Case insensitivity for tags

[ddssff]

If I've bench-marked correctly this appears to reduce performance by about 3%.

[bgamari]

It's quite neat how small this change is. Good work!

[ddssff]

Oops, I think -DCASE_INSENSITIVE should be -DCASE_INSENSITIVE=1

[ddssff]

I think the performance hit may be more like 28%. With case insensitive patch:

benchmarking Forced/html-parser
time                 10.52 ms   (10.52 ms .. 10.53 ms)
                     1.000 R²   (1.000 R² .. 1.000 R²)
mean                 10.53 ms   (10.53 ms .. 10.54 ms)
std dev              11.79 μs   (10.11 μs .. 14.39 μs)

Without case insensitive patch:

benchmarking Forced/html-parser
time                 8.180 ms   (8.174 ms .. 8.186 ms)
                     1.000 R²   (1.000 R² .. 1.000 R²)
mean                 8.182 ms   (8.179 ms .. 8.192 ms)
std dev              14.49 μs   (6.378 μs .. 26.80 μs)

[bgamari]

It looks like there are still testsuite issues, sadly.

[bgamari]

@ddssff It would be worth looking at the Core to see why the overhead here is so high.

[ddssff]

The slowdown is 100% due to the use of foldCase when building the tags, which comes down to a call to Data.Text.toCaseFold. If I substitute unsafeMk for mk it is as fast as ever. I suppose one way forward might be to make TagName a type parameter of Token and have instances for Text and CI Text. But that is a substantial change.

[ddssff]

Not sure what you mean by "Core"

[ddssff]

@bgamari the test suite issues are fixed now.