[Drift] Wildcard schema support added to Sync Streams FROM clause (powersync-service #703)

[benitav]

What's changing: Sync Streams now support a wildcard schema in the FROM clause (e.g. FROM "%".assets), together with new schema(), table_suffix(), and table_name() functions that resolve per-client from the source row. This lets a single stream span every matching Postgres schema, enabling schema-per-tenant replication (one schema per tenant) from a single replication slot without a per-tenant sync-config change.

High priority — existing docs are now inaccurate or incomplete
None identified. Filing under the waiver: the hard precondition is not met (no existing page becomes wrong or misleading), but this is a meaningful new capability with no current doc coverage, and the source PR carries the team's "Product Visibility" label.

For reviewer consideration — may be worth documenting

• Supported SQL guide (https://docs.powersync.com/sync/supported-sql) — the FROM clause is currently only described with plain table names (e.g. FROM users, FROM comments INNER JOIN issues ON ...). There's no mention of schema-qualified sources, wildcard schemas, or the new schema()/table_suffix()/table_name() functions anywhere on this page.
• Grammar Reference (Sync Streams) (https://docs.powersync.com/sync/grammar/sync-streams/index) — the TableSource rule is defined as a bare identifier only (example: users AS u), with no schema qualification. This formal grammar reference doesn't yet reflect the new qualified-schema syntax or the three new functions.
• Partitioned Tables (Postgres) (https://docs.powersync.com/sync/advanced/partitioned-tables) — this page already documents wildcard suffix matching on table names (e.g. "users_%") and a _table_suffix special column, a different mechanism (table-name wildcarding, not schema wildcarding). The new table_suffix() function operates on schema-wildcard sources. Worth a reviewer's judgment on whether the similar naming needs disambiguating if/when this is documented.
• Per the PR discussion, a literal "%" schema name previously only matched a schema literally named % (i.e. matched nothing in practice); it now matches any schema. This is shared with the older Sync Rules compiler via TablePattern.matches(), though the new schema()/table_suffix()/table_name() functions are Sync Streams edition-3 only.
• A follow-on PR (powersync-service#717, open as of filing) introduces sync plan version 2 to support this feature's serialized plan changes, gated so existing deployments not using the new functions are unaffected.

Source PR: powersync-ja/powersync-service#703 — merged 2026-07-15.

Filed automatically by Claude Code (claude-sonnet-5). A human must verify the scope, prepare the docs PR, and publish it only once this update has been released.