An error occurred with Micromamba while creating a workflow containing @pypi

[EveT0n]

When I create a workflow with @pypi by executing the command python3 pypiflow.py --environment=pypi argo-workflows create, the following error occurs:

2024-07-17 19:19:28.652 Bootstrapping virtual environment(s) ...
    Micromamba ran into an error while setting up environment:
    command '/home/hq/.metaflowconfig/micromamba/bin/micromamba create --yes --quiet --dry-run --no-extra-safety-checks --repodata-ttl=86400 --retry-clean-cache --prefix=/tmp/tmptiqsmcnf/prefix --channel=defaults requests==>=2.21.0 boto3==>=1.14.0 python==3.6.8' returned error (1)
    critical libmamba Error parsing version ">=2.21.0". Version contains invalid characters in >=2.21.0.

According to the error message, there is a problem with the use of Micromamba's commands, and there is a problem with the format of requests==>=2.21.0. I think it might have something to do with the versioning of requests in the metaflow_config.py

def get_pinned_conda_libs(python_version, datastore_type):
    pins = {
        "requests": ">=2.21.0",
    }
    if datastore_type == "s3":
        pins["boto3"] = ">=1.14.0"
    elif datastore_type == "azure":
        pins["azure-identity"] = ">=1.10.0"
        pins["azure-storage-blob"] = ">=12.12.0"
        pins["azure-keyvault-secrets"] = ">=4.7.0"
    elif datastore_type == "gs":
        pins["google-cloud-storage"] = ">=2.5.0"
        pins["google-auth"] = ">=2.11.0"
        pins["google-cloud-secret-manager"] = ">=2.10.0"
    elif datastore_type == "local":
        pass
    else:
        raise MetaflowException(
            msg="conda lib pins for datastore %s are undefined" % (datastore_type,)
        )
    return pins

When I change "requests": ">=2.21.0" to "requests": "2.21.0" in the file, the error disappears. The python I use locally is 3.6.8. My question is that this seems to be an obvious problem, and I found that the code has not been modified in version 2.12.7, is there a problem with my use?

[savingoyal]

@cndota123 this can happen if you are relying on the RC version of micromamba. To get around this issue, please nuke - ~/.metaflowconfig/micromamba/, upgrade Metaflow to the latest version and try re-executing the flow.

[EveT0n]

@savingoyal Yes, I found out that the version of micromamba installed via metaflow is 2.0.0, in the new environment I manually installed version 1.5.8 of micromamba, this issue does not appear. However, I tried modifying the solve method of metaflow/plugins/pypi/micromamba.py to fix the issue as well.

class Micromamba(object):
    def solve(self, id_, packages, python, platform):  
          ... 
           for package, version in packages.items():
                version_specifiers = ['==', '>=', '<=', '>', '<', '!=']
                contains_specifier = any(specifier in version for specifier in version_specifiers)
                if contains_specifier:
                    cmd.append("%s%s" % (package, version))
                else:
                    cmd.append("%s==%s" % (package, version))