Closed bernt-matthias closed 8 months ago
ping @ebolyen
@bernt-matthias, is this urgent on your end? @ebolyen is out till this Thursday, but I may be able to get ahold of him. Let me know.
is this urgent on your end?
not so urgent :) no worries.
Let me know if I can help :)
Thanks @bernt-matthias!
Ok. I am satisfied that this works and am now merging it as a stop gap until the next q2galaxy release makes more permanent changes. Thank you @bernt-matthias and @mvdbeek
I think this was not deployed. The test toolshed is currently not working. Probably you need to https://github.com/galaxyproject/tools-iuc/blob/0da1ca313d02ca780a671247e434a424fa67887b/.github/workflows/pr.yaml#L392
For this PR its probably the easiest to push manually to the toolshed.
@bernt-matthias I made that change, and it got us past the previous block, but there is now an issue with templating out one of the tools that I can't properly track down. I know where it's coming from, but I don't know what's causing it to happen (or honestly it's more like I don't know why it hasn't always been a problem). I'll most likely need to consult with @ebolyen about it when he comes in on Thursday since he knows the most about this code-base. I apologize for the delay.
If you point me to the error I might have a look.
I guess it would be sufficient to create another PR (introducing a mock change in the tool) and merge it. Then it should be deployed.
@bernt-matthias The issue we're seeing now doesn't seem to have anything to do with Galaxy itself it appears to be purely a q2galaxy thing to do with how the validator for the input for the quality-filter q-score
action is being rendered. Previously the validator was checking if the type of the artifact provided was in the list:
['SampleData[JoinedSequencesWithQuality]', 'SampleData[SequencesWithQuality]', 'SampleData[PairedEndSequencesWithQuality]']
Now it is checking if the provided type is in the list:
['SampleData[JoinedSequencesWithQuality]', 'SampleData[SequencesWithQuality | PairedEndSequencesWithQuality]']
Causing an artifact of the type 'SampleData[SequencesWithQuality]'
to fail the validator. This is the test that is failing in CI and when I render the tools locally I also get the later of the two validators. The strange thing is that even when I go back to older versions of QIIME 2 and q2galaxy I still get the bottom validator.
The bottom one is more accurate to how the type expression is written inside of q2-quality-filter
, but the tool for q-score
that currently lives in this repo has the first validator, and I can't for the life of me figure out how we ended up getting that first validator into the tool when the second one is all I can get rendered on any version of things, and the relevant code has not changed for some time, and the second one is a more accurate representation of the type used in the actual underlying QIIME 2 action.
RE: mock changing the tool, that's just the thing the existing tool in this repo works (though I have no idea how it was templated in the way it is) it's the one that's being re-rendered in this CI action that fails, so we are unable to re-render the tools. CI has gotten past the action it was failing on yesterday. I'm not sure if you need all the tools to re-render to be able to do what you need to do to deploy the import you manually modified. I suspect not now that I think about it.
I'm not sure if you need all the tools to re-render to be able to do what you need to do to deploy the import you manually modified.
I think it would be sufficient to push the latest version of the import tool to the Galaxy toolshed. Maybe you can do this manually?
The following Traceback can be found in the logs (the test itself is not executed)
I also tried to understand what the config file is doing. If I get it right it generates a json (where some elements are stripped).
Did you know that a json file containing the inputs can also be generated with:
see https://docs.galaxyproject.org/en/master/dev/schema.html#tool-configfiles-inputs. Guess stripping elements could be done with a little helper script that could run in the tool (before
q2galaxy
).Seems that this fails with 23.1 but worked in 23.0. I will try to bisect this, but it's unlikely that this is a Galaxy bug ... the configfile just seems to access some Galaxy internals that may be subject to change.
I also would be a huge fan of having a few tests for the fastq import here.
In order to get the test running I needed to add
collection_type="list"
to the collection parameters. Might also be a good thing to add.