AGENT SKILL

firebase-firestore-data-modeling

Model Firestore data for the queries you actually run, not for normalization

License 许可
Apache-2.0
所属仓库
linktogo/maggie
最近更新
2026-11-14
Stars
★ 2
Forks
2

01 前言 Front Matter

name
firebase-firestore-data-modeling
globs
[ "**/*.ts", "**/*.js", "firestore.indexes.json" ]
description
Model Firestore data for the queries you actually run, not for normalization

03 SKILL.md 正文

Firestore data modeling

Firestore is a NoSQL document store. Model around the reads your app performs; denormalize on purpose rather than mirroring a relational schema.

Rules

  • Design collections from your query patterns first. There are no joins — if you need data together, store it together (embed) or duplicate the few fields you render in lists.
  • Keep documents small and bounded. Use subcollections for unbounded growth (a messages subcollection under a chat) instead of a growing array field.
  • Every query must be backed by an index. Composite queries need a composite index declared in firestore.indexes.json — add it rather than relying on the console's auto-suggest link in production.
  • Denormalized copies drift: update all copies in a single batched write or a transaction, and keep the write path that maintains them in one place.
  • Avoid document-level write hotspots (a single counter hammered by everyone); use distributed counters (sharded) or increment on the server.
  • Paginate with cursors (startAfter(lastDoc)), never large offsets — Firestore bills and scans every skipped document.

Anti-patterns

  • A relational schema with "foreign key" ids you then N+1 fetch client-side.
  • Unbounded arrays inside a document (they must be read/written whole and hit the 1 MiB document limit).
  • Reading a whole collection to count it — keep a maintained counter instead.

04 同仓库更多 Skills