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] Do not cache lockfile generation #12674

Merged
merged 1 commit into from
Aug 27, 2021

Conversation

Eric-Arellano
Copy link
Contributor

Closes #12591.

[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]
@@ -126,6 +126,17 @@ async def generate_lockfile(
input_digest=pyproject_toml_digest,
output_files=("poetry.lock", "pyproject.toml"),
description=req.description,
# Instead of caching lockfile generation with LMDB, we instead use the invalidation
Copy link
Contributor

@jsirois jsirois Aug 27, 2021

Choose a reason for hiding this comment

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

It seems like the oddness would all be solved if the inputs to a lock rule were (<resolve inputs>, <lockfile snapshot from FS if it exists>):

  1. lockfile snapshot DNE: generate using resolve inputs.
  2. lockfile snapshot does exist: compare resolve inputs to snapshot header and only resolve if diff unless --force, in which case do like 1.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

This is already the case thanks to @chrisjrn 's work. Whenever we consume a lockfile, we read its metadata like this:

# --- BEGIN PANTS LOCKFILE METADATA: DO NOT EDIT OR REMOVE ---
# {
# "requirements_invalidation_digest": "76ad65c62c8e6ded78129ae57a934c85cdf562cbe8a8746cf8502d0b07a7e25b",
# "valid_for_interpreter_constraints": [
# "CPython<3.10,>=3.7"
# ]
# }
# --- END PANTS LOCKFILE METADATA ---

We then check that the resolve inputs are compatible with that lockfile. If not, we warn/error that you need to manually regenerate the lockfile. Thanks to #12676, that warning will tell you to only regenerate for that specific lockfile, without needing to regenerate unrelated ones.

--

In all those cases, I don't think we should be caching in LMDB. If you have set to regenerate a lockfile—either because Pants instructed you to or you just want to see if there are new releases—we should be pulling the current state of the world rather than what's cached in LMDB.

@Eric-Arellano Eric-Arellano merged commit d4ef1a6 into pantsbuild:main Aug 27, 2021
@Eric-Arellano Eric-Arellano deleted the lock-cache branch August 27, 2021 20:08
@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.

Lockfiles: Design caching semantics for Python lockfile generation
3 participants