Some time ago I have opened a thread about optimization in Python. I
have now done a simple benchmark and used Python for that too. If s.o is
interested, take a look at: http://edt.vol.cz:81/bench
Any comment is very welcome. I got an answer from the ISE-Eiffel guys
this morning ;-) You might imagine what they think. If any of you have
simular tests, pleas let me know, allow me to put that copy onto that
page. I hope we can found some points which speak for one language under
certain conditions. It would be really helpful if you would include
information about
1) implementation time
2) debug time
3) sort of bug
4) maybe a short estimate how important speed is for that special test
I didn't have done that very detailed. And I now regret that. I just can
give a rought scetch
1) Perl (stolen from an implementation form Markus Mottl on the
OCAML-Mailing-list)
2) OCAML (the same as for Perl)
3) Python first solution < 3 Minutes I guess, If I remember correctly
no bug at all (sheer luck, normally I produce at most on typo ;-))
4) optimizations (<1 minute each, but the last)(typos)
5) fastest solution (point of discussion in this group so this take
quite some time but not for implementation but discussion)
6) Eiffel sol. Because I didn't know gelex to that time I need some more
time the source was done very fast, but not very Eiffelish. No
assertions and the like
(but all implemantations take me more than a week). Some parts have to
be completely new written (split_string which is of course easy to use
in Python)
7) C. This was not difficult to do after all the other stuff, but I
didn't have a hash-table implementation at hand. Found one in the conde
snippets. But found too bugs in it. So adaption take me quite a while.
I can't tell how much time I need for debugging. Either estimate would
be much to low (I woul look like a hacker, I'm not) or to high (would
look like an idiot, but who cares ;-) And I'm using a debugger just to
look into the code that I can see if anything works as inteded even
without a direct bug. I don't know if that is good style, but I feel
comfortable with that.
Regards
Friedrich
have now done a simple benchmark and used Python for that too. If s.o is
interested, take a look at: http://edt.vol.cz:81/bench
Any comment is very welcome. I got an answer from the ISE-Eiffel guys
this morning ;-) You might imagine what they think. If any of you have
simular tests, pleas let me know, allow me to put that copy onto that
page. I hope we can found some points which speak for one language under
certain conditions. It would be really helpful if you would include
information about
1) implementation time
2) debug time
3) sort of bug
4) maybe a short estimate how important speed is for that special test
I didn't have done that very detailed. And I now regret that. I just can
give a rought scetch
1) Perl (stolen from an implementation form Markus Mottl on the
OCAML-Mailing-list)
2) OCAML (the same as for Perl)
3) Python first solution < 3 Minutes I guess, If I remember correctly
no bug at all (sheer luck, normally I produce at most on typo ;-))
4) optimizations (<1 minute each, but the last)(typos)
5) fastest solution (point of discussion in this group so this take
quite some time but not for implementation but discussion)
6) Eiffel sol. Because I didn't know gelex to that time I need some more
time the source was done very fast, but not very Eiffelish. No
assertions and the like
(but all implemantations take me more than a week). Some parts have to
be completely new written (split_string which is of course easy to use
in Python)
7) C. This was not difficult to do after all the other stuff, but I
didn't have a hash-table implementation at hand. Found one in the conde
snippets. But found too bugs in it. So adaption take me quite a while.
I can't tell how much time I need for debugging. Either estimate would
be much to low (I woul look like a hacker, I'm not) or to high (would
look like an idiot, but who cares ;-) And I'm using a debugger just to
look into the code that I can see if anything works as inteded even
without a direct bug. I don't know if that is good style, but I feel
comfortable with that.
Regards
Friedrich