Support disk-level experiment purging in access-profiling

[minghangli-uni]

closes #49

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (09d20f5) to head (624576c).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main       #50   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           16        16           
  Lines          834       854   +20     
=========================================
+ Hits           834       854   +20     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[micaeljtoliveira]

@minghangli-uni Thanks for creating this PR! I know this is still marked as draft, but I took the liberty of having a look and adding some suggestions. I hope that's fine.

[minghangli-uni]

@micaeljtoliveira would you mind having another look?

[micaeljtoliveira]

@minghangli-uni Thanks for the changes. This looks almost ready, just a couple of suggestions.

[micaeljtoliveira]

Suggested change

	def purge_experiments(
	def delete_experiments(

Since this now also deletes the experiments from the manager, I think "delete" would be clearer than "purge". This requires some updates to the docstring and comments for consistency.

[minghangli-uni]

Yeah agreed. Re the existing delete_experiment, I think renaming it to something clearer, like clean_experiment? I'd prefer to do that in a separate PR.

Another thing is that the runner's delete_experiments currently maps onto payu sweep semantics. Without hard flag it's really a sweep/cleanup only, and with hard it removes the directory. To avoid ambiguity in this repo, I'm leaning towards making delete_experiments deterministic - delete always means a hard delete, so users dont have to remember a flag to know whether delete actually deletes.

[micaeljtoliveira]

Re the existing delete_experiment, I think renaming it to something clearer, like clean_experiment? I'd prefer to do that in a separate PR.

Yes, I agree, that one needs to be renamed (I'm thinking to rename it to something like drop_experiment so it's a bit clearer it does not touch the filesystem). I can take care of it in a different PR, no worries.

Another thing is that the runner's delete_experiments currently maps onto payu sweep semantics. Without hard flag it's really a sweep/cleanup only, and with hard it removes the directory. To avoid ambiguity in this repo, I'm leaning towards making delete_experiments deterministic - delete always means a hard delete, so users dont have to remember a flag to know whether delete actually deletes.

I agree, here it's better to always to a hard delete. Maybe later we can add a Payu-specific method to do a sweep, but I'm not sure that's really useful here.

[micaeljtoliveira]

Suggested change

	"""Purges Payu experiments from the work directory.
	"""Delete Payu experiments from the work directory and remove them from the manager.

[minghangli-uni]

Thanks @micaeljtoliveira for your suggestion. Would you mind taking another look at this?

[minghangli-uni]

I'll clean up the history before merge