Chunking : Matching Issue

[Alexis EidelMan]

Hi everyone,

I saw in the longterm "cleanup" roadmap of the TODO file in the 0.6 release the will of chunking all operation.
I want to point out there is a special issue about the matching function. In that particular case, working with all the database or just with a part of it will produce different results.

[Gaëtan de Menten]

I am aware of this. This would only be a problem if a naive chunking is
done. If for each "women", we compute the best score for *all* men
(using chunking or not), this will produce the same result as we have
currently.

Besides, this is probably the last function where I will implement
chunking, as if you have so many objects in your "filtered" sets (the
actual sets to match) that need chunking, you are in trouble anyway
because the matching will take forever...

FWIW, I am aware of the different challenges that will be created by
implementing all operations using chunking. This will be hard to get
right (and efficient) but not impossible.

Ga�tan


----------------------------------------------------------------------------

Disclaimer: please see "www.plan.be/disclaimer.html"

Please consider your environmental responsibility before printing this email

----------------------------------------------------------------------------