Struct PgDocumentBuilder 

pub struct PgDocumentBuilder { /* private fields */ }

Builds index documents from a Postgres database, driven by a SourceSpec — the enabled indexes and their schemas, translated from the top-level config by the composition root. Cheap to clone — the pool, spec, and primary-key cache are shared.

Implementations

impl PgDocumentBuilder

pub fn new(pool: PgPool, spec: Arc<SourceSpec>) -> Self

Create a builder over a connection pool and the source spec.

pub async fn connect( connection_url: &str, spec: Arc<SourceSpec>, ) -> Result<Self>

Connect a pool from a Postgres connection URL and build over it.

Trait Implementations

impl Catalog for PgDocumentBuilder

The Postgres source’s view of its own catalog. The index mapping is derived from the self-describing schema in schema_core; this is the one store-specific piece used for validation — how Postgres types and constrains a column — so a declared schema can be checked against the live database.

fn column<'life0, 'life1, 'life2, 'life3, 'async_trait>( &'life0 self, schema: &'life1 DatabaseSchema, table: &'life2 TableName, column: &'life3 ColumnName, ) -> Pin<Box<dyn Future<Output = Result<ColumnInfo>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait, 'life3: 'async_trait,

The type and nullability of column in table (within schema), as the store defines it.

impl Clone for PgDocumentBuilder

fn clone(&self) -> PgDocumentBuilder

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for PgDocumentBuilder

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl DocumentBuilder for PgDocumentBuilder

fn resolve<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, table: &'life1 TableName, key: &'life2 RowKey, ) -> Pin<Box<dyn Future<Output = Result<Vec<DocumentId>>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

The documents the changed row affects. Empty if it touches nothing any index cares about.

fn build<'life0, 'life1, 'async_trait>( &'life0 self, id: &'life1 DocumentId, ) -> Pin<Box<dyn Future<Output = Result<Document>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Assemble one document, or report it deleted if its root row is absent.

fn build_many<'life0, 'life1, 'async_trait>( &'life0 self, ids: &'life1 [DocumentId], ) -> Pin<Box<dyn Future<Output = Result<Vec<Document>>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Assemble many documents at once. Returns one Document per requested id — an Upsert, or a Delete tombstone when the root row is absent — in any order; callers match results back by Document::id.

fn backfill_scopes(&self) -> Vec<IndexScope>

The enabled indexes this builder serves, each with the root table to snapshot when seeding it. The engine uses this to scope an initial backfill per index. The default is empty — a builder with no backfillable indexes, which the engine simply never seeds.

fn index_mappings<'life0, 'async_trait>( &'life0 self, ) -> Pin<Box<dyn Future<Output = Result<Vec<IndexMapping>>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait,

The resolved mapping of every index this builder serves: each field typed from the schema’s explicit mapping where one is given, and from the source’s own column types otherwise. Sinks that own their index use this to create it up front. The default is empty — a builder that leaves index creation to whatever the sink does on first write.