[npillar]

Summary

Accept arrays with degenerate size-1 "middle" dims between the time axis and the
trailing spatial dims in the polygon reader — e.g. [time, 1, lat, lon].

• assertArrayRank now accepts rank = expected + k when the k middle dims
  are all size 1, and returns k. A non-singleton middle dim (a genuine
  multi-level axis, e.g. a depth profile) still throws SliceError, now with a
  message stating the size-1 rule.
• readPolygon selects index 0 for each singleton middle dim in the per-step
  read, collapsing the dim. The block stays C-order over [rows, cols], so the
  mask gather (values[localR * rowStride + localC]) is untouched.
• Mirrors the point-read path's existing rank-4 singleton handling
  (i4sea-aurora-ui#768). npoints (rank 2) and plain rank-3 behavior unchanged.

Motivation

Closes #18. Hydrodynamic current fields (current_vel, current_dir) from
hidro/Delft3D datasets carry a degenerate depth dim of size 1 — shape
[time, 1, lat, lon]. The polygon reader asserted an exact rank match
(2d/1d ⇒ rank 3), so these fields could not stream through it at all and were
skipped-with-warn consumer-side. That meant WATER_FLOW area meteograms served
water_level but lost the current speed + direction pair — a product-visible
gap, since direction-over-speed is a headline of that feature (i4sea-aurora-ui
spec 058 / #1049).

Index-0 selection of a size-1 dim is deterministic (one element), so no new API
surface is needed. An explicit per-dim selection (for non-singleton middle dims,
e.g. picking a depth level) would be a separate, bigger feature and is out of
scope here.

Test Plan

• Unit tests added/updated — new singleton middle dims suite:
  [time, 1, lat, lon] streams identically to the equivalent rank-3 array
  (cells/bbox/stride/values); 2d layout with a leading singleton depth dim;
  two singleton middle dims ([time, 1, 1, lat, lon]); a non-singleton middle
  dim ([time, 3, lat, lon]) still throws with the size-1 rule in the message.
• Integration tests added/updated — n/a (reads through existing stores)
• All existing tests pass (npm test) — 391 passed; the one failing
  s3-store test is pre-existing and unrelated (missing optional peer
  @smithy/node-http-handler in this environment), confirmed via git stash.

[Copilot]

Pull request overview

This PR updates the polygon reader to accept Zarr arrays that include one or more degenerate (size-1) middle dimensions between the leading time axis and trailing spatial axes (e.g. hydrodynamic fields shaped [time, 1, lat, lon]), collapsing those singleton dims deterministically by selecting index 0.

Changes:

• Extend assertArrayRank to allow rank = expected + k when the k “middle” dims are all size 1, and to throw a clearer SliceError when any middle dim is non-singleton.
• Update readPolygon to inject index-0 selections for each singleton middle dim so per-step reads stream identically to the equivalent rank-3 arrays.
• Add unit tests covering [time, 1, lat, lon], [time, 1, 1, lat, lon], 2d layouts with singleton depth, and rejection of non-singleton middle dims; document the fix in the changelog.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
src/spatial/polygon-reader.ts	Allows and collapses singleton middle dims during per-timestep polygon reads; improves rank-validation error messaging.
tests/unit/polygon-reader.test.ts	Adds coverage for singleton-middle-dim acceptance and ensures non-singleton middle dims still throw with the new rule.
CHANGELOG.md	Documents the polygon-reader fix for singleton-depth arrays and references issue #18.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.