Add a detachAsStream to SkDynamicMemoryWStream.

Add a detachAsStream to SkDynamicMemoryWStream.

R=reed@google.com

Committed: https://code.google.com/p/skia/source/detail?r=10171

Committed: https://code.google.com/p/skia/source/detail?r=10178

Committed: https://code.google.com/p/skia/source/detail?r=10218

[bungeman-skia, 2013-07-17 20:35:43]

This still needs some tests and I've only verified that basic things work, but
vandebo was asking to see something. The intent is to enable single copy from
SkDynamicMemoryWStream to a target buffer. Note that this does not provide for
zero copy reads like Blink's MemoryBuffer, and it would be interesting to see if
we want that instead of (or in addition to) this.

[reed1, 2013-07-17 20:58:47]

unittests?

https://codereview.chromium.org/19677002/diff/4001/include/core/SkStream.h
File include/core/SkStream.h (right):

https://codereview.chromium.org/19677002/diff/4001/include/core/SkStream.h#ne...
include/core/SkStream.h:418: void copyTo(void* dst) const;
I wish this guy would return size_t number of bytes, so it can be used with NULL
and then w/o to size the dst buffer... just saying.

[bungeman-skia, 2013-07-17 21:29:48]

Yep, as stated, this still needs a bunch of tests. Mostly I'm throwing it out
there to see if this meets the current need, or if there is something else which
would be preferable.

https://codereview.chromium.org/19677002/diff/4001/include/core/SkStream.h
File include/core/SkStream.h (right):

https://codereview.chromium.org/19677002/diff/4001/include/core/SkStream.h#ne...
include/core/SkStream.h:418: void copyTo(void* dst) const;
On 2013/07/17 20:58:47, reed1 wrote:
> I wish this guy would return size_t number of bytes, so it can be used with
NULL
> and then w/o to size the dst buffer... just saying.

That's what ::bytesWritten() is for. To clarify, this isn't just about doing one
copy, but also about being able to pass the data around looking like a SkStream.

[vandebo (ex-Chrome), 2013-07-17 21:32:02]

I think the PDF use cases only go from WStream to Stream once.  It looks to me
that this would satisfy that.

[reed1, 2013-07-18 12:53:00]

On 2013/07/17 21:29:48, bungeman1 wrote:
> Yep, as stated, this still needs a bunch of tests. Mostly I'm throwing it out
> there to see if this meets the current need, or if there is something else
which
> would be preferable.
> 
> https://codereview.chromium.org/19677002/diff/4001/include/core/SkStream.h
> File include/core/SkStream.h (right):
> 
>
https://codereview.chromium.org/19677002/diff/4001/include/core/SkStream.h#ne...
> include/core/SkStream.h:418: void copyTo(void* dst) const;
> On 2013/07/17 20:58:47, reed1 wrote:
> > I wish this guy would return size_t number of bytes, so it can be used with
> NULL
> > and then w/o to size the dst buffer... just saying.
> 
> That's what ::bytesWritten() is for. To clarify, this isn't just about doing
one
> copy, but also about being able to pass the data around looking like a
SkStream.

I know, but I can never remember that name when I'm ready to call copyTo(). I
may add this return value in addition.

[bungeman-skia, 2013-07-18 17:37:14]

Ready for review at Patch Set 3.

[reed1, 2013-07-18 20:49:28]

lgtm

[bungeman-skia, 2013-07-18 22:27:24]

Committed patchset #6 manually as r10171.

[robertphillips, 2013-07-19 00:00:16]

This was reverted in r10172 due to test failure in StreamTest
(http://108.170.217.252:10117/builders/Test-Mac10.8-MacMini4.1-GeForce320M-x86...)

[bungeman-skia, 2013-07-19 13:55:43]

Committed patchset #7 manually as r10178.

[ducky, 2013-07-19 18:22:31]

Is it supposed to be detachAsStream, or detatchAsStream (with more emphasis,
deta_t_chAsStream? Your code uses the latter consistently, but the CL title uses
the former (which the dictionary also seems to agree with more).

[bungeman-skia, 2013-07-19 18:31:04]

On 2013/07/19 18:22:31, ducky wrote:
> Is it supposed to be detachAsStream, or detatchAsStream (with more emphasis,
> deta_t_chAsStream? Your code uses the latter consistently, but the CL title
uses
> the former (which the dictionary also seems to agree with more).

Hmmm... yes, where are these reviewers. Will fix.

[bungeman-skia, 2013-07-19 22:32:13]

Committed patchset #8 manually as r10218 (presubmit successful).