couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aurélien Bénel <>
Subject Strategies for inner joins?
Date Thu, 16 Dec 2010 11:26:36 GMT
Dear all,

I'm quite aware that CouchDB developers should avoid to think about joins. However in a few
cases you cannot avoid them.

A few months ago, I implemented one with collating both "parts" in the same view and by creating
a list to filter out the rows whose keys are not in both parts. Thanks to the collation, the
algorithmic complexity of the list is pretty low (O(n)).

However, it does not scale due to the cache strategy of lists. For example, I have 200k rows
to join with 0 rows. The result is obviously 0 rows. But the problem is that it takes 70s
to figure this out EVERY time a different user ask for it.

Would the "multiview handler" be of some help in such a case?  If yes, what is the current
status of this handler? Is there any hope to see it included in CouchDB's next versions?



View raw message