Implement getter of SHA for branches

[mfocko]

Required for passing commit hashes from service to testing farm for
merging of tests that are run on their side or usage in tmt preparation.

Signed-off-by: Matej Focko [email protected]

TODO:

• tests
• consider shared method get_sha_from(branch=None, tag=None) -> Optional[str]

---

We have introduced a new function into ogr that allows you to get commit SHA of the HEAD of the branch.

[softwarefactory-project-zuul]

Build succeeded.

• pre-commit : SUCCESS in 6m 34s
• ogr-tests-rpm : SUCCESS in 6m 45s
• ogr-tests-pip-deps : SUCCESS in 7m 02s
• ogr-reverse-dep-packit-tests : SUCCESS in 16m 19s

[mfocko]

@csomh @TomasTomecek @lachmanfrantisek I would also like some opinion on the TODO and that is changing interface to: get_sha_from(branch=None, tag=None) which seems (at least to me) like a „Python-like“ API, but the implementation looks much worse:

# BaseProject
def get_sha_from(branch: str = None, tag: str = None) -> Optional[str]:
  # check if provided «exactly» one of ‹branch› or ‹tag›
  #   throw exception if not
  return None

# projects
def get_sha_from(branch: str = None, tag: str = None) -> Optional[str]:
  super().get_sha_from(branch=branch, tag=tag)

  if branch:
    # code for SHA of branch
    pass

  # code for SHA of tag
  return None

On the bright side, I like the idea of only one function, doing only one thing (that is getting the SHA, regardless whether of tag or branch). On the other side, I don't like the way it has to be implemented, SRP of it and I also feel like factoring out getting SHA of branch and tag, which basically means we could end up with 3 functions:

get_sha_from(branch=None, tag=None) # this one calls internally the other 2
get_sha_from_tag(tag)
get_sha_from_branch(branch)

↑ this also changes the TODO from «changing» to «extending» interface

And my judgement might be also affected by „indentation hell“, because both Git(Hub|Lab) throw an exception if the branch doesn't exist, so it looks meh even without being nested in another block…

[softwarefactory-project-zuul]

Build succeeded.

• pre-commit : SUCCESS in 4m 59s
• ogr-tests-rpm : SUCCESS in 8m 51s
• ogr-tests-pip-deps : SUCCESS in 8m 59s
• ogr-reverse-dep-packit-tests : SUCCESS in 16m 46s

[csomh]

To me this seems like a syntactic nuance from a user point of view:

get_sha_from(branch="branch")
get_sha_from_branch("branch")
get_sha_from(tag="tag")
get_sha_from_tag("tag")

And I don't have the "more Pythonic"-feeling towards any of the above :-)

As you said get_sha_from() would complicate implementation (and now that I think about it, could be somewhat confusing for users: it's not obvious just by looking at the method signature if providing no or both arguments is a valid option or not), so I would argue against going that way.

[TomasTomecek]

I'd say having two separate methods for each is a better design :) get_sha_from sounds like it would be too complicated and prone to errors - also with modern editors you are immediately being offered a list methods so it's easier to pick one which has narrow scope

[mfocko]

removed TODO from the code

[softwarefactory-project-zuul]

Build succeeded.

• pre-commit : SUCCESS in 3m 55s
• ogr-tests-rpm : SUCCESS in 8m 05s
• ogr-tests-pip-deps : SUCCESS in 6m 35s
• ogr-reverse-dep-packit-tests : SUCCESS in 13m 39s

[softwarefactory-project-zuul]

Build succeeded (gate pipeline).

• pre-commit : SUCCESS in 3m 55s