python - Performance of Pypy's islower(), isupper() -

i have program in str.isupper() called in tight loop, , takes around 25% of running time.

it strikes me isupper , islower slower other isx methods, such isalnum, isalpha, isdigit , on. (note: in cpython methods equally slow.)

$ pypy -mtimeit "'a'.islower()"       - 0.0158  usec per loop $ pypy -mtimeit "'a'.isalpha()"       - 0.00251 usec per loop $ pypy -mtimeit "'a'.isupper()"       - 0.0164  usec per loop $ pypy -mtimeit "'a'.isdigit()"       - 0.00236 usec per loop 

i can make faster isupper method myself by

isupper = lambda c: 'a' <= c <= 'z' 

which runs in reasonable 0.00393 usec per loop. of course version doesn't 'work' unicode strings, strings 8bit ascii.

---

my question if there way 'tell' python skip fancy unicode tests when working simple 'a-z'?

i don't know reason difference, know benchmarks flawed. expression "a".islower() in loop constant-folded away. disappears... might time pass.

we can check "a".isalpha() , pass execute @ same speed: it's speed of empty loop. "a".islower(), i'm not sure understand, it's slower me too, , not --- though it's measuring speed of empty loop, can check (see "jitviewer" if you're interested). mysteries of cpus?

(note has been measured on recent pypy; time ago islower() , isupper() suffered missing optimization, , not constant-folded.)