Struct TableWrite 

pub struct TableWrite { /* private fields */ }

One write of a new table’s first version or of a table’s next version, under its own write attempt. Large payloads upload into the attempt first, as blobs; rows appended afterwards carry them, encoded with Blob::array. A write of a table’s next version can also patch and delete the base version’s rows by row ID. TableWrite::finish returns the written version, one version for every part, which is published only once a Commit stages it and publishes.

Implementations

impl TableWrite

pub fn attempt(&self) -> Uuid

The write attempt whose directory holds this write’s files and uploads.

pub async fn upload_blob( &self, media_type: &str, input: BlobStream, ) -> Result<Blob, Error>

Upload input as one blob of this write, however large, in parts of 64 MiB sent several at once. Each part is retried on its own, and a failed upload is aborted. The blob’s digest is the BLAKE3 digest of the whole payload.

pub fn blob_pack(&self) -> BlobPack<'_>

Begin a pack of payloads to store as blobs of this write, packed together as the inline payloads of appended rows are.

pub async fn append(&mut self, input: impl IntoArrowStream) -> Result<(), Error>

Append rows, which may carry blobs this write uploaded.

pub async fn patch_rows( &mut self, rows: impl IntoArrowStream, ) -> Result<(), Error>

Record new values for existing rows beside the data: each row names its target in a ROW_ID_COLUMN column, and its other columns are columns of the table, with their stored types, whose values it replaces. A row ID appears at most once per write.

pub async fn delete_rows( &mut self, ids: impl IntoArrowStream, ) -> Result<(), Error>

Delete existing rows, named by their row IDs in a ROW_ID_COLUMN column, which is the input’s only column. Each row must be visible in the base version.

pub async fn finish(self) -> Result<Table, Error>

Prepare the version from every part this write took and register it.