Closed
Bug 861113
Opened 12 years ago
Closed 12 years ago
Text prediction ignores capitalization of Polish letters
Categories
(Firefox OS Graveyard :: Gaia::Keyboard, defect)
Tracking
(blocking-b2g:-, b2g18+ fixed)
VERIFIED
FIXED
blocking-b2g | - |
People
(Reporter: stas, Unassigned)
References
Details
(Whiteboard: c=auto-suggest)
Attachments
(1 file)
(deleted),
text/html
|
Details |
Typing "Swiatl" correctly suggests "Światło", capitalized (among other words). Typing "Światl", however, suggests "światła", "światło", "światowej". v1.0.1 is not affected, as it uses a different prediction engine. This is similar to bug 851565, which seems to have been fixed by bug 860462.
Reporter | ||
Comment 1•12 years ago
|
||
David, is this by any chance related to the auto-correction feature from bug 860462?
Flags: needinfo?(dflanagan)
Comment 2•12 years ago
|
||
> v1.0.1 is not affected, as it uses a different prediction engine. Then marking it as leo? rather than tef?
blocking-b2g: tef? → leo?
Reporter | ||
Comment 3•12 years ago
|
||
Sorry, my bad, I forgot tracking flags get cloned along with bugs…
Reporter | ||
Comment 4•12 years ago
|
||
FWIW, this doesn't happen with Spanish "Á" at the beginning of "Álbum."
Comment 5•12 years ago
|
||
The new autocorrect code I landed restricts its uppercase letters to those in the Latin-1 encoding: const UC = /^[A-ZÀ-ÖØ-Þ]+$/; // all uppercase latin characters We need to extend this regular expression so that it also includes all Unicode uppercase letters. Stas: do you happen to know a complete list of Latin-like uppercase letters?
Blocks: 797170
Flags: needinfo?(dflanagan) → needinfo?(stas)
Comment 6•12 years ago
|
||
Tracking, not blocking, this looks like a low-risk fix, nominate for uplift when ready.
Comment 8•12 years ago
|
||
unicode has information, I found http://www.unicode.org/charts/PDF/U0180.pdf by looking at http://www.unicode.org/charts/charindex.html. There's lots of data there. I'm wondering if you can just use toLowerCase or toLocaleLowerCase and take it from there? Not sure which logic deals best with non-letter characters.
Flags: needinfo?(stas)
Comment 9•12 years ago
|
||
Evelyn, I'm asking for your review here because you worked on a similar bug for the 1.0.1 branch. In this case, I was defining uppercase letters with a regexp that only included the Latin-1 charset. So it was excluding polish letters and couldn't tell that accented uppercase letters were uppercase. I've changed that so it defines an uppercase letter as any letter x where x.toLowerCase() is not the same as x. This is a more general definition and works anywhere that toLowerCase() works. In my testing, this fixes the bug that Stas reported. Note, however that when I land 865484, that is going to alter the prediction engine, and the results it reports and his particular examples may no longer work.
Attachment #742230 -
Flags: review?(ehung)
Comment 10•12 years ago
|
||
The code makes sense but I need to verify it. I will test this patch later because I'm working on a leo+ and there is a tef+ review request pending. Sorry!
Comment 11•12 years ago
|
||
Landed on master: https://github.com/mozilla-b2g/gaia/commit/58665c2e62206c59356e1a4c040fa81251294e55
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
Comment 12•12 years ago
|
||
Comment on attachment 742230 [details] link to patch on github Clearing the review request. I landed this unreviewed because it was a very simple patch that needed to land before a much larger complicated patch in bug 865484.
Attachment #742230 -
Flags: review?(ehung)
Updated•12 years ago
|
Whiteboard: c=keyboard → c=auto-suggest
Comment 13•11 years ago
|
||
Word Suggestions in the Polish language setting are now being capitalized correctly. Tested on b2g/nightly/mozilla-central-unagi/latest Kernel Date: Dec 5 Gecko: http://hg.mozilla.org/mozilla-central/rev/3c6f2394995d Gaia: d2ad0f1a5a6be8a2d6df721ba3c2b33e037c423b
Status: RESOLVED → VERIFIED
Comment 14•11 years ago
|
||
Please add a testcase for this bug to moztrap for 1.1 testsuite. If yes, mark this in-moztrap+ when completed. If not, mark this in-moztrap-.
Flags: in-moztrap?(cschmoeckel)
Comment 15•11 years ago
|
||
Added Keyboard Suite Test Case #8450 - [Keyboard] Text prediction text entry in Polish keeps the capitalization entered by the user
Flags: in-moztrap?(cschmoeckel) → in-moztrap+
You need to log in
before you can comment on or make changes to this bug.
Description
•