Bug 98224 – MarkupAccumulator should optimally handle 8 bit Strings

• Component: HTML Editing
• Status: RESOLVED FIXED

[Michael Saboff]

Currently MarkupAccumulator::appendNodeValue() and the related function appendCharactersReplacingEntities() only handle UChar* string data.  These should handle 8 bit string data as well.

[Michael Saboff]

Created attachment 166798 [details]
Patch

[Eric Seidel (no email)]

Sorry, I meant rniwa, not abarth.

[Ryosuke Niwa]

Comment on attachment 166798 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=166798&action=review

> Source/WebCore/editing/MarkupAccumulator.cpp:74
> +            for (size_t m = 0; m < WTF_ARRAY_LENGTH(entityMaps); ++m) {

Please don't use abbreviations like m.

> Source/WebCore/editing/MarkupAccumulator.cpp:98
> +        const UChar* text = source.characters16() + offset;
> +
> +        size_t positionAfterLastEntity = 0;
> +        for (size_t i = 0; i < length; ++i) {
> +            for (size_t m = 0; m < WTF_ARRAY_LENGTH(entityMaps); ++m) {
> +                if (text[i] == entityMaps[m].entity && entityMaps[m].mask & entityMask) {
> +                    result.append(text + positionAfterLastEntity, i - positionAfterLastEntity);
> +                    result.append(entityMaps[m].reference);
> +                    positionAfterLastEntity = i + 1;
> +                    break;
> +                }
>              }
>          }
> +        result.append(text + positionAfterLastEntity, length - positionAfterLastEntity);

It seems like we should be able to use a template to avoid code duplication here.

> Source/WebCore/editing/markup.cpp:247
> +        MarkupAccumulator::appendCharactersReplacingEntities(buffer, content, 0, content.length(), EntityMaskInPCDATA);

StyledMarkupAccumulator inherits from MarkupAccumulator so we shouldn't have to quality it.

> Source/WebCore/editing/markup.cpp:994
> -    appendCharactersReplacingEntities(markup, title.characters(), title.length(), EntityMaskInPCDATA);
> +    MarkupAccumulator::appendCharactersReplacingEntities(markup, title, 0, title.length(), EntityMaskInPCDATA);

Ditto.

[Michael Saboff]

Created attachment 167764 [details]
Patch with some updates from reviewer's comments

(In reply to comment #3)
> (From update of attachment 166798 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=166798&action=review
> 
> > Source/WebCore/editing/MarkupAccumulator.cpp:74
> > +            for (size_t m = 0; m < WTF_ARRAY_LENGTH(entityMaps); ++m) {
> 
> Please don't use abbreviations like m.

This was from the original code, but I changed 'm' to 'entryIndex'.

> > Source/WebCore/editing/MarkupAccumulator.cpp:98
> > +        const UChar* text = source.characters16() + offset;
> > +
> > +        size_t positionAfterLastEntity = 0;
> > +        for (size_t i = 0; i < length; ++i) {
> > +            for (size_t m = 0; m < WTF_ARRAY_LENGTH(entityMaps); ++m) {
> > +                if (text[i] == entityMaps[m].entity && entityMaps[m].mask & entityMask) {
> > +                    result.append(text + positionAfterLastEntity, i - positionAfterLastEntity);
> > +                    result.append(entityMaps[m].reference);
> > +                    positionAfterLastEntity = i + 1;
> > +                    break;
> > +                }
> >              }
> >          }
> > +        result.append(text + positionAfterLastEntity, length - positionAfterLastEntity);
> 
> It seems like we should be able to use a template to avoid code duplication here.

I thought about that with the original patch.  Given there are static variables in the method that are used in the loops, either the statics would need to be duplicated in the template or a pointer to the array along with it's size needs to be passed into the templated loop function.  Both of these options seemed less desirable to having the two loops.  If you think either of these would be more readable, I could be convinced.

> > Source/WebCore/editing/markup.cpp:247
> > +        MarkupAccumulator::appendCharactersReplacingEntities(buffer, content, 0, content.length(), EntityMaskInPCDATA);
> 
> StyledMarkupAccumulator inherits from MarkupAccumulator so we shouldn't have to quality it.

Removed the class name here.

> > Source/WebCore/editing/markup.cpp:994
> > -    appendCharactersReplacingEntities(markup, title.characters(), title.length(), EntityMaskInPCDATA);
> > +    MarkupAccumulator::appendCharactersReplacingEntities(markup, title, 0, title.length(), EntityMaskInPCDATA);
> 
> Ditto.

Can't remove the class name here as this isn't a member function.

[WebKit Review Bot]

Comment on attachment 167764 [details]
Patch with some updates from reviewer's comments

Clearing flags on attachment: 167764

Committed r130795: <http://trac.webkit.org/changeset/130795>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Geoffrey Garen]

Comment on attachment 167764 [details]
Patch with some updates from reviewer's comments

View in context: https://bugs.webkit.org/attachment.cgi?id=167764&action=review

> Source/WebCore/editing/MarkupAccumulator.cpp:70
> +    if (source.is8Bit()) {
> +        const LChar* text = source.characters8() + offset;

Is there some advantage to writing this function out twice instead of using a function template?

[Michael Saboff]

(In reply to comment #7)
> (From update of attachment 167764 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=167764&action=review
> 
> > Source/WebCore/editing/MarkupAccumulator.cpp:70
> > +    if (source.is8Bit()) {
> > +        const LChar* text = source.characters8() + offset;
> 
> Is there some advantage to writing this function out twice instead of using a function template?

See the response in comment #4 to Ryosuke's comment.  Templating this function is complicated by the function static variables.

[Ryosuke Niwa]

(In reply to comment #7)
> (From update of attachment 167764 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=167764&action=review
> 
> > Source/WebCore/editing/MarkupAccumulator.cpp:70
> > +    if (source.is8Bit()) {
> > +        const LChar* text = source.characters8() + offset;
> 
> Is there some advantage to writing this function out twice instead of using a function template?

Static variables :( Let us know if you have an idea on how to share more code here.