Glenn Fowler requests:
 > if it works post a timing comparison

Here are some timing tests. For ksh, I used the dgk code posted later.
I ran each test listed below 3 times. The results below are the middle
time. The file I used is the zsh configure script, because:

   * it is over 20,000 lines long, so 
   * it is representative of things I had in mind and 
   * I had previously used and reported on that elsewhere.

wc says of this file:
21124  67719 555992 ./zsh/configure

I ran the tests below on an Ubuntu 8.3 thinkpad laptop with an Intel
Centrino CPU: 
  Model: 6.15.13 "Intel(R) Core(TM)2 Duo CPU T5450 @ 1.66GHz"

First, the fast ksh code with the IFS and glob settings:

$ time /tmp/kshrf.sh

real 0m0.099s
user 0m0.080s
sys  0m0.008s

This is a comparable and a little faster than the code for zsh:

$ time /tmp/zshrf.sh

real 0m0.193s
user 0m0.176s
sys  0m0.008s

But what's also impressive is that even the slow ksh code isn't all
that slow:

$ time /tmp/kshrs.sh
real    0m1.456s
user    0m0.760s
sys     0m0.696s

Especially when you compare with zsh:

real    0m15.874s
user    0m1.420s
sys     0m0.092s

or bash:

real    0m4.980s
user    0m4.208s
sys     0m0.760s


But it looks like the custom array reading code I wrote does pretty
good here:

real 0m0.032s
user 0m0.024s
sys  0m0.008s

(I wrote this builtin out of frustration and desperation, not a desire
to get the last ounce of speed. At the time, the speed difference was
more on the order of minutes than seconds. So it's possible bash has
improved or I was encountering files even larger than 20,000 or I had
a slower computer. For zshdb and kshdb I am very happy with the
results reported above.)

Thanks, everyone!
_______________________________________________
ast-users mailing list
[email protected]
https://mailman.research.att.com/mailman/listinfo/ast-users

Reply via email to