Add getFamilyNames to SkTypeface.

Add getFamilyNames to SkTypeface.

Committed: https://code.google.com/p/skia/source/detail?r=10589

Committed: https://code.google.com/p/skia/source/detail?r=10592

Committed: https://code.google.com/p/skia/source/detail?r=10608

Committed: https://code.google.com/p/skia/source/detail?r=10615

[bungeman-skia, 2013-08-05 20:29:24]

This adds getFamilyNames to SkTypeface. Most of the implementations are fairly
straight forward, I imagine most of the interest here will be about how to
expose the API. Currently I'm just returning an SkRefCnt object so that the
state can be carried out and deleted in a sane fashion. If there's a better
means of exposing this sort of thing, that would be of interest.

[reed1, 2013-08-05 20:45:34]

https://codereview.chromium.org/21716005/diff/15001/include/core/SkTypeface.h
File include/core/SkTypeface.h (right):

https://codereview.chromium.org/21716005/diff/15001/include/core/SkTypeface.h...
include/core/SkTypeface.h:230: virtual bool next() = 0;
I vote for a reset() as well -- expected pattern for other skia iterators, and
trivial to implement.

https://codereview.chromium.org/21716005/diff/15001/include/core/SkTypeface.h...
include/core/SkTypeface.h:232: virtual const SkString& getLanguage() const = 0;
Returning strings forces the impl to really have these objects. How about either
returning const char*, or have a struct that gives it all back at once.

e.g.

struct Rec {
    const char* fName;
    const char* fLanguage;
    size_t fNameLength;
    size_t fLanguageLength;
};

bool next(Rec*);

[bungeman-skia, 2013-08-06 15:48:22]

https://codereview.chromium.org/21716005/diff/15001/include/core/SkTypeface.h
File include/core/SkTypeface.h (right):

https://codereview.chromium.org/21716005/diff/15001/include/core/SkTypeface.h...
include/core/SkTypeface.h:230: virtual bool next() = 0;
On 2013/08/05 20:45:34, reed1 wrote:
> I vote for a reset() as well -- expected pattern for other skia iterators, and
> trivial to implement.

I went and looked around for an example, and I can't find any. reset() is common
on array like things, and means discard your data. SkStream::rewind() exists and
seems to be the closest thing, but even then its part of the 'extended'
interface. I think to be consistent such a method would be called rewind, but
currently we have no use case. It would be easy enough to add later if we need
it.

https://codereview.chromium.org/21716005/diff/15001/include/core/SkTypeface.h...
include/core/SkTypeface.h:232: virtual const SkString& getLanguage() const = 0;
On 2013/08/05 20:45:34, reed1 wrote:
> Returning strings forces the impl to really have these objects. How about
either
> returning const char*, or have a struct that gives it all back at once.
> 
> e.g.
> 
> struct Rec {
>     const char* fName;
>     const char* fLanguage;
>     size_t fNameLength;
>     size_t fLanguageLength;
> };
> 
> bool next(Rec*);

Added a LocalizedString struct.

[bungeman-skia, 2013-08-06 22:53:58]

Committed patchset #5 manually as r10589 (presubmit successful).

[bungeman-skia, 2013-08-07 02:45:30]

Committed patchset #6 manually as r10592 (presubmit successful).

[reed1, 2013-08-07 12:06:41]

https://codereview.chromium.org/21716005/diff/26002/include/core/SkTypeface.h
File include/core/SkTypeface.h (right):

https://codereview.chromium.org/21716005/diff/26002/include/core/SkTypeface.h...
include/core/SkTypeface.h:242: LocalizedStrings* getFamilyNames() const;
Perhaps copyFamilyNames() would be clearer that the returned ptr is now owned by
the caller. Our getters almost never carry that obligation, and we use create or
new ore ref when ownership events are happening.

[bungeman-skia, 2013-08-07 15:46:33]

Committed patchset #7 manually as r10608 (presubmit successful).

[robertphillips, 2013-08-07 15:57:12]

This was reverted in r10609 due to compilation errors.

[bungeman-skia, 2013-08-07 17:09:26]

Committed patchset #8 manually as r10615 (presubmit successful).