Create mapping on Jira user id -> GitHub account

[mocobeta]

To correctly map Jira user ids in issues (reporter/assignee/author) to GitHub account, we need an account mapping file.
This could be inferred from https://github.com/orgs/apache/people?

[mocobeta]

Maybe we could take a brute force approach:

1. List all GitHub users or Search Users by their full names
   • Save the accounts with their names (this is available via Users API)
   • At the same time, it'd be good to log the ASF organization membership for the user and check if the user has push access to lucene repo for manual checks and possible future use
2. Associate Jira usernames with GitHub accounts by comparing "display" names (in other words, "full name") [1]

[1] E-mail address could be another clue, but it's not publicly exposed in Jira as far as I know, and also it's an optional field in GitHub. I would rather rely on users' Full Names than E-mails, although names can be ambiguous (if there are multiple accounts with same full names, it'd be possible to manually identify the correct account).

Jira Profile

GitHub Profile

[mocobeta]

Tasks to be done:

• regenerate a candidate mapping (on July 24th)
• manually make a "verified" mapping and commit it to main (on July 24th or 25th)
• send a mail to the dev list to let others browse/check both "candidate" and "verified" mappings (on July 25th)
• accept pull requests to add/edit the mapping
• fix the final mapping (on August 7th)

[mikemccand]

Could we maybe look for Jira issues that have GitHub PRs attached and "correlate" the ids of who opened the PR against who commented on the issue?

It would clearly not be perfect, but it could provide input for a human to sift through and carry over some verified accounts.

[mocobeta]

We already include merged pull requests' authors (if their GitHub full names are set to the same string as Jira full names).
Maybe we could also consider all opened pull requests' authors.

[mikemccand]

We already include merged pull requests' authors (if their GitHub full names are set to the same string as Jira full names).
Maybe we could also consider all opened pull requests' authors.

OK thanks, but does this only work for committers?

I was thinking if a contributor who is not a committer comments on a Jira issue and also opens a PR, linked to the issue, we could maybe correlate those two events to speculate about ID mapping. And then verify by hand after.

[mocobeta]

"Authors" are not necessarily committers; they literally pull request authors (contributors).
For example apache/lucene@2cf12b8

[mocobeta]

Properly speaking, the current "verified" account mapping includes both committers and commit authors. "commit authors" can be committers or contributors.

4. Verify the candidate GitHub accounts by checking  if (1) the GitHub account has push access to apache/lucene repository, or (2) the GitHub account has been logged as commit author in the repo's commit history at least once.

Here, (1) means committers and (2) means committers or contributors.

[mikemccand]

OK got it.

Could we expand the matching so that if the userid in jira == the userid in GitHub we strongly suggest a match? E.g. mdmarshmallow would have been matched this way.

Hmm, actually, his presented name (Marc D'mello) looks the same in GitHub and Jira. Oh, wait, no! One is Marc D'mello and the other is Marc D'Mello (m vs M). Maybe we can do a case insensitive comparison?

But I'll push his account to the verified file separately.

[mocobeta]

Could we expand the matching so that if the userid in jira == the userid in GitHub we strongly suggest a match? E.g. mdmarshmallow would have been matched this way.

It'd be easy to pick up such candidates - I think we'd need manually verify all of them if there are no clues other than id strings.

[mocobeta]

I'll try to improve candidate generation and verification steps maybe next week.

[mocobeta]

I'm closing this, but we'll accept improvements on mapping until the actual migration.