Note asciiCI badness
using `asciiCI` regresses performance terribly, nearly doubling runtime
and significantly increasing code size. Didn't look into why.
```
-rw-r--r-- 1 ben ben 2.2M Apr 6 13:48 Text/HTML/Parser.o
$ ./Benchmark "length/mine Text" +RTS -t -rticky
benchmarking length/mine Text
time 84.76 ms (82.53 ms .. 87.48 ms)
0.998 R² (0.993 R² .. 1.000 R²)
mean 88.22 ms (86.39 ms .. 91.87 ms)
std dev 4.142 ms (2.397 ms .. 6.218 ms)
<<ghc: 12785825784 bytes, 24417 GCs, 4386215/4663584 avg/max bytes
residency (72 samples), 17M in use, 0.000 INIT (0.000 elapsed), 6.891
MUT (6.906 elapsed), 0.489 GC (0.488 elapsed) :ghc>>
```
commitdd116a6da69f0f3788f308cdb008cec8f50d5aa9
authorBen Gamari <ben@smart-cactus.org>
dateWed 6 Apr 2016, 13:49 (11 years ago)
parent3f61cb1
Text/HTML/Parser.hs+1
@@ -182,6 +182,7 @@ markupDeclOpen =
182182 where
183183 comment = string "--" >> commentStart
184184 docType = do
185+ -- switching this to asciiCI slowed things down by a factor of two
185186 s <- take 7
186187 guard $ T.toLower s == "doctype"
187188 doctype