for a language with 2 forms (1 singular and 1 plural).
2. Select this string in the list
3. Press tab until you reach the "Singular" button (i.e. at least 5 times)
4. Press tab once again to go to the "plural" button
5. Press "enter" to press the "plural button
6. Press "space" to press the "plural button. Then the focus jumps to the translation edit field.
7. Listen to the name of the edit field

Actual result

• Steps 3. and 4. one of the two buttons should be active (and actually it is visually underlined), but this information is not conveyed by any mean to the accessibility APIs.
• Step 5: nothing happens
• Step 7: the accessible name of the edit field is named "singular"; I am unsure if this name is provided by poedit to accessibility APIs or if it's derived from something else by the APIs themselves or by the screen reader NVDA.

Expected result

• Steps 3. and 4. one of the two buttons should report a state indicating to accessibility APIs that it is selected (e.g. pressed for a button, checked for a checkbox, selected for a tab, etc.). Alternatively, if it's difficult to provide a state to accessibility APIs, the name of the button should include its state.
• Step 5: the button should have been activated (but this issue is quite minor since we may use space).
• Step 7: the accessible name of the edit field should match the name of the selected (underlined) button, i.e. "plural" here

Notes

By the way, thanks to @LeonarddeR and you for the recent accessibility improvement in poedit.

[vslavik]

Poedit uses a "fake" notebook control here, assembled from some panels and buttons. This is actually broken in multiple interesting ways:

1. For screen readers, there's no notebook. I think it would be desirable to just use a standard notebook in the presence of screen reader, appearance be damned, no? (This is already done e.g. for other visual flair like menus or title-less windows.)
2. TAB navigation is thoroughly broken and while 1 would fix it, this needs to be fixed in the "visual" version too: TAB needs to move into the tab, not among the buttons, and left/right arrows need to be supported.
3. A11y names for text controls are missing (and may be hard to read).

I'm not sure where does "singular" come from; I don't see that in Accessibility Insights.

Regarding the last point, I also wonder about the usability of the tab names. For Roman languages, they are just plural/singular, but for many other languages, there are more than 2 forms and Poedit shows descriptive labels, e.g. for Czech it shows "One", "n → 2,3,4" and "n → 0,5,6,7" as the tabs. Many other systems use CLDR nomenclature ("One", "Few", "Many") that is easier to pronounce, but I, as a translator, never know what they apply to w/o checking, so I tried to do better for Poedit. Now I wonder if NVDA chokes on that or not and if a different phrasing (e.g. "2,3,4 etc.)" would be more helpful...?

[CyrilleB79]

Yes, a standard notebook controller should probably be more usable. If the focus stops once on the tabs list and if we can change the selected tab with left/arrow, that would be nice and completely usable.

Re the name of the buttons/tabs for Slavic languages:
I am not a translator of these languages. But I have worked on pluralization support in NVDA and thus have conducted tests with some of these languages. For someone who do not know the specificity of each langauge, I'd say than having the examples numbers (as today) is much more helpful than "some", "few", etc.