Skip to content

fix: re-apply verified bug fixes on synced master#180

Open
gitlzzz wants to merge 11 commits into
LopezGroup-ICIQ:masterfrom
gitlzzz:master
Open

fix: re-apply verified bug fixes on synced master#180
gitlzzz wants to merge 11 commits into
LopezGroup-ICIQ:masterfrom
gitlzzz:master

Conversation

@gitlzzz

@gitlzzz gitlzzz commented Jul 2, 2026

Copy link
Copy Markdown
Collaborator

Summary

Re-applies the verified bug fixes from the earlier (discarded) bug-fix PR, rebased onto the current synced master. Each fix is its own commit. Two fixes from the original set were dropped because the synced upstream already resolved them.

Applied (10 fixes)

Fix File
MANIFEST.in casing (src/ATLAS/datasrc/atlas/data) so data/ ships MANIFEST.in
get_cache_path honors XDG_CACHE_HOME (dropped literal $) + test core/code_utils.py, tests/test_code_utils.py
select_structures_fps called with correct args (was shifted positionals) active_learning/active_learning_utils.py
MLIPModelFileData methods defined inside the class workflows/datatypes/mlip.py
species fraction / len(struct) (was / len(symbol)) core/database/diversity_metrics.py
get_vendi_score_db_rbf returns its result core/database/diversity_metrics.py
np.maximum in tanimoto_distance (was np.max) core/database/diversity_metrics.py
is_ase initialized before use in apply_replacement core/utils.py
concave-hull ndarray branch in check_traj_in_domain active_learning/md/atl_process_structure.py
filter base-skip reads the dataframe row (was duplicate empty .info key) core/filtering/structure_filters.py

Dropped (already fixed upstream on the synced master)

  • clusters.py is False → upstream already uses == False in the refactored apply_replacement_cluster_db.
  • can_submit_calculation tuple-as-bool → the buggy file workflows/active_learning.py was deleted upstream; remaining callers already unpack correctly.

Test plan

  • py_compile clean on all changed files
  • pytest tests/157 passed, 0 failed (4 errors are test_structure_viewer.py missing optional PySide6, unrelated)
  • test_gen_db (schema/data/ packaging) and the XDG_CACHE_HOME test pass

gitlzzz and others added 11 commits June 30, 2026 11:09
MANIFEST.in referenced src/ATLAS/data (wrong case), so config_schema.yaml and the rest of data/ were never shipped, breaking atl_gen_configuration_file in installs.
Read os.environ['XDG_CACHE_HOME'] instead of the literal '$XDG_CACHE_HOME'; update the test that pinned the bug.
The fps data-reduction path passed shifted positional args (int as selected_db) -> TypeError. Pass explicit keywords and an empty selected_db.
__init__/architecture/load_model were at module scope (and nested), so the class was non-functional. Remove a bogus kwargs.pop and flush/seek the temp file before torch.load.
Divided the atom count by len(symbol_string) (1 or 2) instead of len(struct), corrupting the species-fraction feature.
The computed score was discarded, so the function always returned None.
np.max's second positional arg is axis, not a second array; element-wise max needs np.maximum.
is_ase was only set in the ase.Atoms branch, raising UnboundLocalError for ATL/pymatgen inputs.
An ndarray hull has len()==n_vertices, falling through to an empty MultiPolygon so every frame was flagged extrapolating. Add the ndim==2 branch (mirrors concave_hull.py).
The base-skip checked the same empty .info key twice; read base/isolated_atom from the row instead. (The except/continue half is already handled upstream.)
fix: re-apply verified bug fixes on synced master
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.

1 participant