Struct TableCommit 

pub struct TableCommit { /* private fields */ }

Table commit logic for Paimon write operations.

Provides atomic commit functionality including append, overwrite and truncate

Implementations

impl TableCommit

pub fn new(table: Table, commit_user: String) -> Self

pub async fn commit(&self, commit_messages: Vec<CommitMessage>) -> Result<()>

Commit new files in APPEND mode.

pub async fn commit_with_identifier( &self, commit_messages: Vec<CommitMessage>, commit_identifier: i64, ) -> Result<()>

Commit new files with a caller-provided commit identifier.

Identifiers must increase monotonically for a given commit_user. All messages for one identifier must be submitted in a single call. This method does not filter previously committed identifiers. Use Self::filter_and_commit_with_identifier when retrying an uncertain commit result.

pub async fn filter_and_commit_with_identifier( &self, commit_messages: Vec<CommitMessage>, commit_identifier: i64, ) -> Result<()>

Filter a previously committed identifier, then commit if it is new.

Identifiers must increase monotonically for a given commit_user. This method is intended for retrying the same uncertain result; regular commits should use Self::commit_with_identifier.

pub async fn overwrite( &self, commit_messages: Vec<CommitMessage>, static_partitions: Option<HashMap<String, Option<Datum>>>, ) -> Result<()>

Overwrite partitions with new data.

When static_partitions is None, extracts the set of partitions touched by commit_messages and overwrites only those (dynamic partition overwrite). When static_partitions is Some, uses the caller-provided partition spec to determine which partitions to replace (static partition overwrite). A partial spec (not all partition keys specified) uses predicate-based filtering so that all matching partitions are overwritten. For unpartitioned tables this is a full table overwrite.

When static_partitions is Some but commit_messages is empty, the specified partitions are truncated (all existing data removed, nothing written).

pub async fn overwrite_with_identifier( &self, commit_messages: Vec<CommitMessage>, static_partitions: Option<HashMap<String, Option<Datum>>>, commit_identifier: i64, ) -> Result<()>

Overwrite partitions with a caller-provided commit identifier.

A previously committed identifier is filtered so retrying an uncertain result cannot re-execute the destructive operation.

pub async fn truncate_partitions( &self, partitions: Vec<HashMap<String, Option<Datum>>>, ) -> Result<()>

Drop specific partitions (OVERWRITE with only deletes).

pub async fn truncate_partitions_with_identifier( &self, partitions: Vec<HashMap<String, Option<Datum>>>, commit_identifier: i64, ) -> Result<()>

Drop specific partitions with a caller-provided commit identifier.

A previously committed identifier is filtered so retrying an uncertain result cannot delete data committed in between.

pub async fn drop_partitions( &self, partitions: Vec<HashMap<String, Option<Datum>>>, ) -> Result<()>

Python-compatible alias for dropping partitions.

pub async fn drop_partitions_with_identifier( &self, partitions: Vec<HashMap<String, Option<Datum>>>, commit_identifier: i64, ) -> Result<()>

Python-compatible alias for dropping partitions with a caller-provided commit identifier. Unlike truncate_partitions, an empty partition list is rejected just like FileStoreCommit.drop_partitions.

pub async fn truncate_table(&self) -> Result<()>

Truncate the entire table (OVERWRITE with no filter, only deletes).

pub async fn truncate_table_with_identifier( &self, commit_identifier: i64, ) -> Result<()>

Truncate the entire table with a caller-provided commit identifier.

A previously committed identifier is filtered so retrying an uncertain result cannot delete data committed in between.

pub async fn abort(&self, commit_messages: &[CommitMessage]) -> Result<()>

Abort a prepared commit by deleting newly written data, changelog and index files.

Deletion is best-effort and mirrors Python FileStoreCommit.abort: missing files or storage errors are ignored so abort cleanup never masks the original write failure.