jaketf commented on a change in pull request #5602: [AIRFLOW-4983] Feature/dataflow operator python3 URL: https://github.com/apache/airflow/pull/5602#discussion_r304972812
########## File path: airflow/contrib/hooks/gcp_dataflow_hook.py ########## @@ -34,8 +34,12 @@ # This is the default location # https://cloud.google.com/dataflow/pipelines/specifying-exec-params + DEFAULT_DATAFLOW_LOCATION = 'us-central1' +# These are the supported python interpreters +DATAFLOW_SUPPORTED_PY_INTERPRETERS = {'python2', 'python3'} Review comment: @tvalentyn Good call out on minor versions. Currently [only python 3.5 is considered stable for beam.](https://beam.apache.org/roadmap/python-sdk/#python-3-support) However, adding minor version to interpreter might be good. This gets a little tricky because we don't know what is available in whatever system is running airflow (could be on a standalone VM, on Kubernetes, in a Hadoop cluster, etc. etc.). In [cloud composer it will be 2.7.15 or 3.6.6](https://cloud.google.com/composer/docs/concepts/python-version). By not hard coding minor version it softens assumptions about the infra that composer is running on. Not sure what the best path forward is here. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
