Rating: add rating to organisations and tools

[orchestr7]

We need to have the following addition on the backend:

1. We would like to have proper calculations of contest rating
2. To do this we need to have new fields in LnkContestProject.kt: bestExecution: Execution and bestScore: Int
3. In Organization and Project entities we should have totalScore (or something like this - please check that these fields are not yet added). It should be taken as sum of the list of BEST EXECUTIONS
4. When contest execution finishes - we update all these fields and rating

[kgevorkyan]

First phase:

• Pass type of execution to the backend
• Fill lnk_contest_execution
• Move score from lnk_contest_execution into the Execution entity

[petertrr]

Move score from lnk_contest_execution into the Execution entity

I think the assumption there was that we'll only need scores in contests. Are you planning to change this?

[kgevorkyan]

I think the assumption there was that we'll only need scores in contests. Are you planning to change this?

We decide with @akuleshov7 to move it anyway to Execution

[petertrr]

Ok, to clarify reasons - score will be used not only for contests, even if it's calculation may depend on the exact type of execution.

[petertrr]

It should be taken as sum of the list of BEST EXECUTIONS

Why do we need to sum all executions? For me this metric is not descriptive, because it mostly represents not the quality of the tool but rather how active they are using our platform. I think historical diagram of best score or variance of best scores over time may be a better option.

[orchestr7]

No-no, it should be a sum of ALL best executions on all contestsXprojects @petertrr

Imagine if an organization participated in 100000 contests and made 1000000000000000 executions. They will anyway have 100000 best results (same to the number of contests). It is much more valuable than some organization participated in 1 contest and won.

We will ban those guys who are cheating and are creating same project again and again.

[petertrr]

The only thing that concerns me is that score in different contests may be calculated differently. Now we support only F-measure (i.e. combined precision and recall for warn-plugin-based tests), but we may have different types in the future. Then we'll need to change rules of aggregation and recalculated already stored values.

Right now, to calculate contestRating (this is the name of the field we already have) we can do something like
contestRating = lnkContestProject.filter { it.project == this }.sumBy { bestScore }

And to calculate organization rating,
organizationRating = projects.findAllByOrganization(this).sumBy { it.contestRating}

I think at least the latter can be calculated per request, because it doesn't involve a lot of data. Project.contestRating can be cahced as a column, however.

[petertrr]

For other plugins calculation of rating requires saveourtool/save-cli#449

[petertrr]

#1162 (comment)

If we have LnkContestProject.bestExecution we don't need LnkContestProject.bestScore

[orchestr7]

@petertrr what's left?

[petertrr]

Only a follow-up from #1162 (comment), everything else is implemented

[orchestr7]

@nulls can you please finalise it? :)