RFR: 8248655: Support supplementary characters in String case insensitive operations

[Roger Riggs]

Hi Naoto,

Looks fine. (with Joe's suggestion)

On 7/22/20 4:20 PM, Joe Wang wrote:
> Hi Naoto,
>
> The change looks good to me. "supLower" is indeed super slow :-)
>
> The only minor comment I have is that the compareCodePointCI method 
> performs toUpperCase unconditionally. That's not a problem for the 
> regular case, where a check on cp1 == cp2 (line 337) is done prior to 
> the method call. But for the sup case (starting at line 341), the 
> method is called unconditionally while in webrev.04 there was a check 
> "cp1 != cp2".  One option to fix it is to include the "cp1 != cp2" 
> check in the method compareCodePointCI, then cp1 == cp2 at line 337 
> can be omitted.
I would have added to line 353 the same cp1 == cp2 check as 337 to avoid 
the method call
unless it was needed.

Roger

>
> Regards,
> Joe
>
> On 7/22/20 10:23 AM, naoto.sato at oracle.com wrote:
>> Hi,
>>
>> I revised the fix again, based on further suggestions:
>>
>> https://cr.openjdk.java.net/~naoto/8248655.8248434/webrev.05/
>>
>> Changes from v.04 are (all in StringUTF16.java):
>>
>> - The short cut now does case insensitive comparison that makes the 
>> fix closer to the previous implementation (for BMP characters).
>> - Changed the bit operation to negating for detecting needed index 
>> increment.
>> - Method name is changed to better reflect what it is doing, with 
>> more descriptive comments.
>>
>> Here is the benchmark results:
>>
>> before:
>> Benchmark                                Mode  Cnt   Score Error  Units
>> StringCompareToIgnoreCase.lower          avgt   25  49.960 ? 1.923  
>> ns/op
>> StringCompareToIgnoreCase.supLower       avgt   25  21.003 ? 0.354  
>> ns/op
>> StringCompareToIgnoreCase.supUpperLower  avgt   25  30.863 ? 4.529  
>> ns/op
>> StringCompareToIgnoreCase.upperLower     avgt   25  15.417 ? 1.046  
>> ns/op
>>
>> after:
>> Benchmark                                Mode  Cnt    Score Error  Units
>> StringCompareToIgnoreCase.lower          avgt   25   46.857 ? 0.524  
>> ns/op
>> StringCompareToIgnoreCase.supLower       avgt   25  148.688 ? 6.546  
>> ns/op
>> StringCompareToIgnoreCase.supUpperLower  avgt   25   37.160 ? 0.259  
>> ns/op
>> StringCompareToIgnoreCase.upperLower     avgt   25   15.126 ? 0.338  
>> ns/op
>>
>> Now non-supplementary operations ("lower" and "upperLower") are on 
>> par with the "before" result (I am not quite sure why the "after" 
>> results are somewhat faster though). For supplementary test cases, 
>> "supLower" is very slow. The reason is two fold; one is because 
>> "before" one exits at the very first character (which I am addressing 
>> here) while "after" continues to compare to the last characters, the 
>> other reason is the test suffers from the change where supplementary 
>> cases double the case insensitivity checks (compared to the "after" 
>> result just below). Also "supUpperLower" gets slower for the same 
>> reason. These are expected results for supplementary comparisons (as 
>> we discussed).
>>
>> Naoto
>>
>> On 7/17/20 4:36 PM, naoto.sato at oracle.com wrote:
>>> Hi,
>>>
>>> Based on the suggestions, I modified the fix as follows:
>>>
>>> https://cr.openjdk.java.net/~naoto/8248655.8248434/webrev.01/
>>>
>>> Changes from the initial revision are:
>>>
>>> - Shared the implementation between compareToCI() and regionMatchesCI()
>>> - Enabled immediate short cut if two code points match.
>>> - Created a simple JMH benchmark. Here is the scores before and 
>>> after the change:
>>>
>>> before:
>>> Benchmark                                Mode  Cnt   Score Error  Units
>>> StringCompareToIgnoreCase.lower          avgt   25  53.764 ? 2.811  
>>> ns/op
>>> StringCompareToIgnoreCase.supLower       avgt   25  24.211 ? 1.135  
>>> ns/op
>>> StringCompareToIgnoreCase.supUpperLower  avgt   25  30.595 ? 1.344  
>>> ns/op
>>> StringCompareToIgnoreCase.upperLower     avgt   25  18.859 ? 1.499  
>>> ns/op
>>>
>>> after:
>>> Benchmark                                Mode  Cnt   Score Error  Units
>>> StringCompareToIgnoreCase.lower          avgt   25  58.354 ? 4.603  
>>> ns/op
>>> StringCompareToIgnoreCase.supLower       avgt   25  57.975 ? 5.672  
>>> ns/op
>>> StringCompareToIgnoreCase.supUpperLower  avgt   25  23.912 ? 0.965  
>>> ns/op
>>> StringCompareToIgnoreCase.upperLower     avgt   25  17.744 ? 0.272  
>>> ns/op
>>>
>>> Here, "sup" means all supplementary characters, BMP otherwise. 
>>> "lower" means each character requires upper->lower casemap. 
>>> "upperLower" means all characters are the same, except the last 
>>> character which requires casemap.
>>>
>>> I think the result is reasonable, considering surrogates check are 
>>> now mandatory. I have tried Roger's suggestion to use 
>>> Arrays.mismatch() but it did not seem to benefit here. In fact, the 
>>> performance degraded partly because I implemented the short cut, and 
>>> possibly for the overhead of extra checks.
>>>
>>> Naoto
>>>
>>> On 7/15/20 9:00 AM, naoto.sato at oracle.com wrote:
>>>> Hello,
>>>>
>>>> Please review the fix to the following issues:
>>>>
>>>> https://bugs.openjdk.java.net/browse/JDK-8248655
>>>> https://bugs.openjdk.java.net/browse/JDK-8248434
>>>>
>>>> The proposed changeset and its CSR are located at:
>>>>
>>>> https://cr.openjdk.java.net/~naoto/8248655.8248434/webrev.00/
>>>> https://bugs.openjdk.java.net/browse/JDK-8248664
>>>>
>>>> A bug was filed against SimpleDateFormat (8248434) where 
>>>> case-insensitive date format/parse failed in some of the new 
>>>> locales in JDK15. The root cause was that case-insensitive 
>>>> String.regionMatches() method did not work with supplementary 
>>>> characters. The problem is that the method's spec does not expect 
>>>> case mappings of supplementary characters, possibly because it was 
>>>> overlooked in the first place, JSR 204 - "Unicode Supplementary 
>>>> Character support". Similar behavior is observed in other two 
>>>> case-insensitive methods, i.e., compareToIgnoreCase() and 
>>>> equalsIgnoreCase().
>>>>
>>>> The fix is straightforward to compare strings by code point basis, 
>>>> instead of code unit (16bit "char") basis. Technically this change 
>>>> will introduce a backward incompatibility, but I believe it is an 
>>>> incompatibility to wrong behavior, not true to the meaning of those 
>>>> methods' expectations.
>>>>
>>>> Naoto
>