frontend: add failed-to-succeeded-stats command

[FrostyX]

Fix #2779

[FrostyX]

how much time pass between failed job and first succeeded of the same NEVRA.

I think the same NEVRA probably isn't what you are interested in? Because I can bump the package version/release while fixing it.

So the script goes through every package and its builds. If any of them fails, it will find the next successful build of the same package (It doesn't care for NEVRA, it only cares for package.id) and compares their build.ended_on.

@xsuchy, I didn't exactly remember what you are going to use this for, but I know we talked about average/median time. I think this output might be more helpful. But I can change it any way you need.

I run this against the current production database dump, and the result looks like this:

Which can be read as "In 367 cases, a new successful build was submitted before the failing one could even finish. 59 failures were fixed in seconds, 9210 builds were fixed within hours after they failed, etc. The time ranges are very much understandable from the script itself. I can add more of them if you need.

Lastly, I am ignoring all packages that have only one build, packages that never failed, packages that have never been fixed, etc.

[xsuchy]

For the record here are few artifacts that Jakub created:

Statistics for Packit user only:

Statistics for all users for 2021

Statistics for all users for 2022

[lbarcziova]

with this, are the builds filtered via the Copr project owner? (or via the submitter of the build?)

[nikromen]

afaik the user_id in Copr model is only project owner since this model represents the project as a whole, not the individual builds. To filter builds according to submitter:

query = query.filter(models.Build.user_id=PACKIT_USER_ID)

[FrostyX]

Exactly as @nikromen is saying.