Closed ajschmidt8 closed 1 year ago
Yeah looks like we are just doing a basic sort for CalVer which will fail in oct/nov/dev.
If you set it to be semver instead it will more intelligently sort, but that was causing problems for some users who were using calver tags that were not semver compliant.
For example 2022.09.0
is not valid semver because you cannot have leading zeroes. (Note that this also sorts correctly with the lexicographic sort).
It looks like dask-sql
tags are valid semver because there is no leading zero, so I wonder if switching to the semver mode would be a good workaround?
- name: Get latest cuDF nightly version
id: cudf_latest
uses: jacobtomlinson/gha-anaconda-package-version@0.1.3
with:
org: "rapidsai-nightly"
package: "cudf"
- version_system: "CalVer"
+ version_system: "SemVer"
I will try that
that seemed to work. latest log w/ debugging enabled is below. closing.
We use the workflow below to keep
dask-sql
up-to-date in our Docker images.https://github.com/rapidsai/docker/blob/branch-22.10/.github/workflows/update-dask-sql.yml
However, from the latest workflow run logs here, it seems that the action is picking up
2022.9.0
as the latestdask-sql
version despite2022.10.0
being out.It seems the API is returning the correct
2022.10.0
entry, so I think this has to do with lexicographic sorting issues between2022.9.0
and2022.10.0