Nearby Places Search · Approach

01 · Develop the design

Plan the design

Use a spatial index to obtain bounded candidates and authoritative records to apply visibility and current coordinates. Exact distance filtering and a clear snapshot or best-effort contract make the results explainable.

Stage 1

Use a coarse-to-fine search

Enumerate cells intersecting the search circle, retrieve candidates and calculate exact distances under a documented earth model.

Stage 2

Handle geographic boundaries

Normalize longitude and support date-line wrapping and polar coverage. Do not treat latitude and longitude degrees as uniform meters.

Stage 3

Pin the search contract

A stable cursor needs a pinned index snapshot and deterministic tie-breaks; otherwise label continuation as best effort.

02 · Work through the method

Step-by-step approach

Validate finite coordinates, radius and page size. Select an index resolution that bounds coverage without excluding intersecting cells, including geographic wraparound.
Retrieve candidates from every intersecting cell using a supported spatial query. Deduplicate IDs and apply exact radius checks with a tested geospatial distance implementation.
Load authoritative place records, enforce visibility and use current coordinates for final filtering. A stale index may still miss moved-in places, so disclose the freshness contract.
Sort by distance and stable ID. Pin an index snapshot for consistent continuation, or document best-effort pages. Enforce work budgets and mark incomplete results explicitly.

03 · Justify the choices

Why this approach works

Why query multiple cells?

A circle can cross cell boundaries

Cell coverage yields candidates; exact distance filtering removes the excess.

What should a dense query do at its work limit?

Return explicit truncation or require a narrower query

Bounded work must not masquerade as exhaustive nearest-neighbor results.

How should hidden places be handled?

Recheck authoritative visibility before response

Candidate retrieval and authorization have different freshness requirements.

04 · Plan the evidence

Validation strategy

Places across a cell boundary are returned within the radius.
Date-line and polar queries do not drop covered cells.
A hidden record is excluded despite a stale public index entry.
Dense query truncation is explicit and deterministic ties use place ID.

Trade-offs to explain

Fine cells reduce false positives but increase fan-out.
Authoritative filtering prevents disclosure but cannot recover candidates absent from a stale index.
Stable pagination consumes snapshot retention; best-effort pages may shift under updates.