Is there any way to run task w/o running its dependencies?

[mbergal]

Will be very useful when writing Taskfiles where deps are very heave and long to run

[ghostsquad]

The general pattern is to encode logic in your dependent tasks that will indicate whether or not that task needs to run. This is done through sources and generates and/or status.

https://taskfile.dev/usage/#using-programmatic-checks-to-indicate-a-task-is-up-to-date

https://taskfile.dev/usage/#by-fingerprinting-locally-generated-files-and-their-sources

Related #788

[mbergal]

I am developing my Taskfile. build depends on other tasks. I run build the first time. clean and clone execute successfully. Their execution took ~5min. build task failed. I don't need to repeat clean and clone, they've been done already. I just need to repeat build.

[ghostsquad]

Is there a way to programmatically identify if clean/clone has already been done?

E.g. clean runs rm -rfd ./tmp, then you can skip clean if tmp doesn't exist.

Clone could be skipped or alternative steps performed if the directory that was cloned into already exists. Check if it exists, then run git pull instead.

[ghostsquad]

If clone does a clone or pull depending on the directory existence, then it doesn't require clean.

[mbergal]

Clone could be skipped or alternative steps performed if the directory that was cloned into already exists. Check if it exists, then run git pull instead.

This becomes very complicated and not very general. Not every step creates/removes a directory. In reality clone is not just git pull, but also dvc pull ... and still takes a long time. Of course, I can just put a marker file for every successful step, but --only is way easier, IMHO.

[ghostsquad]

Pull requests are welcome and regularly reviewed.