Decouple DmpFlagger from git
The DmpFlagger is tightly coupled to the concept of the integration of git commit hashes into our flagging pipelines. A concept I am not sure about anymore and a coupling that causes more troubles than it solves. For example:
- On systems without a system wide installation of git (e.g. Windows) the
DmpFlagger.__init__
always generates warnings - During the setup phase of a pipeline, we usually generate quite a few new commits and might also run stuff in 'git dirty mode'. In cases, where we only want to continue processing on values with changed flags, this constant modifications break the idea of a 'minimal-work-pipeline'.
To circumvent these issues, I already added the parameter version_info
to skip the inclusion of commit hashes in the comment strings, but now I think we should go even further. IMO the commit hash should in fact be an optional parameter to the DmpFlagger itself, if a certain implementation has a use for it, the hash should be generated outside of SaQC and simply passed in, if there is a need to control this from the CLI, it would be straight forward to add a new CLI-option for that.