missing iceland landcover #46
Labels
No labels
claude-wip
map-report
map:is
map:se
needs-review
No milestone
No project
No assignees
2 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
art/ismap#46
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
something went wrong with the northern parts of iceland, everything is cut off past 66.5 north.
Location: z11 / 66.47022 / -16.42047
Feature:
cover— country=is, cover_class=sport_leisureArea: circle r=4560 m @ 66.55541, -18.02804
Location: z9 / 66.42155 / -16.70454
Area: circle r=23959 m @ 66.52587, -16.11064
Investigation complete — #46 is the Iceland manifestation of the #47 absorption bug, and the current live DB already satisfies the report.
Diagnosis (evidence-based):
336635e, min_area_pixels: 16 / min_width_pixels: 1) WITHOUT the absorb_boundary pixel floor. At coarse zoom the pixel-area thresholds are enormous (z4 ~1 531 km²/px, z8 ~6 000 km²), so every small northern polygon — the tapering NE peninsula, Grímsey — becomes a sub-threshold candidate whose only larger neighbour is the sea; it inheritswaterand vanishes. The "66.5°N line" is simply where Iceland narrows below screen-pixel width.Status: no code change needed on this issue — the fix is the #47/#41 absorb_boundary floor, already implemented and live. Closing decision belongs to the maintainer once #55/#56 land (or now, since the live DB satisfies the report). The investigation worker hit its iteration cap before opening a PR; nothing was committed, so this issue is being left blocked rather than falsely released as untouched.
After all branches (except station names in Warsaw) have been merged, the tree was updated to latest main with the #47 fix in. I made sure to disable caches in the developer console, then when I saw that it still used cached tiles, I went into private window and checked that it downloaded new tiles and it did. The result are the two screenshots below.
Grimsey completely vanishes at z12, the url for that image is: https://experiments.blahonga.org/?map=is#12.71/66.54267/-17.99846
The second screenshot is around https://experiments.blahonga.org/?map=is#13.18/66.50164/-16.17706
Grimsey is literally the whole screen, it shouldn't be absorbed or generalized away anyway even if the water absorption bug was still there.
Result is in #61 — #61; the review note is the PR description. In short: the LMI sea polygon is broken at the 66.5°N data cut (self-intersecting, inverted ring roles), the z12 rollup's MakeValid promotes its island loops into marine_water, and the #47 fix never touches it because the island is already water before absorption runs. The PR excludes the broken polygon and computes a valid sea from the official coastline instead. Verified in a real browser at both screenshot locations.
Resolved by PR #61