Re: BK-kernel-tools/shortlog update

[Jeff Garzik <jgarzik@pobox.com>]

On Wed, Mar 26, 2003 at 09:10:31PM +0100, Matthias Andree wrote:
> On Wed, 26 Mar 2003, Linus Torvalds wrote:
> 
> > Btw, one feature I'd like to see in shortlog is the ability to use 
> > regexps for email address matching, ie something like
> > 
> > 	'torvalds@.*transmeta.com' => 'Linus Torvalds'
> > 	... 
> > 	'alan@.*swansea.linux.org.uk' => 'Alan Cox'
> > 	...
> > 	'bcrl@redhat.com' => 'Benjamin LaHaise',
> > 	'bcrl@.*' => '?? Benjamin LaHaise',
> > 	..
> > 
> > I don't know whether you can force perl to do something like this, but if 
> > somebody were to try...

Perl is very regex-friendly.  Sure it can do this :)


> I'd like to keep the hash for all those addresses that aren't wildcards
> and that aren't regexps -- we have fast, that is O(1) to O(log n),
> access to the hash (depending on Perl's implementation) and we have
> worse than O(n) for regexp, where n is the count of address strings or
> regexps.
> 
> Would you agree to a version that has a set of fixed addresses and a
> separate list of regexps, tries the hash first and then a list of
> regexps?  That sounds like a) easy addition, b) good performance to me
> (before implementing it). If so, I could add some code for that feature.

Do we really care about performance here?

I think maintain-ability is probably more important.

In any case, splitting the lists into "fixed" and "regex" doesn't seem
like a bad idea, provided that the change was fairly easy and
self-contained.

	Jeff