Suite srv files mgr

[oliver-sanders]

This is a small internal change with no associated Issue.

This came out of the import time investigations we were looking into the other day.

Something which has long bugged me it that we have to initialise the SuiteSrvFilesManager class every time we want to use it despite the fact that most of it's methods could be functions.

This PR:

• Makes the suite_srv_files_mgr module purely functional.
• Removes some unused code.
• Tidies up and documents some constants.

Note I didn't use Python enum.Enum for the constants but just dumped them in classes (for namespaceing). The reason for this is that enums are a bit of a faff (you end up adding lots of MyEnum.Foo.value lines).

Requirements check-list

• I have read CONTRIBUTING.md and added my name as a Code Contributor.
• Contains logically grouped changes (else tidy your branch by rebase).
• Does not contain off-topic changes (use other PRs for other changes).
• Already covered by existing tests.
• No change log entry required (internal only, invisible to users).
• No documentation update required.

[oliver-sanders]

This will fail unit-tests.

Mock-based unit-tests are inherently fragile to implementation changes (which is annoying as that is when you need them the most). Two are still failing, I can't work out how to fix them without re-implementing them.

• TestSuiteSrvFilesManager.test_register
• TestScan.test_get_scan_items_from_fs_with_owner_active_only

I can't seem be able to patch imported functions the way it was done before e.g:

mocked_os.path.basename = os.path.basename

@kinow any advice on how to fix these tests?

[kinow]

Created a PR in case that's easier to copy/adapt the code from: oliver-sanders#10

[oliver-sanders]

not able to mock os by itself normally. So the only way is to be specific about what needs to be mocked.

Aah, OK, I've no idea why this change would break that.

Created a PR in case that's easier to copy/adapt the code from: oliver-sanders#10

Thanks, would have taken me forever to work that out.

[oliver-sanders]

Looks like there might be a small speed-up, I've measured an improvement of ~10% in my slow, bloated Conda environment.

$ git checkout master
$ cylc run generic --hold
$ prof --n=10 --shell=python 'from cylc.flow.network.client import SuiteRuntimeClient; SuiteRuntimeClient("generic")'
1.91900000000000000000
$ git checkout suite-srv-files-mgr
$ prof --n=10 --shell=python 'from cylc.flow.network.client import SuiteRuntimeClient; SuiteRuntimeClient("generic")'
1.69400000000000000000

[oliver-sanders]

Deconflicted.

[oliver-sanders]

Merged in latest master with TB tweaks.

[matthewrmshin]

Quick comment.

[hjoliver]

Approving as @oliver-sanders has addressed some comments and argued his way out of the others 👍

[kinow]

Added some comments, but I think these are no blockers @oliver-sanders . Feel free to press the merge button if OK (I guess we can more renaming class name and adding docstrings to another follow-up PR or issue?)

[kinow]

And code looking much cleaner 👏 👏 👏

[oliver-sanders]

Addressed comments, waiting for tests to pass...

[oliver-sanders]

Deconflicted.

[oliver-sanders]

(coverage is actually up slightly but the patch coverage is low because suite_srv_files_mgr coverage was low before)

[kinow]

One unused import, but not a blocker. Re-approved 👍