[BEAM-8224] Fix bug in _fn_takes_side_inputs

[udim]

Bug was introduced in the conversion to inspect.signature, and manifests
as assigning a wrapper function that accepts a single positional
argument when the wrapped function can accept an arbitrary number.

---

Thank you for your contribution! Follow this checklist to help us incorporate your contribution quickly and easily:

• Choose reviewer(s) and mention them in a comment (R: @username).
• Format the pull request title like [BEAM-XXX] Fixes bug in ApproximateQuantiles, where you replace BEAM-XXX with the appropriate JIRA issue, if applicable. This will automatically link the pull request to the issue.
• If this contribution is large, please file an Apache Individual Contributor License Agreement.

Post-Commit Tests Status (on master branch)

Lang	SDK	Apex	Dataflow	Flink	Gearpump	Samza	Spark
Go		---	---		---	---
Java
Python		---			---	---
XLang	---	---	---		---	---	---

Pre-Commit Tests Status (on master branch)

---	Java	Python	Go	Website
Non-portable
Portable	---		---	---

See .test-infra/jenkins/README for trigger phrase, status and link of all Jenkins jobs.

[udim]

R: @robertwb

[udim]

run pytthon 2 postcommit

[udim]

run python 2 postcommit

[udim]

run python 3.7 postcommit

[aaltay]

Changes LGTM.

Py2 tests are failing due to: https://issues.apache.org/jira/browse/BEAM-8229

[udim]

Run Python 3.7 PostCommit

[udim]

Run Python 3.7 PostCommit

[robertwb]

This should fix it, just one comment about tests.

[robertwb]

Let's check that it actually gets the values right, e.g. do lambda *args: args. (Similarly below.)

[markflyhigh]

#9570 is merge which should unblock Python2_PC failure.

[udim]

Run Python 3.7 PostCommit

[robertwb]

Run Portable_Python PreCommit

[aaltay]

New tests are failing in the precommit tests:

17:59:06 ======================================================================
17:59:06 ERROR: test_var_positional_only_side_input_hint (apache_beam.typehints.typed_pipeline_test.SideInputTest)
17:59:06 ----------------------------------------------------------------------
17:59:06 Traceback (most recent call last):
...
17:59:06 TypeError: Expected bytes, got tuple [while running 'Map()']

[robinyqiu]

Run Python PreCommit

[udim]

trivial_inference.py defaults to Any in 3.6, 3.7 since it doesn't support kwargs (added by the wrapper).
In 2.7, 3.5 these are supported, but since side-input hints are not passed along the inferred output type is str, and the coder for that fails to encode a tuple.

Even if we did pass side-input hints, there is still missing logic that would bind the arguments for the var_positional into a single tuple (like getcallargs or signature.bind).

Perhaps we should skip trivial_inference when side-inputs are present.

[aaltay]

trivial_inference.py defaults to Any in 3.6, 3.7 since it doesn't support kwargs (added by the wrapper).
In 2.7, 3.5 these are supported, but since side-input hints are not passed along the inferred output type is str, and the coder for that fails to encode a tuple.

Even if we did pass side-input hints, there is still missing logic that would bind the arguments for the var_positional into a single tuple (like getcallargs or signature.bind).

Perhaps we should skip trivial_inference when side-inputs are present.

Would we need to skip it for all python version or for some python versions?

I am in favor of skipping in order to not block the release and continue with the proper fix in master.

/cc @tvalentyn in case anything needs to be documented as known issues for python 3.

[markflyhigh]

Run Portable_Python PreCommit

[udim]

Run Portable_Python PreCommit

[udim]

I've opened a couple of bug for the portable precommit flakes:
https://issues.apache.org/jira/browse/BEAM-8248
https://issues.apache.org/jira/browse/BEAM-8249

[udim]

I believe we can safely ignore the portable python precommit failure, since this PR is holding up the release.