[BEAM-3142] Apply futurize on gen_protos #4077
Conversation
holdenk
changed the title
|
cc @aaltay any idea why this would fail here (or how to get the log file out of jenkins)? It hasn't even it the point of changes when it fails to generate. |
|
SUCCESS --none-- |
2 similar comments
|
SUCCESS --none-- |
|
SUCCESS --none-- |
|
Hey @holdenk, sorry I missed the question earlier. Do you still have the issue? |
holdenk
changed the title
|
No, its working now I did some poking in another PR to get the logs sorted out. |
sdks/python/gen_protos.py
Outdated
| @@ -111,6 +111,15 @@ def generate_proto_files(force=False): | |||
| '%s' % ret_code) | |||
|
|
|||
|
|
|||
| if sys.version_info[0] >= 3: | |||
| ret_code = subprocess.call( | |||
| ["futurize", "-0", "-wv", "--no-diff", out_dir]) | |||
There was a problem hiding this comment.
Could you expand the arguments to their long form: futurize --both-stages --write --verbose --no-diff ?
holdenk
force-pushed
the
BEAM-3142-fix-proto-gen-for-py3
branch
from
d967dc6
to
0a1737a
Compare
|
ping? would be nice to get this in. |
|
LGTM. Should we use futurize for Python 2 as well and remove the branch? |
|
Is this sufficient? Or should we be generating (and somehow selectively choosing) proto files generated targeting Py3? |
|
Futurizing for Python2 would require futurize to be installed by Python2 users so I'd say no. |
|
This PR has been open for a while, I will merge it. If the changes are not sufficient we can revisit. |
Python 3 removes relative import support, we can apply futurize on the generated protos to fix the imports. It's less than ideal, but it works.