We're now gating on pep8, so make it pass. This removes a couple of
unused imports and fixes a typo in the CSV output code in the reviewers
The list of reviewers in the code actually includes people that didn't
do any reviews, but received them. Since tracking received reviews,
this value became inaccurate.
Remove the number of reviews received from the reviewers report output.
I think this information is valuable, but the values we're getting are
not accurate. The problem is that if a patch gets rebased, the trivial
rebase detection will automatically re-apply reviews. Those are getting
counted even though they aren't new reviews being received.
This involves updating one of the shell scripts which now depends on
'pip install -e .' or 'pip install reviewstats' (for a release) having
One step on a long path. Note that there is some noddy stuff in here
like a pointless test - its in cookiecutter and I don't want to muddle
code moves up with that, so 'meh'.
Previously, a disagreement was defined as a core reviewer giving a
negative vote after the person had given a positive vote. The
percentage was then the number of disagreements / all positive votes.
Later, disagreements were extended to cover the opposite case: when a
core member gives a positive vote after a person had given a negative
The percentage calculation was not updated to reflect this new
definition. It still showed disagreements as a percentage of the
positive votes, instead of all of them (+ and -).
There is a wholly separate dev community around jenkins-job-builder
that it would be nice to track. Leaving jjb also in infra though,
because it's interesting to track there as well.
Adds more per day stats:
- Changes with at least one patchset active in the time period
- New changes uploaded
- Changes merged
- Changes abandoned
- Changes left in WIP
Estimation of queue growth based over the time period
Add a new column that shows how many reviews each person received in the
same time frame. It also shows the ratio between reviews given and
As a side effect, the result now also includes those that submitted
patches but did no reviews. I think that's still useful information to
show when looking at the heath of the review queue.