RFR: 8300207: Add a pre-check for the number of canonical equivalent permutations in j.u.r.Pattern [v2]

[Stuart Marks]

On Tue, 17 Jan 2023 12:23:25 GMT, Alan Bateman <alanb at openjdk.org> wrote:

>> The choice of a `<p>` paragraph rather than `@apiNote` is for consistency with similar commentary paragraphs in the specs of `CASE_INSENSITIVE`, `UNICODE_CASE`, and `UNICODE_CHARACTER_CLASS`.
>> 
>> I have no problems in using `@apiNote` instead, but then it would be better to apply the same for the other mentioned flags as well.
>
> Okay, I see your point and to use apiNote consistently would require "converting" some of the existing text to apiNote too.
> 
> I'm still mulling over Pattern.compile throwing OOME. An implNote is probably the right category for this, in which case it can start with "In the the JDK Reference Implementation ...".  I assume the static Pattern.matches needs same, and also Pattern.matcher for the lazy compilation case.

@AlanBateman

I had previously discussed with @rgiulietti whether OOME or PatternSyntaxException was more appropriate. The issue is that you might try to compile a pattern that contains a character with N combining diacritics, and it might work fine. But if you change that character to have N+1 combining diacritics, it might throw OOME. There's no syntax difference, but rather the issue is hitting an internal implementation limit.

There's a bit of a precedent for throwing OOME in such cases. Various places in the library try to grow arrays. If the required array size is greater than MAX_VALUE, the library pre-emptively throws OOME without even trying to allocate the array.

-------------

PR: https://git.openjdk.org/jdk/pull/12027