Separate and update PDF_DIFF_TRACE_IN_PNG

Separate and update PDF_DIFF_TRACE_IN_PNG

Move its functionality out of readToken() and into its own class.
Callers of the previous readToken() now call
SkPdfNativeTokenizer::readToken(), which in turn calls a function
for writing the diff to a file, if the caller requests it and
PDF_TRACE_DIFF_IN_PNG is defined.

Do not attempt to draw a diff for compatibility sections, which we
do not draw.

Use SkString to handle string manipulation.

Hide globals only used by PDF_TRACE_DIFF_IN_PNG behind that flag.

Remove hasVisualEffects, which always returns true.

Rename gLastOpKeyword to gOpCounter for clarity.

In SkPdfNativeTokenizer, set fEmpty to true when the entire stream
has been read.

Use SkBitmap::copyTo instead of manually copying an SkBitmap.

Builds on https://codereview.chromium.org/79933003/

R=mtklein@google.com

Committed: https://code.google.com/p/skia/source/detail?r=12436

[mtklein, 2013-11-22 14:46:12]

Feel free to ignore all the comments in the first CL that this one makes moot.
:)

https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/i...
File experimental/PdfViewer/inc/SkPdfDiffEncoder.h (right):

https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/i...
experimental/PdfViewer/inc/SkPdfDiffEncoder.h:20: static void
WriteToFile(PdfToken*);
Do you anticipate this class will become more than just a function?

If not, or at least not for now, I'd just change this to

namespace SkPdfDiffEncoder {
void WriteToFile(PdfToken*);
}

to make it clear SkPdfDiffEncoder is a namespace.  I don't think the callers
would have to change at all.

https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/p...
File experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h (right):

https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/p...
experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h:170: bool
readToken(PdfToken* token, bool write = false);
maybe writeDiff?

[scroggo, 2013-11-22 15:47:16]

https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/i...
File experimental/PdfViewer/inc/SkPdfDiffEncoder.h (right):

https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/i...
experimental/PdfViewer/inc/SkPdfDiffEncoder.h:20: static void
WriteToFile(PdfToken*);
On 2013/11/22 14:46:12, mtklein wrote:
> Do you anticipate this class will become more than just a function?
> 
> If not, or at least not for now, I'd just change this to
> 
> namespace SkPdfDiffEncoder {
> void WriteToFile(PdfToken*);
> }
> 
> to make it clear SkPdfDiffEncoder is a namespace.  I don't think the callers
> would have to change at all.

reed@ encouraged me to use a class in another instance  for consistency- see
(https://codereview.chromium.org/25581002/diff/1/include/utils/SkFrontBuffered...).

As discussed in person, I think we're content to do either one.

https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/p...
File experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h (right):

https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/p...
experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h:170: bool
readToken(PdfToken* token, bool write = false);
On 2013/11/22 14:46:12, mtklein wrote:
> maybe writeDiff?

Done.

[mtklein, 2013-11-22 15:51:49]

On 2013/11/22 15:47:16, scroggo wrote:
>
https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/i...
> File experimental/PdfViewer/inc/SkPdfDiffEncoder.h (right):
> 
>
https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/i...
> experimental/PdfViewer/inc/SkPdfDiffEncoder.h:20: static void
> WriteToFile(PdfToken*);
> On 2013/11/22 14:46:12, mtklein wrote:
> > Do you anticipate this class will become more than just a function?
> > 
> > If not, or at least not for now, I'd just change this to
> > 
> > namespace SkPdfDiffEncoder {
> > void WriteToFile(PdfToken*);
> > }
> > 
> > to make it clear SkPdfDiffEncoder is a namespace.  I don't think the callers
> > would have to change at all.
> 
> reed@ encouraged me to use a class in another instance  for consistency- see
>
(https://codereview.chromium.org/25581002/diff/1/include/utils/SkFrontBuffered...).
> 
> As discussed in person, I think we're content to do either one.
> 
>
https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/p...
> File experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h (right):
> 
>
https://codereview.chromium.org/80463005/diff/150001/experimental/PdfViewer/p...
> experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h:170: bool
> readToken(PdfToken* token, bool write = false);
> On 2013/11/22 14:46:12, mtklein wrote:
> > maybe writeDiff?
> 
> Done.

lgtm

[scroggo, 2013-12-02 20:22:39]

Committed patchset #6 manually as r12436 (presubmit successful).