Closed (cannot reproduce)
Project:
Project
Version:
6.x-1.x-dev
Component:
Releases
Priority:
Normal
Category:
Task
Assigned:
Unassigned
Reporter:
Created:
26 Nov 2009 at 02:03 UTC
Updated:
12 Nov 2015 at 03:16 UTC
Jump to comment: Most recent, Most recent file
Comments
Comment #1
marvil07 commentedI originally used
project_releases_version_minor_version_patch_version_extra_weight_version_extraas the name for one of the indexes, but it gave me a Identifier name is too long error, so I usedproject_releases_versions.Comment #2
damien tournoud commentedDo you have examples of the queries you want to optimize?
The database engine cannot use an index to optimize a ORDER BY if it hasn't used the same index to filter the rows, so you might need to add more columns to your second index.
Comment #3
marvil07 commentedActually I did not take a look to the queries mentioned in #642094: Move the code to build the query that drives project_release_find_latest_releases() into a shared helper function, and it seems that this need to be re-thinked.
Here a example query generated by
project_release_query_releases_by_branch():Not sure about what index would help here.
Comment #4
damien tournoud commentedNo index can help for the order by in this case. It could help to have an additional condition on
n.type = 'project_release'. After that, supposing we have an index on r.pid already I don't think we can help much more.That said, we could make this query lightning fast by denormalizing the status column in the project_release_nodes table.
Comment #5
hass commented#1: 0001-Add-more-indexes-to-project_release_nodes.patch queued for re-testing.
Comment #7
drummThis table no longer exists in D7.