feat(operations.files): Ensure file ops don't fail if source missing in prepare phase

[aktiur]

Updated the move, copy and unarchive operations so that they won't raise an operation error when the source file (or target directory) is missing while in the prepare phase. Instead, they simply log an info message.

During the execution phase, the operations raise errors as before.

This commit is based on previous work by @vaz-ar in PR #1787.

All tests pass. However, the FakeState used by pyinfra-testing always has is_executing set to false. This means only the prepare phase code path is actually tested. On a similar note, I haven't seen a way to assert what is logged during the operation. How should I proceed to make the tests more complete?

[wowi42]

LGTM

The core logic is solid. During the prepare phase, move, copy, and unarchive now log an info message and proceed when the source or destination directory doesn't exist yet, rather than raising OperationError. The execution phase still raises errors as before. The False vs None distinction in the fact return values is handled correctly: False (path exists but is wrong type) always raises, None (doesn't exist yet) is lenient during prepare.

Two small things:

First, the host.log → logger.info change in the line function and the import reordering of format_text are unrelated to this PR. Functionally identical, just a style change. Please split these out into a separate PR.

Second, move/invalid_dest.json was changed from testing null (dest doesn't exist) to false (dest is not a directory). Both are useful, but the "doesn't exist" prepare-phase path for move's dest is now untested. A follow-up fixture would close that gap — the copy equivalent is already covered by copy/invalid_dest.json.

On execution-phase testing: you're right that FakeState always has is_executing=False. The is_executing branches are simple enough to verify by inspection, so this isn't a blocker. If you want to add coverage later, a small patch to pyinfra-testing to allow setting is_executing on the fake state would be the cleanest approach.

[aktiur]

Thanks for your review! I updated the PR to address your two remarks:

First, the host.log → logger.info change in the line function and the import reordering of format_text are unrelated to this PR. Functionally identical, just a style change. Please split these out into a separate PR.

My mistake. I originally used host.log in the prepare phase code path of the operations I modified, but realised I could not, because log is not defined on the FakeHost used during tests. I was not careful and/or too zealous when I did some find/replace to handle that. It's gone from the PR.

Second, move/invalid_dest.json was changed from testing null (dest doesn't exist) to false (dest is not a directory). Both are useful, but the "doesn't exist" prepare-phase path for move's dest is now untested. A follow-up fixture would close that gap — the copy equivalent is already covered by copy/invalid_dest.json.

I created two additional tests for both files.copy and files.move named missing_dest that cover the case where the destination directory is missing. The original tests (named invalid_test) both cover the case where the destination is not a directory.

On execution-phase testing: you're right that FakeState always has is_executing=False. The is_executing branches are simple enough to verify by inspection, so this isn't a blocker. If you want to add coverage later, a small patch to pyinfra-testing to allow setting is_executing on the fake state would be the cleanest approach.

I'm going to look into that.