Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[internal] Sort input requirements for LockfileMetadata #12615

Merged
merged 4 commits into from
Aug 23, 2021

Conversation

Eric-Arellano
Copy link
Contributor

This ensures that callers can't mess up the determinism of the lockfile's invalidation digest. It will also make #12610 more presentable for users, when we store the requirement strings in the lockfile rather than just a hash.

[ci skip-rust]
[ci skip-build-wheels]

# Rust tests and lints will be skipped. Delete if not intended.
[ci skip-rust]

# Building wheels and fs_util will be skipped. Delete if not intended.
[ci skip-build-wheels]
# Rust tests and lints will be skipped. Delete if not intended.
[ci skip-rust]

# Building wheels and fs_util will be skipped. Delete if not intended.
[ci skip-build-wheels]
# Rust tests and lints will be skipped. Delete if not intended.
[ci skip-rust]

# Building wheels and fs_util will be skipped. Delete if not intended.
[ci skip-build-wheels]
Copy link
Contributor

@chrisjrn chrisjrn left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good to me; stylistic gripes follow:

@@ -113,7 +114,7 @@ def calculate_invalidation_digest(requirements: Iterable[str]) -> str:
"""Returns an invalidation digest for the given requirements."""
m = hashlib.sha256()
inputs = {
"requirements": list(requirements),
"requirements": sorted(FrozenOrderedSet(requirements)),
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we might be able to safely leave this as a set() and save the import: sorted takes care of the ordering for us.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I should comment why I did this: I wanted to avoid set because it would break if the input was already ordered. Python uses timsort https://en.wikipedia.org/wiki/Timsort, which is O(n) in the best case if it's already sorted, and O(n logn) in the worst case.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

two points:

  1. set is guaranteed to be ordered as of Python 3.6, so running sorted on a set is equivalent to running it on a list.
  2. n is quite small for requirements, so if (1) weren't true, it probably wouldn't matter

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

dict is ordered, set is not. (We really on dict's ordering to implement FrozenOrderedSet)

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🤦 fine :)

],
)
def test_invalidation_digest(requirements, expected) -> None:
assert calculate_invalidation_digest(FrozenOrderedSet(requirements)) == expected
a = "flake8-pantsbuild>=2.0,<3"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Stylistically, you can re-write this test in a parametrized way by using the params:

def test_invalidation_digest(reqs_1, reqs_2, are_equivalent):
    assert (calculate_invalidation_digest(reqs_1) == calculate_invalidation_digest(reqs_2)) == are_equivalent

That will give you the benefit of the test runner showing all failure cases if things start going wrong, otherwise a failure early on hides the extent of things being broken.

If you're worried about the bit where you generate test cases using a for loop, note that the arguments to parametrize could be generated by another function.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ack that's possible. Imo, it's simpler to have this current style, e.g. the variables a, b, and c being locally scoped. Keep it simple here.

# Rust tests and lints will be skipped. Delete if not intended.
[ci skip-rust]

# Building wheels and fs_util will be skipped. Delete if not intended.
[ci skip-build-wheels]
@Eric-Arellano Eric-Arellano merged commit b37ef21 into pantsbuild:main Aug 23, 2021
@Eric-Arellano Eric-Arellano deleted the sort-requirements branch August 23, 2021 17:45
@illicitonion illicitonion mentioned this pull request Aug 30, 2021
illicitonion added a commit to illicitonion/pants that referenced this pull request Aug 30, 2021
* Clean up the config parsing code. ([pantsbuild#12678](pantsbuild#12678))

* [internal] `--generate-lockfiles-resolve` does not determine `PythonLockfileRequest` for unspecified tools ([pantsbuild#12692](pantsbuild#12692))

* [internal] Add missing `resources` targets for default tool lockfiles ([pantsbuild#12670](pantsbuild#12670))

* [internal] Reorganize lockfile code to no longer be in `backend/python/experimental` ([pantsbuild#12669](pantsbuild#12669))

* [internal] Test that default tool lockfiles work on macOS ([pantsbuild#12666](pantsbuild#12666))

* [internal] Improve the error message for stale/invalid lockfiles ([pantsbuild#12618](pantsbuild#12618))

* [internal] Rename `./pants lock` to `./pants generate-user-lockfile` and `./pants tool-lock` to `./pants generate-lockfiles` ([pantsbuild#12641](pantsbuild#12641))

* [internal] Only run macOS tests in CI annotated with `@pytest.mark.platform_specific_behavior` ([pantsbuild#12665](pantsbuild#12665))

* [internal] Test that Python tools work with all expected interpreter versions ([pantsbuild#12656](pantsbuild#12656))

* [internal] Factor out a helper rule for setting up `--resolve-all-constraints` ([pantsbuild#12630](pantsbuild#12630))

* [internal] Add links to changelog and Twitter in Pants PyPI project page. ([pantsbuild#12636](pantsbuild#12636))

* [internal] Update CoC with proper markdown + https links. ([pantsbuild#12637](pantsbuild#12637))

* GitHub issue templates ([pantsbuild#12617](pantsbuild#12617))

* Register subsystems and target types by backend/plugin. ([pantsbuild#12623](pantsbuild#12623))

* [internal] Sort input requirements for `LockfileMetadata` ([pantsbuild#12615](pantsbuild#12615))

* [internal] Expect lockfile metadata to be defined ([pantsbuild#12616](pantsbuild#12616))

* Simplify subsystem registration. ([pantsbuild#12620](pantsbuild#12620))

* [internal] Use constants for magic strings `<none>` and `<default>` for tool lockfiles ([pantsbuild#12613](pantsbuild#12613))

* [internal] Refactor `lockfile_metadata.py` to be more class-based ([pantsbuild#12611](pantsbuild#12611))

* Embrace that help strings are markdown. ([pantsbuild#12606](pantsbuild#12606))

* Stop fetching a lockfile request just to figure out the requirements digest ([pantsbuild#12607](pantsbuild#12607))

* [internal] Refactor MyPy first party plugins ([pantsbuild#12599](pantsbuild#12599))

* [internal] Refactor MyPy config file setup ([pantsbuild#12593](pantsbuild#12593))

* [internal] use `is_union()` rather than `union.is_instance()`. ([pantsbuild#12595](pantsbuild#12595))

* fs_util can use mTLS and ignore certificates ([pantsbuild#12586](pantsbuild#12586))

* [internal] Fix Pylint lockfile to actually be wired up ([pantsbuild#12590](pantsbuild#12590))

* [internal] Update lockfile TODOs with current state of project ([pantsbuild#12592](pantsbuild#12592))

* [internal] Check for lockfile staleness by using context's interpreter constraints, rather than global constraints  ([pantsbuild#12566](pantsbuild#12566))

* [internal] Refactor Pylint first-party plugins ([pantsbuild#12583](pantsbuild#12583))

* [internal] Fix Pytest and IPython to check for stale lockfiles ([pantsbuild#12582](pantsbuild#12582))

* [internal] Do not cache lockfile generation ([pantsbuild#12674](pantsbuild#12674))
illicitonion added a commit that referenced this pull request Aug 30, 2021
* Clean up the config parsing code. ([#12678](#12678))

* [internal] `--generate-lockfiles-resolve` does not determine `PythonLockfileRequest` for unspecified tools ([#12692](#12692))

* [internal] Add missing `resources` targets for default tool lockfiles ([#12670](#12670))

* [internal] Reorganize lockfile code to no longer be in `backend/python/experimental` ([#12669](#12669))

* [internal] Test that default tool lockfiles work on macOS ([#12666](#12666))

* [internal] Improve the error message for stale/invalid lockfiles ([#12618](#12618))

* [internal] Rename `./pants lock` to `./pants generate-user-lockfile` and `./pants tool-lock` to `./pants generate-lockfiles` ([#12641](#12641))

* [internal] Only run macOS tests in CI annotated with `@pytest.mark.platform_specific_behavior` ([#12665](#12665))

* [internal] Test that Python tools work with all expected interpreter versions ([#12656](#12656))

* [internal] Factor out a helper rule for setting up `--resolve-all-constraints` ([#12630](#12630))

* [internal] Add links to changelog and Twitter in Pants PyPI project page. ([#12636](#12636))

* [internal] Update CoC with proper markdown + https links. ([#12637](#12637))

* GitHub issue templates ([#12617](#12617))

* Register subsystems and target types by backend/plugin. ([#12623](#12623))

* [internal] Sort input requirements for `LockfileMetadata` ([#12615](#12615))

* [internal] Expect lockfile metadata to be defined ([#12616](#12616))

* Simplify subsystem registration. ([#12620](#12620))

* [internal] Use constants for magic strings `<none>` and `<default>` for tool lockfiles ([#12613](#12613))

* [internal] Refactor `lockfile_metadata.py` to be more class-based ([#12611](#12611))

* Embrace that help strings are markdown. ([#12606](#12606))

* Stop fetching a lockfile request just to figure out the requirements digest ([#12607](#12607))

* [internal] Refactor MyPy first party plugins ([#12599](#12599))

* [internal] Refactor MyPy config file setup ([#12593](#12593))

* [internal] use `is_union()` rather than `union.is_instance()`. ([#12595](#12595))

* fs_util can use mTLS and ignore certificates ([#12586](#12586))

* [internal] Fix Pylint lockfile to actually be wired up ([#12590](#12590))

* [internal] Update lockfile TODOs with current state of project ([#12592](#12592))

* [internal] Check for lockfile staleness by using context's interpreter constraints, rather than global constraints  ([#12566](#12566))

* [internal] Refactor Pylint first-party plugins ([#12583](#12583))

* [internal] Fix Pytest and IPython to check for stale lockfiles ([#12582](#12582))

* [internal] Do not cache lockfile generation ([#12674](#12674))
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants