Open anton-khodak opened 8 years ago
@anton-khodak Did you look at https://travis-ci.org/common-workflow-language/workflows/builds/134234968 ?
I think it is fine to just check in the generated descriptions; don't worry about writing a specific test. As long as the generated output parses, that is good enough for now.
I'm totally with @mr-c, we should focus first on CWL, not on specific tools since the amount of work can be quite substantial. If you want to see whether one of the CNVkit subtools works it's fine to dedicate some focused effort, but by no means aiming to cover the whole suite of tools.
Hope that makes sense ;)
OTOH, for a good example on how to test different tools (in my case SV callers), MetaSV has it quite well wrapped up:
https://github.com/bioinform/metasv
But this is just an example, don't spend too much time looking through it.
@brainstorm , that's great! I misinterpreted the goal of the PR, it was not to pass Travis checks but to merely validate those tools. In that case, I'll fix the job file (@mr-c pointed indirectly on that issue) and push all other tools.
UPD. I should have looked more closely at test/cwltest.py
... Travis CI checks the mere validity of tools, not how they are executed (with or without errors).
Hi guys, I'm happy to help with testing CNVkit and/or tweaking the test suite to play better with argparse2cwl. You can skip wrapping anything marked "deprecated" (e.g. loh
, genome2access
), those parts will be removed in the next release. Just let me know anything else you need.
@etal, very happy to have you help Anton with that. I was looking at the outputs
generated by argparse2cwl
yesterday but since I never used CNVkit before, I'm missing a few bits of domain expertise there, so help is super welcome, thanks!
I've released a new minor version of CNVkit that drops the deprecated parts and introduces a few new options. I think the current CWL wrappers in Anton's repo should still work, but batch
has a new --method
option that's worth exposing. Let me know if there's anything else I can do to help complete and maintain these wrappers.
Standing PR to add tool descriptions (created by argparse2cwl) and tests for CNVkit tools .
Issues I encountered on first steps:
I think this error might be caused by irrelevant data.
Also, I couldn't find copy number reference profile sample files (.cnn) at all. If somebody who uses CNVkit frequently could give me a hint where to take proper data, my work in testing would have been much facilitated.
tool-name-test.yaml
file format is specified. It was intuitively understandable what to write there, but I wish somebody pointed the standard for those files.