add new validation type not_contains

[asraa]

This commit adds a not_contains validation type for string fields. The goal is to validate that field doesn't contain a blacklisted character, like nul characters.

Signed-off-by: Asra Ali [email protected]

[asraa]

Hey, this is my initial commit for a not_contains type. Had a quick question:
I used protoc-gen-go to generate validate.pb.go, but the formatting looks very different than the old version. Was there a better way to do this?
I also updated vettools, but I'm having doubts it was necessary since the "-shadow" removal was a while back (?) see https://go-review.googlesource.com/c/go/+/154584/3/doc/go1.12.html
/review @akonradi

[akonradi]

Not sure what's up with validate.proto.go but the rest LGTM modulo the one doc comment. I'd suggest getting an additional reviewer since this is an API change, though.

[asraa]

/review @rodaine

[rodaine]

you're identifying the first two tests as "valid" but then setting their validity to false, and vice-versa for the last two. Either the labels are wrong, or the implementation is backwards.

[asraa]

Oops, thank you so much for catching that. I didn't catch the mislabeling.

[rodaine]

So as stated elsewhere, please set the file permissions back to 0644; this file has no business being executable 😅

Also, let's generate this file with a version of protoc-gen-go before the addition of this new constant. For folks not using this package via Bazel, there is no significant changes that warrant the runtime upgrade (and this will cause compilation errors for them).

[asraa]

Was it enough to manually change the proto version of the file? (I saw some instructions about installing protoc-gen-go with a custom tag, but this seemed like enough from a previous PR).

[rodaine]

No, unfortunately. protoc-gen-go emits those constants to control the minimum runtime version of the protobuf package used. The emitted code emitted with "3" has optimizations that don't exist from when the value used to be "2". That is to say, it will still be broken.

[asraa]

Gotcha, I installed v 1.2.0 of protoc-gen-go and used that to generate the pb.go file.

[asraa]

Just added python support as well.

I had to change sys.stdin.read() to sys.stdin.buffer.read() in harness.py. @kothariaditya, did you encounter errors like:

Traceback (most recent call last):
  File "/usr/local/google/home/asraa/.cache/bazel/_bazel_asraa/846d06093714e38d686b02e69f690c56/execroot/com_envoyproxy_protoc_gen_validate/bazel-out/k8-fastbuild/bin/tests/harness/executor/linux_amd64_stripped/executor.runfiles/com_envoyproxy_protoc_gen_validate/tests/harness/python/harness.py", line 37, in <module>
    message = sys.stdin.read()
  File "/usr/lib/python3.6/codecs.py", line 321, in decode
    (result, consumed) = self._buffer_decode(data, self.errors, final)
UnicodeDecodeError: 'utf-8' codec can't decode bytes in position 61-62: unexpected end of data

when you ran the python tests?

[kothariaditya]

Hey @asraa I don't think I encountered that before. Did you get the error before adding the not_contains testcases as well?

[asraa]

Yep, I hit that for any Python testcase including floats, etc, as a matter of just reading in the file.

[kothariaditya]

Wow that's really weird. Never had that issue while running locally or in the CI. Thanks for updating it to sys.stdin.buffer.read(). Sorry for the trouble!

[asraa]

@rodaine Hey Chris, when you have a chance, would you mind taking a look at this PR again? Thank you!

[rodaine]

Thanks for the patch @asraa!