I’ve developed a simple testing framework for Snakemake workflows. Each test case has some number of input/output files. Tests are run like so:

1. An empty working directory is created.
2. The input files are copied to the working directory (or symlinked to from it).
3. Snakemake is invoked in the working directory using the CLI to generate the output files.
4. The framework checks that the expected output files exist and their contents match example versions.

This mostly works, but I run into some situations where Snakemake decides that the input files are out of date and need to be recreated from scratch. I would like to prevent this from happening, so that Snakemake only runs the small set of rules that are meant to be tested. I have looked into the following solutions:

• The --rerun-triggers option seems like it may do what I need but the documentation isn’t very clear about what the possible values mean. There is not a “none” or “never” value, and it will not accept an empty value.
• Providing a whitelist of the rules that can be run. If Snakemake can’t produce the requested files with these rules then it should fail. Unfortunately there doesn’t seem to be a way to do this. The --omit-from option does sort of the opposite, acting as a blacklist instead of a whitelist.
• Running --touch on all input files first. This might work, but I’m not completely sure how this is supposed to work from the documentation (if this is tracked within the .snakemake/ directory, or if it actually works like the touch command and just updates the mtime of the files themselves [and if so if that works when the files are symlinks]).