Covers the pure-Python helpers and YAML-applied transforms that the
build pipeline relies on. Subprocess calls (mapshaper, NE download)
are not exercised — those are integration concerns covered by the
regen workflow itself.
Test categories:
- _matches (4 tests): scalar equality, AND'd conditions, `in: [...]`
list-membership matcher, missing property
- _bbox_center (2 tests): unit square, offset square
- _translate_and_scale (4 tests): pure translate, scale-around-centroid,
combined transform, multipolygon handling
- _translate_and_scale_with_pivot (1 test): shared pivot preserves
relative positions of grouped features (the Paris-petite-couronne case)
- _drop_parts (2 tests): drops specified indices, polygon unchanged
- _bbox_contains (2 tests): inside-bbox, outside-bbox-west
- apply_name_overrides (1 test): applies only to matching features,
respects match conditions across countries (FRA "Seien" vs GBR "Seien"
don't collide)
- apply_flying_islands (2 tests): repositions matched features,
drop_outside_bbox guarded to Admin 1 only (the bug we fixed earlier)
Wired into the regen workflow as a step that runs BEFORE the build,
so a broken transform fails CI before producing potentially-bad output
files.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
