mortie 1.0 floor bump: migrate retired call sites (arrow from_wkbs rename; shardmap fallback off morton_coverage_moc)

[espg]

🤖 from Claude

Filed at espg's direction (in-session 2026-08-24), from the mortie PR espg/mortie#195 question-6/7 rulings. mortie's API consolidation (merging soon; next mortie tag after it is 1.0.0) retires two names zagg calls today, so the zagg floor bump to mortie 1.0 carries two mechanical call-site migrations:

1. src/zagg/catalog/sources.py:614/623 — from mortie.arrow import from_wkbs → from_wkb (renamed under the one-name-per-operation invariant; the column-in batch shape is unchanged). Migration note from the mortie CHANGELOG: tolerance= must be passed as a keyword on the new spelling — never positionally (the third positional is now moc; mortie also grows a None/bool type refusal on moc that makes a positional float loud rather than silent).
2. src/zagg/catalog/shardmap.py:383/395 — the per-ring fallback arm calls the retired scalar morton_coverage_moc(lats, lons, order=). Migrate to a surviving public route: polygons_to_morton_mocs on the single ring (the hot path one line up already uses it — likely the smallest diff), or the geometry route (from_geometry(..., moc=True)). Do NOT import the surviving private _morton_coverage_moc.
3. Sweep docstrings/comments naming the retired spellings (shardmap.py:22/44/372/1048/1600, sources.py:574/609).

Gated on: mortie 1.0.0 release + the zagg pyproject.toml mortie floor bump (one PR: floor + both call sites + docstring sweep + tests green). No behavior change intended — pin parity between old and new call forms in the tests if practical.

Refs: espg/mortie#195 (retirement + rulings), espg/mortie#187, zagg PR #400 (the batch path), zagg PR #447 (the unindexed from_wkbs path).

[espg]

🤖 from Claude

Still gated — mortie 1.0.0 is not released. Checked the PyPI release index for mortie on 2026-08-26: latest is 0.9.11, and no 1.0.0 exists yet. This issue's own gate is "mortie 1.0.0 release + the zagg pyproject.toml mortie floor bump", so it is not workable today and is being left out of the 2026-08-26 small-fix bundle rather than half-landed.

Confirming the two call sites are still on the retired spellings at main (c56221b), so the migration list is unchanged:

• src/zagg/catalog/sources.py:614 / :623 / :625 — from mortie.arrow import from_wkbs, called twice (the indexed and the keep-filtered arm).
• src/zagg/catalog/shardmap.py — morton_coverage_moc still named in the module docstring (:22), the order-clamp comment (:44) and the per-granule comment (:552), alongside the fallback arm itself.

Both still resolve against the current floor (mortie>=0.9.10, pyproject.toml:89), so nothing is broken today — this is purely pre-staged work waiting on the upstream tag.

Two things worth noting for whoever picks it up:

1. The floor bump is a dependency change. CLAUDE.md §4 requires that be raised before it lands. It is arguably already sanctioned by this issue's text, but a bump from >=0.9.10 to >=1.0.0 is a real floor move on the lambda extra's exact pins too (build_layer.sh derives MORTIE_SPEC from pyproject.toml per issue Move the mortie decimal-parse boundary off the private _decimal_to_word #322), so it wants an explicit yes rather than an inferred one.
2. The layer rebuild is the part with teeth. Issue ATL03 worker memory doubled between the 0.48 and 0.52 fleets (47/251 NEON shards OOM at 4096 MB) — not #537, bisect layer-vs-code #538 is currently tracking a worker-memory doubling whose better suspect is exactly a layer-resolved mortie change (0.9.9 → 0.9.11) rather than zagg's own code. Bumping the floor to 1.0 moves the layer again and would confound that bisect if the two land in the wrong order. Recommend sequencing this after ATL03 worker memory doubled between the 0.48 and 0.52 fleets (47/251 NEON shards OOM at 4096 MB) — not #537, bisect layer-vs-code #538's layer-vs-code bisect resolves.

Applying waiting — nothing for the routine to do until the upstream tag exists. It will be picked up on a later run once 1.0.0 is on PyPI and the floor bump is signed off.

---

Generated by Claude Code