Re: IncrementalFaultList behavior

From: Andrus Adamchik (andru..bjectstyle.org)
Date: Mon Sep 08 2003 - 11:32:15 EDT

  • Next message: Fabricio Voznika: "Re: Object validation"

    On Monday, September 8, 2003, at 11:07 AM, Scott Finnerty wrote:

    > However, some times I would like to use the Iterator to process the
    > result
    > set, but this always forces a full retrieval and in the case of large
    > result
    > sets with PostgreSQL, a subsequent failure. Monkeying with the
    > PostgreSQL
    > parameter is not an ideal solution. How about returning a custom
    > Iterator
    > that resolves pages at a time as the first element in an unresolved
    > page is
    > requested from the Iterator? This would space out the hits to the
    > database
    > in page size chunks and should address the problem with PostgreSQL?
    > I'd be
    > willing to code up such a modification to IncrementalFaultList if this
    > solution sounds acceptable.

    Hi Scott,

    +1 for a lazy iterator. I guess I never cared too much about this in
    the past, since I always used sublists. But your point is valid, and
    your patch is very welcomed.

    As for the huge IN query that happens if "resolveAll" is called on
    IncrementalFaultList, I think we can patch it too, independently from
    the Iterator. It may run a SELECT...WHERE...IN in chunks of reasonable
    (configurable) size that should be bigger than a typical page size for
    better performance. I'll open a bug for this second part, but will
    probably mark it as "after 1.0".

    Andrus



    This archive was generated by hypermail 2.0.0 : Mon Sep 08 2003 - 11:29:43 EDT