RFR: 8035424: (reflect) Performance problem in sun.reflect.generics.parser.SignatureParser

[Peter Levart]

Hi Claes, Andrej,

On 11/29/2016 11:53 PM, Claes Redestad wrote:
>>
>> http://cr.openjdk.java.net/~plevart/jdk9-dev/8035424_SignatureParser.performance/webrev.02/
>>
>>
>> What do you think of this one?
>
> Fair point. I think this is easier to follow and should do the trick 
> just as nicely.
>
> What you have here is fine, but in this day and age of compact strings 
> I wonder if return mar.replace('/', '.') might have both a readability 
> and a slight performance edge (assuming signatures are almost always 
> ASCII).

Good idea. It simplifies code and improves performance.


On 11/30/2016 08:38 AM, Andrej Golovnin wrote:
> Hi Peter,
>
> I know it is matter of taste, but don't you think this:
>
>   269         for (char c = current();
>   270                c != ';' && c != '.' && c != '/' && c != '[' &&
>   271                c != ':' && c != '>' && c != '<' &&
>   272                !Character.isWhitespace(c);
>   273              c = current())
>   274         {
>
> is easier to read than the long or-condition which is then negated?

Indeed it is. Another good idea, thanks.

Here's a webrev incorporating both suggestions. All 73 java/jang/reflect 
jtreg tests are passing...

http://cr.openjdk.java.net/~plevart/jdk9-dev/8035424_SignatureParser.performance/webrev.03/


Regards, Peter