On Mon, 19 Sep 2022 22:18:42 GMT, Naoto Sato <na...@openjdk.org> wrote:
>> Yeah, I saw there's a mismatch between the src and limit in this call that
>> led to the index check Exception:
>> 286 for (int b = ci.getBeginIndex(); b < end;) {
>> 287 boundaries.add(b);
>> 288 b = Grapheme.nextBoundary(text, b, end);
>> 289 }
>>
>> and nextBoundary could walk through the entire CharSequence. Maybe it's
>> worth a note to the length() method.
>
> Thanks, Joe. Added comment to the method.
OK yeah this is really confusing. One might ask a similar question in `charAt`
about why it doesn't call
src.setIndex(index + src.getBeginIndex());
The answer is that this is a special-purpose `CharSequence` that represents
characters in the index range [0..endIndex) of the underlying
`CharacterIterator`, **even if** that `CharacterIterator` represents the
subrange of some string. As Joe noted, the calling code in
`GraphemeBreakIterator` takes care to ensure that only valid indexes into the
`src` are used.
I think the comment on `length` helps a little bit but maybe a class-level
comment would be better, since it applies to the whole model of this special
`CharSequence` and not just the `length` method. Just a sentence or two is
sufficient. Feel free to crib from the above.
-------------
PR: https://git.openjdk.org/jdk/pull/10349
