Skip to content

faq.md

Latest commit

 

History

History
178 lines (103 loc) · 5.98 KB

faq.md

File metadata and controls

178 lines (103 loc) · 5.98 KB

SwiftSync FAQ

This is the single FAQ for SwiftSync.

Source-of-truth docs:

  • docs/project/parent-scope.md (parent-scoped sync/query behavior)
  • docs/project/property-mapping-contract.md (mapping/import/export semantics)
  • docs/project/reactive-reads.md (@SyncQuery / @SyncModel mental model)
  • docs/project/relationship-integrity.md (many-to-many inverse anchor rule)
  • docs/project/backend-contract.md (recommended backend shape)

If an answer needs more than a short explanation, this FAQ points to the source-of-truth doc instead of duplicating it.

1) Do I need to import multiple modules?

No. Use only:

import SwiftSync

2) Can two different parents have children with the same id?

Yes, if the model uses parent-scoped identity.

  • ParentScopedModel defaults to .scopedByParent
  • explicit parent-relationship sync on non-ParentScopedModel types defaults to .global
  • @Attribute(.unique) on raw id still enforces global uniqueness at the store level

See docs/project/parent-scope.md.

3) Why does parent-scoped sync need relationship?

Because SwiftSync must know which child->parent relationship defines the sync scope (especially for scoped delete/diff).

  • parent-scoped sync always passes an explicit relationship: key path

See docs/project/parent-scope.md.

4) How do I think about @SyncQuery filtering?

Use this mental rule:

  • relationship: + relationshipID: = relationship-scoped query by ID
  • relationship: always names the relationship path used by the query
  • predicate: = custom business filters or scalar-only filters

See docs/project/reactive-reads.md.

5) What does refreshOn: mean in @SyncQuery?

It expands what related model changes should invalidate/refetch the query.

Use it when the UI reads related data that is not part of the query model's own scalar fields.

See docs/project/reactive-reads.md.

6) Is missing field the same as null?

No.

  • missing key => no-op (absent = ignore)
  • explicit null => clear
  • explicit [] for to-many => clear membership

See docs/project/property-mapping-contract.md and docs/project/backend-contract.md.

7) How strict is relationship FK (*_id) linking?

Strict by type.

  • Int FK links to Int identity
  • String FK links to String identity
  • no cross-type coercion at relationship-link step

Scalar attribute coercion is broader; relationship FK linking is intentionally stricter.

See docs/project/property-mapping-contract.md.

8) Does SwiftSync support ordered to-many relationship syncing?

Not as ordered relationship semantics.

Treat to-many sync as unordered membership and store an explicit scalar order field if UI order matters.

See docs/project/property-mapping-contract.md and docs/project/backend-contract.md.

9) How do key mapping defaults work (snake_case, camelCase, @RemoteKey)?

Convention-first.

  • inbound key style is configured once at SyncContainer (.snakeCase default, .camelCase optional)
  • @RemoteKey overrides conventions
  • import/export follow the same mapping contract

See docs/project/property-mapping-contract.md.

10) Are any model property names blocked?

Yes for @Syncable models: description, hashValue.

Use names like descriptionText / hashValueRaw, and map backend keys with @RemoteKey if needed.

See docs/project/property-mapping-contract.md.

11) Can I control relationship insert/update/delete work per sync call?

Yes. Use relationshipOperations (default: .all).

Use this when you want row updates but need to narrow relationship-side work for a specific sync pass.

12) What if payload has duplicate rows with the same identity?

Later payload rows win (payload order is applied in order).

13) What if local storage already has duplicate rows for the same identity?

SwiftSync deduplicates local identity collisions during sync and keeps one logical row per identity.

14) What if a row has missing or null primary key?

That row is skipped for matching/diffing. Sync continues for valid rows.

15) What happens when a scalar payload value is null?

  • optional scalar -> nil
  • non-optional primitive scalar -> default fallback ("", 0, false, epoch date, zero UUID)

See docs/project/property-mapping-contract.md.

16) What happens if two sync calls run at the same time?

SwiftSync serializes sync calls per ModelContainer.

  • same ModelContainer => queued, no overlap
  • different stores => can run concurrently

This is enforced by an internal per-container sync lease, not by app code needing to add its own lock.

17) How do I cancel a sync?

Use Swift Concurrency task cancellation (task.cancel()).

SwiftSync cooperatively cancels and rolls back unsaved in-memory work for that run.

18) In app code, should views perform saves directly?

Recommended pattern: no.

  • views collect intent and display local reactive state
  • a domain/service layer performs backend mutations and syncs local storage
  • views re-render from @SyncQuery / @SyncModel

See docs/project/reactive-reads.md ("App Best Practices").

19) Should I pass SwiftData model objects between views/sheets?

Recommended default: no.

  • pass IDs/scalars
  • child view owns/queries the data it needs
  • avoid model-object handoff across navigation/sheet boundaries

This keeps view ownership explicit and avoids stale retained-reference assumptions.

See docs/project/reactive-reads.md ("App Best Practices").

20) What is the inverse rule for many-to-many relationships?

The corrected rule is narrower than what we first documented:

  • this is a many-to-many issue (not a general all-to-many issue)
  • one-to-many relationships work fine without explicit inverses
  • many-to-many pairs should have one explicit inverse anchor (@Relationship(inverse: ...)) on either side
  • do not force both sides if SwiftData throws the circular @Relationship macro compiler error

We removed the earlier broad @Syncable warning/allowlist because it encoded the wrong rule.

See docs/project/relationship-integrity.md.