multiple `cml pr` runs

[DavidGOrtega]

If the data is changed after a PR it's going to run again generating strange branch names.

experiment-cml-pr-da72cf16-cml-pr-da72cf16-cml-pr-da72cf16

Things to consider:

• If we revert the state use send-comment to comment into the new PR via cml send-comment --pr --commit-sha HEAD wont work
• Do not take into account current branch name to avoid concatenation
• Take into account file changes to generate the branch name also we have to consider #792

I wonder if is practical to fix it, I mean, should cml pr allow double execution? Which ones would be those plausible scenarios?

[0x2b3bfa0]

Related to / duplicate of https://github.com/iterative/cml/issues/758?

[DavidGOrtega]

Related (not duplicated) to

• #758
• #778
• #792

[0x2b3bfa0]

Enough material for an epic? 🤔

[DavidGOrtega]

@0x2b3bfa0 Im attacking them in a PR

[casperdcl]

@DavidGOrtega since there are lots of approaches I'd suggest writing CLI docstrings for your ideas so we can discuss them before spending time implementing a polished PR.

[DavidGOrtega]

@casperdcl not really that complex, assuming that we are not rolling back the only issue here is more:

• a naming issue
• the only rollback to have into account is the user.name user.email
• add the parameters to even allow the pr name as @daavoo suggested.

I would not consider it even an epic in terms of big.