Re: [Nano-devel] WIP: Use caching for regex

[Benno Schulenberg]

Op 03-11-18 om 04:39 schreef Devin Hussey:
> Here is my patch, I am not going to duplicate the details.
>     https://savannah.gnu.org/patch/?9713

What the heck?  How am I supposed to comment on it here if
you don't attach the patch?

> For a real example, instead of this O(n^2) strcmp tree:

??  How does a series of comparisons have a performance in the
order of n squared?  If we add two more bindable functions, on
average it will take one more comparison to rebind a function.
Linear.  Unless you assume that the user always wants to rebind
all available functions.

<Snip all other irrelevant stuff>.

Your testing of the regex caching patch is misleading.  You
include the same bunch of nanorc more than twenty times.  Yes,
if you include the same regex twentyfold, sure you are going
to see some performance improvement from hashing/caching.

The test that matters, though, is how much you gain when you
include all the nanorc files just once, as will be the case
in most circumstances.  And instead of writing the hashing
and caching stuff first, it would have been better to check
how many identical regexes there are in the current nanorcs
(that are included with nano).  And again when you delete the
near duplicate objc syntax.  Less than ten percent?  Then it's
not worth expending much effort to make this faster.  And for
sure not when it adds more than three hundred lines of code.

Benno

signature.asc
Description: OpenPGP digital signature