[svk-devel] Refactored log filters

[David Glasser]

On 8/3/06, Michael Hendricks <michael at ndrix.org> wrote:
> The refactor has the following improvements:
>
> * log filter classes can now be subclassed
> * reduced chance of name collisions between log filters
> * cleaner code without POE-ish constants

This is great. I've really found log filters to be an awesome feature
of SVK. (And folks who haven't seen the stats filter on CPAN (which I
hadn't until yesterday) should really check it out.)

There's one thing that confuses me. It seems that there exists a
pretty clear division between output and selection filters.  Things
seem to break in odd ways (examples available on request :) ) if you
throw an output filter into the --filter argument, or a selection
filter (or for that matter a |) into the --output argument.  I don't
*think* you can ever have a filter that's both.

Shall we just make SVK::Log::Filter::Output and
SVK::Log::Filter::Selection subclasses, and enforce this?

--dave

-- 
David Glasser | glasser at mit.edu | http://www.davidglasser.net/