Struct PgSink 

pub struct PgSink { /* private fields */ }

Implementations

impl PgSink

pub async fn new(client: PgClient, limits: SinkLimits) -> Result<Self>

Prepare a connected target client: create the metadata schema and configure the session for replica apply.

pub fn restrict_to(self, tables: &[TableName]) -> Self

Restrict apply to the configured tables (run path). Without this, every published relation that exists on the target is accepted (status/cleanup never apply anything, so they do not need it).

pub fn client(&self) -> &PgClient

pub async fn record_table_copied( client: &PgClient, job_id: &str, schema: &str, table: &str, rows: u64, bytes: u64, ) -> Result<()>

Record a table as copied during the snapshot (inside the caller’s transaction).

Trait Implementations

impl ReplicationSink for PgSink

async fn load_checkpoint(&mut self, job_id: &str) -> Result<Option<Checkpoint>>

Load the job’s checkpoint; None if the job has never been bootstrapped.

async fn save_checkpoint(&mut self, cp: &Checkpoint) -> Result<()>

Persist a checkpoint outside of any data transaction (autocommit).

async fn begin(&mut self) -> Result<()>

Start a destination transaction.

async fn relation(&mut self, rel: Arc<Relation>) -> Result<()>

Relation metadata; may arrive inside or outside a transaction and must be cached.

async fn apply(&mut self, op: Op) -> Result<()>

Apply one change inside the open transaction. Implementations may batch.

async fn commit(&mut self, cp: &Checkpoint) -> Result<()>

Flush pending changes, write cp, and commit atomically. After Ok, the data and the checkpoint are durable.

async fn rollback(&mut self) -> Result<()>

Abort the open transaction, if any. Best effort; must not panic.