208789 - "text-transform: capitalize" needs to understand word boundaries

Status: RESOLVED FIXED

(Core :: Layout: Text and Fonts, defect, P4)

Description

[Travis Seitler]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.4b) Gecko/20030516 Mozilla Firebird/0.6
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.4b) Gecko/20030516 Mozilla Firebird/0.6

In
http://www.rnd.knoxcounty.org/purchasing/testing/supplierdiversityprogram.html,
near the bottom of the page, the heading 'What are "small and disadvantaged
businesses"?' has the 'text-transform: capitalize' rule applied, and so it
should appear as:

What Are "Small And Disadvantaged Businesses"?

In Firebird 0.6, the word "small" is not capitalized.  It appears that Firebird
sees the quote character as the first letter of the word, and tries to
"capitalize" the quote instead of the character following it.

Reproducible: Always

Steps to Reproduce:
1. Visit page with text-transform: capitalize applied to text within quotes,
such as
http://www.rnd.knoxcounty.org/purchasing/testing/supplierdiversityprogram.html
Actual Results:  
The heading appeared as:

What Are "small And Disadvantaged Businesses"?

Expected Results:  
Recognize that the quote character is not a letter, and capitalize the letter
following it instead.

Comment 1

[David Tenser [:djst]]

Not a Mozilla Firebird-specific bug. It's most probably a (known?) Gecko bug.

Comment 2

[Jo Hermans]

Well, the bug is confirmed in Mozilla too (build 2003060503 on Mac OS X 10.2.6),
but I can't find any comments in the CSS-spec about it. It must be a cultural
thing too, but I think that skipping any " or ' prefix might be ok. Maybe other
characters too ?

The code that has to be changed is in
http://lxr.mozilla.org/seamonkey/source/layout/html/base/src/nsTextTransformer.cpp
(search for NS_STYLE_TEXT_TRANSFORM_CAPITALIZE). We might need something like :

static void
AsciiToTitleCase(unsigned char* aText, PRInt32 aWordLen)
{
   while (aWordLen-- > 0) {
     if ( (*aText == '"') || (*aText == '\'') )
        continue;
     *aText = toupper(*aText);
     break;
   }
}

and we'll have to fix nsCaseConversionImp2::ToTitle() too (see
http://lxr.mozilla.org/seamonkey/source/intl/unicharutil/src/nsCaseConversionImp2.cpp#236)

Comment 3

[Christopher Hoess (gone)]

IIRC, there should already be code to achieve the same effect with :first-letter
(ensuring that a letter following a punctuation character gets the :first-letter
style), although I think it's only semi-funcitonal right now. Sharing that code
should probably be investigated.

Comment 4

[Neil Parks]

It isn't only quotation marks.  Capitalization also fails to occur after other
punctuation such as a period or an open parenthesis.  

For examples see:

http://forums.mozillazine.org/viewtopic.php?t=46482&highlight=

Comment 5

[Simon Montagu :smontagu]

*** Bug 273360 has been marked as a duplicate of this bug. ***

Comment 6

[Uri Bernstein]

*** Bug 345456 has been marked as a duplicate of this bug. ***

Comment 7

[Zack Weinberg (:zwol)]

Attachment: test case

The present code is structured rather differently, and I'm not at all sure how it manages to work - but it does the Right Thing for all the examples I can find, except arguably "c.a.g.e" -> "C.a.g.e" (and not "C.A.G.E") - I attach a test case and reference.

Comment 8

[Zack Weinberg (:zwol)]

Attachment: reference rendering

Comment 9

[Mike]

Any status on this bug? It works for most layouts, but still not working properly for some instances, such as the example above c.a.g.e -> still gives C.a.g.e (should be C.A.G.E) as well as super-size -> gives Super-size (should be Super-Size).

Appears to work properly in Safari 5 and Chrome 6.

Maybe a simple workaround might be to just add a check for a few specific, common symbols and punctuation until a proper fix can be written.

Comment 10

[Mike]

< - BUMP ? - >

Still not working in some instances for FF v 17.0.1.

Example: 6-figure ...should be... 6-Figure


IDEA ? (I'm not a coder, outside of HTML/PHP/JavaScript ;-)
but hopefully what I write will make sense to a coder...

There are only 26 lowercase characters on the English keyboard.
What about checking something along the lines of:

if ("abcdefghijklmnopqrstuvwxyz").indexOf(theFirstLetter) == 0) //if the first char is lowercase
    { upperCase(%TheFirstLetter); }


---------------------
Granted, this may be more difficult in other languages and may be processor heavy.
Just an idea.

Thanks for your great work!
-Mike

Comment 11

[Jonathan Kew [:jfkthame]]

I think we should consider adjusting the behavior here, so that a letter after word-internal punctuation gets capitalized (so "Word-Internal" rather than "Word-internal", if applied to this sentence). My sense is that this is more often the desired result than not, and it will align us more closely with how webkit/blink-based browsers behave.

(I'll note, though, that whatever we do, there will inevitably be cases where the result turns out to be inferior. Compare

data:text/html,<span style="text-transform: capitalize;">how does your country spell "colo[u]r"?</span>

between Firefox and the other browsers. There's no perfect solution here.)

Comment 12

[Jonathan Kew [:jfkthame]]

Attachment: Bug 208789 - patch 1 - Create an intl::GeneralCategory enum for UnicodeProperties::CharType() to return, to avoid directly referring to ICU4C constants or mapping via harfbuzz constants. r=#platform-i18n-reviewers

No change in behavior; this just gives us our own version of the general category constants,
so we can avoid depending on ICU's constants elsewhere in the codebase.

Comment 13

[Jonathan Kew [:jfkthame]]

Attachment: Bug 208789 - patch 2 - Make text-transform:capitalize act on a letter that follows punctuation without a space, as in "Tip-Top" or "What.Three.Words". r=#layout-reviewers

This implements the adjustment to our behavior, bringing us closer to the other browsers
(although with a better result for examples like "colo[u]r", which we continue to treat
as one word rather than three).

Simple reftest included; for now, I've put it with our in-house tests rather than under
web-platform-tests because the exact behavior here is somewhat under-specified.

Depends on D173203

Comment 14

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/cc59256f5ef0
patch 1 - Create an intl::GeneralCategory enum for UnicodeProperties::CharType() to return, to avoid directly referring to ICU4C constants or mapping via harfbuzz constants. r=platform-i18n-reviewers,nordzilla
https://hg.mozilla.org/integration/autoland/rev/c5bd0893afc5
patch 2 - Make text-transform:capitalize act on a letter that follows punctuation without a space, as in "Tip-Top" or "What.Three.Words". r=emilio

Comment 15

[Norisz Fay [:noriszfay]]

https://hg.mozilla.org/mozilla-central/rev/cc59256f5ef0
https://hg.mozilla.org/mozilla-central/rev/c5bd0893afc5