Iceland: acceptance tests for the remaining R-checks

art commented

2026-08-15 07:22:14 +00:00

Owner

Part of the #115 plan: the mechanism (config-declared acceptance_tests:, run by the importer on every import and via -acceptance-only) landed in #119 with a representative first set. This issue tracks porting the REST of this country's R-checks into acceptance tests. The general instruction from #115: a test is a sentence citing the R-number, a SQL query whose first column is the pass/fail boolean, and a timeout from {1, 10, 100, 1000}; prefer 1s/10s, put the genuinely slow ones at 100s.

Candidates already measured/written down in docs/country-requirements.md:

R105/R119: every staging table's row count is nonzero after load (cheap; catches the shapefile-polygon-only loader silently skipping rows).
R119/R120: row counts per zoom with no unintended empty partition, and NULL shares in the columns that matter, for the ismap_* per-country layers (contours, peaks, buildings, airports — R13's Iceland-only set).
R46/R49/R50/R51: waterways importance histogram (Horton shape), zero NULL strahler — the shared-contract R46 test covers NULLs but not the ladder shape.
R34: cover area within ~1% of the authoritative national area (metric CRS), the R34 check Iceland is one of the two reference cases for.
#120's fix should land WITH its acceptance test ("0 invalid geometries at maxz") so the regression is caught the same way it was found.

Part of the #115 plan: the mechanism (config-declared `acceptance_tests:`, run by the importer on every import and via `-acceptance-only`) landed in #119 with a representative first set. This issue tracks porting the REST of this country's R-checks into acceptance tests. The general instruction from #115: a test is a sentence citing the R-number, a SQL query whose first column is the pass/fail boolean, and a timeout from {1, 10, 100, 1000}; prefer 1s/10s, put the genuinely slow ones at 100s. Candidates already measured/written down in docs/country-requirements.md: - R105/R119: every staging table's row count is nonzero after load (cheap; catches the shapefile-polygon-only loader silently skipping rows). - R119/R120: row counts per zoom with no unintended empty partition, and NULL shares in the columns that matter, for the ismap_* per-country layers (contours, peaks, buildings, airports — R13's Iceland-only set). - R46/R49/R50/R51: waterways importance histogram (Horton shape), zero NULL strahler — the shared-contract R46 test covers NULLs but not the ladder shape. - R34: cover area within ~1% of the authoritative national area (metric CRS), the R34 check Iceland is one of the two reference cases for. - #120's fix should land WITH its acceptance test ("0 invalid geometries at maxz") so the regression is caught the same way it was found.

art referenced this issue

2026-08-15 07:23:15 +00:00

#115: Formalized acceptance tests. #119

art referenced this issue

2026-08-15 07:23:27 +00:00

Formalized acceptance tests. #115

art-bot was assigned by art

2026-08-15 08:55:04 +00:00

art commented

2026-08-15 09:38:07 +00:00

Author

Owner

Depends on #120 -- the 0 invalid geometries at maxz test in this issue candidate list ships with #120 fix; #120 is ahead of #121 in the queue (2026-08-15).