The pre-commit workflow already installs `python-version: "3.x"`, but ruff was
still grading the archive against py310 -- so it kept withholding the pyupgrade
and type-checking fixes that only become safe at the newer target.
`pin_ci_python.py` rewrites the knob.
The 17 findings that appeared are what the stale target was hiding:
`datetime.timezone.utc` where `datetime.UTC` is now correct (UP017), and 14
imports used only in annotations that belong under `TYPE_CHECKING` (TC002,
TC003) -- safe here precisely because 3.14's PEP 649 defers annotation
evaluation, which is the same reason ruff would not offer them at py310.
`ruff check` and `ruff format --check` are clean again, so the zero-findings
state this repo reached earlier today holds at the new target.
Still contradictory and deliberately untouched: the four
`Programming/ERSMS-project/*/Dockerfile` files say `FROM python:3.10-slim`.
They build an archived coursework service against its own era's pinned
dependencies, and nothing in this repo runs them; bumping the base image would
be a rewrite of that project, not a lint-target change.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SfnJmxxsZ534AKk16rC2Gr
598 findings to none, and the suite ran through `./run.sh run` after each step
rather than at the end -- 18 passed, 27s, every time.
The shape of this one was different from the other courses. PORR is four
implementations of the same linear algebra (sequential, threads, processes,
dask), so nearly every finding came in groups of four:
- 229 missing annotations. Rather than `object`, which mypy strict rejects on
every attribute access, each module declares what these actually are:
`Matrix = np.ndarray | list[list[float]]` and the vector equivalent. The
backends really are called with both -- comparing them is the point of the
course.
- 76 names: `A` -> matrix, `Ab` -> augmented, `Ax` -> product, `I` ->
identity, `dA` -> dask_matrix, and the two CamelCase functions. Renamed
through `tokenize`, not regex, so no string or comment was touched.
- 121 missing docstrings, one per method name since the names repeat across
all four backends.
- Seven functions over the argument limit, all taking the same seven or nine
loose values. They are two frozen dataclasses now, in a new
richardson_problem.py: `Problem(matrix, rhs, lambda_min, lambda_max)` and
`Settings(max_iterations, x0, tol)`. `RichardsonMethod.__init__` also took a
`size` argument it never read.
- threads_indep_old's iteration did "split into chunks, submit, wait" four
times by hand (C901 11, PLR0915 54). One `_run_chunked` helper.
- tests.py's two bare asserts became explicit raises, per the no-ignores rule.
`assert False` was also B011: `python -O` deletes it, so the non-convergence
branch would have silently passed under optimisation.
Real fixes rather than style: np.random -> Generator, `os.path.exists` ->
Path, an `except Exception` around dask narrowed to the errors it actually
raises, and two runnable modules that logged at INFO without ever calling
basicConfig -- caught by scripts/check_refactor_safety.py, which is now doing
the job it was written for.
Repo-wide: 1,781 findings -> 1,183, all of them in the five courses left.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0172dj34FDpao6wsix8ZuCzE
PORR was the last course whose `run` did not work. Four measurements, not
guesses, explain why -- and none of them is the matrix size everyone assumes:
ProcessingType.PROCESSES 20.9s at n=2, 32.8s at n=5, 78.2s at n=10. That
is process-pool setup PER TEST, and it grows.
matrix_type poli3 densifies a sparse .mat into a 2.3 GB npz cache
(nemeth12: 723 MB) -- over the 2G ceiling before
any solving happens.
n=10000 800 MB matrix, and np.linalg.solve wants a copy.
n=5000 spd_5000.npz was a Git LFS pointer whose object
404s on GitHub, so it regenerates from scratch.
Remove those and what is left still exercises the whole Richardson
implementation at every size that fits: 18 tests, 28 seconds through the
harness, exit 0. The submitted parameter set is one documented environment
variable away. `main.py` now propagates pytest's status (verified: exit 5 when
nothing is collected) -- it discarded it before, so `run` could not tell a
passing suite from a failing one.
The rest of this commit is scripts/check_refactor_safety.py, which exists
because three regressions got past both ruff and a smoke run during this
campaign, and I would rather they could not get past the next one:
--signatures diffs two trees function by function, parameter by
parameter, default by default. This is what found the 22
EARIN functions whose defaults an annotation pass silently
dropped.
--logging a module with a __main__ guard that logs at INFO must call
basicConfig, or it prints nothing at all. Found two more
in PORR just now.
--print-keywords refuses to let an automated rewrite near print(x, end="")
or print(x, file=sys.stderr); losing those turns a
rendered board into one glyph per line, or silently moves
error output onto stdout.
All three pass on the whole archive. Also in here: 12 latent TypeErrors mypy
found in EARIN that ruff could not -- call sites still passing positionally to
parameters this campaign had made keyword-only, plus one caller of a function
whose argument I had removed. Every one would have raised at runtime.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0172dj34FDpao6wsix8ZuCzE
