[id2vec] Collect Role-Identifier dataset on full PGA

[zurk]

Now I have a small piece of this dataset for tests as reported here: https://github.com/src-d/backlog/issues/1248#issuecomment-398319941
@vmarkovtsev wants to publish all datasets we have right now (https://github.com/src-d/backlog/issues/1310), so it is a good opportunity to collect full Role-Identifier dataset.

[vmarkovtsev]

@zurk Please update the status.

[zurk]

Since we have jgit-spark-connector (former sourced-engine) deprecated, first we should rewrite sourced-ml and after that, we can collect dataset.

[zurk]

@vmarkovtsev I do not think this dataset add any value for us now.
Let's close this issue as an outdated?

[vmarkovtsev]

Actually no, this is still important and valuable. Once we copy PGAv2 in https://github.com/src-d/infrastructure/issues/788 we should extract the dataset. It is fine to write some ad-hoc code for that.

[vmarkovtsev]

Romain, you will be able to do this one the collected UASTs from PGAv2.

[vmarkovtsev]

Please remember to use src-d/ml-mining for the code. You'd probably like to nuke 50% of the code from there, that's OK.