Skip to content

Fixes by format action - #9

Open
github-actions[bot] wants to merge 6 commits into
feat/keops-algorithmfrom
python-code-format-patches
Open

github-actions[bot] wants to merge 6 commits into
feat/keops-algorithmfrom
python-code-format-patches

Conversation

@github-actions

@github-actions github-actions Bot commented Jul 6, 2026

Copy link
Copy Markdown

This is an auto-generated PR with fixes.

@github-actions
github-actions Bot force-pushed the python-code-format-patches branch 2 times, most recently from 7b370f9 to 275a24c Compare July 14, 2026 14:38
@github-actions
github-actions Bot force-pushed the python-code-format-patches branch from 275a24c to 88ec956 Compare July 16, 2026 10:55
@github-actions
github-actions Bot force-pushed the python-code-format-patches branch from 88ec956 to 11adaad Compare July 23, 2026 14:13
passscoed and others added 2 commits October 23, 2026 09:12
…uard

- DeferredLinearOperator: __init__ set self._matvec/self._rmatvec as instance
  attributes, permanently shadowing the class methods that guard against
  using an unconfigured operator. Confirmed: calling .matvec() on a fresh,
  unconfigured DeferredLinearOperator() silently fell through instead of
  raising NotImplementedError. Removed the shadowing attributes so the class
  methods actually run. That, in turn, exposed a second bug: the guard probed
  "is this configured" by calling the real mfunc/rmfunc with a bare 0 and
  comparing the result to the string "not implemented yet" - which crashes or
  raises "truth value of an array is ambiguous" for any real, properly
  configured transform (confirmed via the existing nfft_U.py test). Replaced
  the string-sentinel probe with a simple None check on _matvec_func/
  _rmatvec_func, set once at construction/`.set()` time - no probing call
  needed. Also removed dead code in the old _matvec/_rmatvec bodies
  (print(self.LO @ x) referencing a nonexistent attribute, and
  self @ x / self.H @ x, which would have recursed infinitely had they ever
  been reached).

- Added missing `raise` in front of several validation ValueError/RuntimeError
  constructions that were built but never raised (GroupedTransform.py x7,
  GroupedTransforms.py, CWWTtools.py). Most guard the "mixed"/NFMT system,
  which the code already flags as untested/unfinished; CWWTtools.indextoN's
  check is on the supported wavelet path. NFMTtools.py needed special
  handling: it's imported unconditionally at package import time, so a
  top-level `raise` would break importing pyGroupedTransforms entirely -
  replaced with a module-level __getattr__ that raises lazily only when the
  (unimplemented) module is actually used.

- Fixed a typo (s.bdanwidths -> s1.bandwidths) and a use-before-assignment of
  `s` in GroupedTransform.get_matrix()'s "mixed" branch.

Verified: full existing test suite (tests/run_tests.py) still passes.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
requires-python is ">=3.9", but scipy>=1.16.0 has no wheel for Python 3.9
or 3.10 (verified via pip download) - the declared floor was already
unsatisfiable in practice, and CI never tested below 3.11 to catch it.
Nothing in this package's own code needs anything past very old, stable
scipy/numpy APIs. Lowered to numpy>=1.26.4 and scipy>=1.11.4 (both have
wheels back to Python 3.9) and verified the full test suite against exactly
those versions.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
@github-actions
github-actions Bot force-pushed the python-code-format-patches branch from 11adaad to 1956b7a Compare October 23, 2026 09:15
passscoed and others added 2 commits October 23, 2026 09:33
"Python 3.9 - ubuntu-latest" was failing at the "Set up Python" step itself
(before any dependency installation), with:

  Error: The version '3.9' with architecture 'x64' was not found for
  Ubuntu 26.04.

ubuntu-latest now resolves to Ubuntu 26.04, and actions/setup-python has no
prebuilt Python 3.9 or 3.10 binary for it (per the upstream
actions/python-versions manifest, those only go up to 22.04/24.04 - Python
3.9 itself has been EOL since October 2025). Confirmed this is new/infra-side
and not caused by anything in this PR: the same job passed fine on `develop`
in February 2026. 3.11/3.12 still resolve fine on ubuntu-latest (apparently
via a fallback), so left those alone.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
@github-actions
github-actions Bot force-pushed the python-code-format-patches branch from 1956b7a to 5248aa5 Compare October 23, 2026 11:20

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants