237344 - Contractions not recognized by spelling checker when real apostrophe character used

Status: RESOLVED FIXED

(Core :: Spelling checker, defect, P1)

Description

[Greg K.]

When using a +ve contraction (e.g., we've) and a real apostrophe character
(U+2019) instead of a foot mark (U+0027), the spelling checker fails to
recognize the contraction. All other contractions (+s, +ll, +t, +re, +m and +d)
are apparently recognized when using either symbol.

To reproduce:
Create a new e-mail message and type in "we've we’ve" (using the real apostrophe
in the second word)
2. Check spelling

Actual results:
Spelling checker flags the second instance of "ve" as a word by itself

Expected results:
Spelling checker should show no misspellings.

Comment 1

[Greg K.]

Attachment: HTML testcase that can be opened in editor and spell-checked to demonstrate the bug

Comment 2

[Miles Libbey]

this appears to be a dup of 218463, but I don't have permissions to change it's
status...

Comment 3

[Michiel van Leeuwen (email: mvl+moz@)]

This is about catching misspellings, the other is about suggestions.

Comment 4

[Matthew Wilson]

This appears to be because s, ll, t, re, m and d all pass spell-checking as
individual words. "we値l" is checked as two separate words.

Comment 5

[Matthew Wilson]

Attachment: Proposed patch (partial - changes dictionary but only adds "we’ve" as a demonstration)

This has two parts. The change to mozEnglishWordUtils.cpp ensures that a U2019
apostophe is included as part of a word. But changing this on its own means
that the part of the word after the apostrophe is never spell-checked, because
the text is transformed into ISO8859-1 and the apostrophe is an illegal
character. So either Unicode apostrophes would need to be replaced with ASCII
apostrophes in a copy of the source text, or, as here, I changed the character
set to CP1252 so that the apostrophe character can be represented. This would
need the dictionaries to be updated, the patch includes only "we’ve" as a
demonstration. If this approach is approved then I could go through the
dictionary in full.

Comment 6

[Michiel van Leeuwen (email: mvl+moz@)]

Instead of changing myspell, i'd rather change ’ to ' before calling myspell.
Then you don't have to maintain your own version of the dictionary and don't
have to have two copies of one word in the dict.

Comment 7

[Matthew Wilson]

Comment on attachment 171546 [details] [diff] [review]
Proposed patch (partial - changes dictionary but only adds "we’ve" as a demonstration)

OK, withdrawing review request.

Comment 8

[Matthew Wilson]

> Instead of changing myspell, i'd rather change ’ to ' before calling myspell.
> Then you don't have to maintain your own version of the dictionary and don't
> have to have two copies of one word in the dict.

On the other hand, if we change the dictionaries, then we can internationalise
the change easily. Changing ` to ' in the code may be too Anglocentric.

Comment 9

[Brett Wilson]

This is already fixed in a patch I'm working on for one of the other spellcheck bugs (which one I'm not sure because it fixes many).

Comment 10

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Brett, does your patch also fix issues with other "typographic" symbols, or should I file another bug on that?  

E.g., using a real em-dash (U+2014) without spaces flags the entire word-em-dash-word as one misspelled word where using the equivalent -- does not.

Bob--the fastest man in the universe--still lost the race.  <--fine
Bob—the fastest man in the universe—still lost the race.  <-- flags

Both this and Greg's original issue are all the more glaring on Mac OS X, where the system spellchecker correctly treats (U+2019) and (U+0027) as identical and treats -- and (U+2014) as identical; it joins elements separated by the former pair and splits elements separated by the latter when spell-checking. :(

Comment 11

[Brett Wilson]

This was fixed by bug 339478.