Repository navigation
Improve type checking in tests - #684
Merged
Merged
Conversation
|
PR revision: No breaking API changes detected. PR public API verdictNo breaking API changes detected. Target: Static analysis only: no behavioral compatibility guarantee or external-dependency signature analysis. Public API changes: 9367681 -> HEADExecutive summaryNo breaking API changes detected. Breaking changesNone. New public APINone. |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #684 +/- ##
=======================================
Coverage 100.0% 100.0%
=======================================
Files 163 163
Lines 9979 9979
Branches 1199 1199
=======================================
Hits 9979 9979 🚀 New features to boost your workflow:
|
Collaborator
Author
|
Because this is a large PR (touching many files), here is an AI-generated summary (generated by gpt-6.1-sol at high effort level). My prompt was: The "I" in the following is gpt-6.1-sol. I reviewed the diffs of all 36 commits in PR #684. The relevant changes are individual commits within that PR, so the links below point to commits. None changes production-library code under Test Logic And Fixtures
Mock Calling Contracts
Other Non-Annotation Changes
The remaining commits contain annotations, behavior-preserving casts or rewrites, formatting, imports needed for annotations, or documentation changes. |
Annotate test returns, mock fixtures, fixture parameters, and source update/reset callbacks to improve static type checking. Remove the unused FakeSource implementation and source_pair fixture from the two-point theory tests. (requiring type annotations noted that it was returning the wrong type; only because it was unused was it not causing runtime failures).
Add precise fixture types and return annotations to improve static type checking. Move a tracer-pair membership assertion before its lookup.
Add a shared TypedDict for cosmology fixtures and annotate fixture returns, test parameters, and dynamically selected fixture values. Use the real XCDM fixture instead of the mock in map_cosmo_spec_nl. Move imports to module scope and document local Pylint suppressions for introspected NumCosmo APIs and intentional protected access.
Annotate FakeSource methods and both integration tests with explicit types (sacc.Sacc, ParamsMap, ModelingTools, Tracer, -> None). FakeSource.create_tracers now raises NotImplementedError instead of returning an empty list, so any accidental call fails loudly. The update and reset tests do not exercise create_tracers.
Replace the .conda-env-specific instructions in docs/agents/python.md
with a generic requirement to run checks in an active, isolated
environment (conda, venv, or agent sandbox) and verify that python,
mypy, and black resolve to that same environment. Also fix two typos
in the monkeypatching guidance ("monkeypatching it the" and the
missing colon after "Avoid monkeypatching").
Represent expected galaxy redshift distributions with TomographicBin rather than two bare NumPy arrays. The domain type makes the fixture contract explicit: a named tomographic bin with redshift samples, distribution values, and supported measurement types. Migrate six single-galaxy-bin fixtures to return (sacc_data, expected_bin). Construct expected bins directly from known inputs, with copied arrays and explicit measurement types. Update consumers to use expected_bin.z and expected_bin.dndz while preserving extraction calls and assertions. Add explicit type annotations to test fixtures and functions. Make TrivialSourceGalaxy.create_tracers raise NotImplementedError because the test subclass only supports source argument handling. Remove the unnecessary galaxy-fixture dependency from test_cmb_convergence_read and create a local empty sacc.Sacc() instead.
Reflow multi-line function signatures in fctools tests onto single lines to match the project's formatter. No behavior or test logic changes.
Annotate test methods, fixtures, and mock functions in tests/fctools/test_list_tools.py with parameter and return types (Path, pytest.MonkeyPatch, NoReturn, object). Adds the typing and pytest imports needed for these annotations. No test logic changes.
Annotate test functions in test_statistic.py with `-> None` return types and parameter types (sacc.Sacc, list[stat.TrivialStatistic]) for mypy coverage. No test logic changes.
Add explicit parameter and return types to two-point tests and fixtures. Rename and simplify the lens0 no-data fixture to return only its SACC tracer. Clarify test names, variables, and docstrings to distinguish theta-grid setup without measurements from physical predictions.
Annotate test function signatures and fixtures in test_const_gaussianPM.py so mypy checks the file. Add typed imports (Callable, cast, TwoPointFactory, GuardedStatistic). Wrap mock statistic containers in cast(GuardedStatistic, ...) where they are passed to PM internals, and annotate the attributes of IncompleteStatistic. No runtime behavior changes.
Annotate fixtures, test functions, and their parameters in tests/likelihood/test_weak_lensing.py with Mock, WeakLensingArgs, and None return types so the module can be checked by mypy.
Compose fixture annotations from redshift distribution and windowed spectrum aliases, and reuse them across Gaussian and metadata tests. Keep NumCosmo imports type-only and complete ConstGaussian test annotations.
Add type hints to test functions, fixtures, and helper callbacks in tests/fctools so they pass mypy strict checking. Parameters are annotated with pytest types (MonkeyPatch, CaptureFixture) and Path, and mock callbacks with requests.Session and Callable. Move the Path import in test_symbol_reference_checker.py to module level, and give the invalid symbol-map test a typed local variable. No test logic changes.
Add parameter and return annotations to test functions, fixtures, and helpers across tests/, so the test suite passes stricter mypy checks. Annotations use project types where available (ModelingTools, TrivialStatistic, CCLCreationMode, sacc.Sacc) and numpy.typing arrays. Where annotations cannot express the code's behavior, add targeted suppressions and a cast: - test_descriptors.py: __str__ returning None is kept as a deliberate negative-test double, marked with type: ignore[override]. - test_modeling_tools_ccl_factory.py: cast the cosmology dict to dict[str, float] before passing it to ParamsMap. - test_utils.py: read pytestmark via type: ignore[attr-defined] in the two YAML serialization tests that check parametrization coverage. No runtime behavior changes; only test annotations, a cast, and suppression comments were added, plus imports needed for the annotations.
Annotate parameters and return types in the likelihood test scripts, lkdir fixtures, and test functions under tests/likelihood/. Script factories now declare NamedParameters inputs and typed (likelihood, ModelingTools) return tuples. Test helpers in test_binned_cluster_number_counts.py and test_student_t.py take typed arguments, with Completeness and Purity imported from crow. No runtime behavior changes.
Comment out the `[[tool.mypy.overrides]]` block for `tests` that set `disallow_untyped_defs = false`. Tests now inherit the global `disallow_untyped_defs = true`, so mypy requires type annotations on test function definitions. The commented-out replacement line sets `disallow_untyped_defs = true`, which matches the inherited value.
marcpaterno
force-pushed
the
improve-type-checking-in-tests
branch
from
October 25, 2026 01:06
af86271 to
45c7d5a
Compare
Collaborator
|
Hello @marcpaterno , three things:
|
Import numcosmo_py.Nc directly in tests/_types.py so NumCosmoCosmologyFixture annotations no longer need string forward references. Remove the unused map_cosmo_spec_nl fixture from test_numcosmo_mapping.py.
vitenti
approved these changes
Oct 27, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Enable strict mypy checking for
tests/and give test functions, fixtures, mocks, and callbacks explicit types. Shared fixture aliases and records intests/_types.pykeep the annotations consistent. The follow-up annotations in the cluster, relocation, and base-likelihood tests are folded into the bottom commit.This PR changes tests and developer guidance; it does not change library code or public API.
Evidence
disallow_untyped_defs; the remaining annotations produced mypy errors.make lintpasses, including mypy across 339 source files. The 36 targeted tests pass.Merge Danger
Door: Two-way. The changes can be reverted without a data migration.
Blast Radius: Tests. Strict checking may expose future untyped test definitions in CI.