Page MenuHomePhabricator

Offer "Avg. Time Open (Days)" average value for Gerrit patchsets per affiliation/organization
Open, NormalPublic0 Story Points

Description

"Gerrit Timing" offers Avg. Time Open (Days) per submitter under Submitters:

Would love to be able to filter on Organizations in general, e.g. if patchsets submitted by WMF folks get reviewed quicker than patchsets submitted by independent folks.

Internal non-public ticket: https://gitlab.com/Bitergia/c/Wikimedia/support/issues/63

Event Timeline

Aklapper triaged this task as Lowest priority.May 31 2019, 8:03 PM
Aklapper created this task.
Aklapper raised the priority of this task from Lowest to Normal.Aug 5 2019, 9:23 PM
Aklapper updated the task description. (Show Details)

For clarification, quoting from the internal ticket:

One changeset can have several patchsets (the last one overwrites the previous ones, afaik). The time we track is for changeset, based on its status. That means we track the time from the moment the changeset is created to the point in which it is resolved or abandoned.

Adding some more background, the more I think about this:
I do realize that a changeset marked as CR-1 (means: needs improvement) might simply remain open forever if nobody ever picks it up and fixes it. And that could make the stats more noisy and less useful.
I'm more interested in the time values per affiliation for changesets that are CR+0 (no review has happened yet) and CR+1 (someone has given positive feedback but someone else must approve the changeset via a CR+2 to get merged, which can sometimes be a bottleneck because the person giving a CR+2 is kind of responsible if the changeset gets merged and then breaks something).
So I'm back to wanting also to be able to filter this on code-review labels, as requested in T224755 and I hope that this is going to be combinable.

As this is a bit related: Thanks to the Status widget on the Gerrit-Timing dashboard, we already know that the review rate (percentage) of Gerrit changesets is different per affiliation, by filtering on the Organization. However, these stats only reflect the status of patches created in the last 90 days as of right now; they include no information about the timing. Comparison for the last 90 days of Wikimedia Foundation vs Individual:

We're running into inconsistencies in Gerrit data in this task which distort the results:

Full list of the 43 problematic changesets can be seen at https://wikimedia.biterg.io/goto/87013a17063b675738cd841d8fb49c6b

Wondering if we could simply use the created_on value instead of the opened value to calculate the painless_delay value (and as there is a patchset_created_on), but need to check with Bitergia who know what they are doing, unlike me.

We also need to provide input on how to visualize this. Input very welcome.
Currently the visualization proposal can be seen in https://wikimedia.biterg.io/goto/d813c19b886e7015c587014af443b2b0 , adding a screenshot here:


Note that this is still a draft and work in progress.

Also wondering about medium vs average here.