16 characterization tests against real wgrib2 binary + checked-in HRRR fixtures: extract_grid/3 (bbox + physical sanity on TMP/DPT + longitude convention), extract_grid_from_file/3, extract_grid_from_file_mapped/4 reducer plumbing, extract_grid_messages/3 metadata, extract_points_from_file/3 nearest-neighbor snap, undefined-value sentinel filtering, empty-match and error shapes. Tagged @describetag :slow so they run on 'mix test --include slow' (project default excludes slow, matches existing HRRR/NARR tests). Surfaces one real bug (not fixed): extract_grid_messages variants don't account for the 8-byte Fortran record overhead between messages in wgrib2 -lola output, so per-message values after the first are shifted. The correct offset math exists in parse_lola_binary/3. Tests characterize current behavior (metadata-only); worth a follow-up fix + value-range assertions. |
||
|---|---|---|
| .. | ||
| fixtures | ||
| microwaveprop | ||
| microwaveprop_web | ||
| mix/tasks | ||
| support | ||
| test_helper.exs | ||