Skip to content

backend-contract.md

Latest commit

 

History

History
103 lines (69 loc) · 3.37 KB

backend-contract.md

File metadata and controls

103 lines (69 loc) · 3.37 KB

Backend Contract For Minimal SwiftSync Boilerplate

This is the recommended API contract when backend and iOS are co-designed for SwiftSync.

The goal is simple:

  • near-zero model mapping boilerplate
  • predictable partial-update semantics
  • deterministic sync behavior

1) Identity Rules

Use one stable identity key per resource:

  • default: id
  • keep identity type stable per resource (Int stays Int, String stays String)

For parent-scoped resources, treat identity as (parent_id, id) at the API-contract level.

2) Naming Rules

Use snake_case API keys that map cleanly to Swift camelCase:

  • scalar: display_name -> displayName
  • to-one FK: <relation>_id (example: assignee_id)
  • to-many FK: <relation>_ids (example: watcher_ids)
  • nested objects/arrays: relation key itself (example: assignee, members)

This minimizes @RemoteKey usage.

3) Null/Missing Semantics (Required)

Treat these as different states:

  • key missing => no-op
  • key present with null => clear
  • key present with [] on to-many => clear membership

Strong rule:

  • clear/remove/delete intent must be explicit (null / []), never inferred from omission

Implementation guidance:

  • serializers must not silently drop explicit nulls
  • write contract tests for null/missing behavior per endpoint

4) Relationship Shape Policy

Default to FK payloads for sync endpoints:

  • to-one: *_id
  • to-many: *_ids

Use nested relationship objects only when you intentionally want child upsert behavior in the same payload.

Do not mix relationship shapes unpredictably within the same endpoint contract.

5) Parent Scope Rules

For parent-scoped endpoints:

  • include parent_id explicitly, or guarantee scope by endpoint path and keep it stable
  • apply delete/missing-row semantics only within that parent scope

6) Ordering Policy (Current SwiftData Reality)

Do not rely on ordered relationship semantics.

Use explicit scalar order fields when order matters:

  • position or sort_index
  • query via sortBy in SwiftSync read layer

7) Sync Metadata

Include updated_at on all syncable resources.

Recommended next step for incremental sync:

  • add monotonic version/revision per row (or equivalent server sequence)

Optional for soft-delete flows:

  • add deleted_at tombstones with explicit retention policy

8) Integration Guidance

For existing APIs:

  • preserve backward compatibility with versioned endpoints
  • align resource-by-resource to this contract
  • add fixture-based contract tests to avoid drift

Implemented In This Repo

Current demo backend payloads now follow:

  • stable UUID id on all entities (no opaque integer or slug IDs)
  • snake_case naming
  • *_id / *_ids relationship keys
  • explicit null emission for optional fields in task payloads
  • created_at and updated_at on all demo resources
  • no ordered-relationship assumptions

Client-Authoritative Create

Task creation in the demo uses client-authoritative identity:

  • The client generates id (UUID string), created_at, and updated_at before sending.
  • The server validates presence and uniqueness of id; duplicate id returns a validation error.
  • Server-side PATCH endpoints remain authoritative for updated_at on updates.

This pattern is enforced in DemoServerSimulator.createTask(body:): the body dict must include id, created_at, and updated_at or the call is rejected with a validation error.