[Zope3-dev] Re: Making a separate project for zope.bforest

[Gary Poster]

On May 6, 2006, at 6:13 PM, Martin Aspeli wrote:

> On Sat, 06 May 2006 21:14:57 +0100, Gary Poster <gary at zope.com> wrote:
>
>> Hey all.  Seems like no-one but ZC has used zope.bforest, and it  
>> probably fits better in the "a million itsy-bitsy projects" story  
>> now that we have it (and an accompanying automated test story).   
>> Anyone object if I move it out of the trunk?
>
> A b-forest? That's like so many b-trees you can't see it for them?

:-)

Yes, it's a simple, reasonably-well-tested building block that  
aggregates multiple btrees into a single dict-like interface.  You  
can decide whether you want to see the trees, or the forest.

We've used it for two very different purposes:

- the guts for simple rotating caches (here, the "forest" is the  
engine for the public API, and the "trees" are just used internally  
as the buckets that can be cleared and rotated);

- the guts for an object that should be seen both as an aggregated  
container for all jobs, and as an object with two discrete child  
containers for active jobs and for completed jobs, respectively.

I think many of us have written similar code (especially the cache  
with rotating buckets).  It's simple code.  It's a generic task,  
though, so why not start with a shared, tested engine?

But anyway, it's only in the core because we didn't have the mini- 
package story when we packaged it: it's never been in a release, AFAIK.

Gary