Page MenuHomePhabricator

Redundant results in ORES review tool
Closed, ResolvedPublic

Event Timeline

Restricted Application added subscribers: Jay8g, Luke081515, TerraCodes, Aklapper. · View Herald Transcript
Ladsgroup renamed this task from Redandat results in ORES review tool to Redundant results in ORES review tool.Aug 29 2016, 8:59 PM

enwiki, nlwiki, and fawiki are clean now. Working on wikidata and other wikis

All wikis are clean now. Then I need to figure it out why it shows duplicate rows.

Ladsgroup lowered the priority of this task from Unbreak Now! to High.Aug 29 2016, 9:09 PM

I was able to reproduce it, I'm getting a sense why it's happening but I'm not sure how I can fix it.

Change 307870 had a related patch set uploaded (by Ladsgroup):
Not including results when oresm_is_current = 0

https://gerrit.wikimedia.org/r/307870

Can you explain what was wrong and how you fixed it?

Okay,
Imagine we had this in ores_model

oresm_idoresm_nameoresm_versionoresm_is_current
1damaging0.0.10
2damaging0.0.21

and we have two scores for each model for every revision:

oresc_idoresc_revoresc_modeloresc_score...
123451510.06...
123461520.8...

In this case we would have two rows in user contribs or changes list query because we do LEFT JOIN on ores_model (using other types will lead to omitting cases when we don't have score which not desirable).

So I solved it by explicitly adding a where condition that oresm_is_current != 0.

It will all cases except when the oresc_model id doesn't exist in ores_model (which can happen due to T144195: Check model version replaces every time it runs.) but hopefully we won't see it again since the patch got merged.

Change 307870 merged by jenkins-bot:
Not including results when oresm_is_current = 0

https://gerrit.wikimedia.org/r/307870