Table

Per-table staging area for a tx's writes.

Rows are written column-by-column into a shared relation.

writePut / writeDelete / writeErase — and the batch writePuts / writeDeletes / writeErases / patchDocs — wrap this protocol for callers that don't need the column-level (zero-alloc) control.

If you elect to write the values directly, each op must have exactly one value written to each of the columns:

• the entity id — writeId / (internal only) writeIid;

• the valid-time range — writeValidTimes / writeValidTimeMicros / writeDefaultValidTime;

• put only: the document columns, via putDocWriter;

Then, a terminal endPut / endDelete / endErase — this is what records the row.

In columnar style, you may choose to write all your IDs first, then all the valid-times, then all the documents, then the batch endPuts / endDeletes / endErases - it is your responsibility to ensure a consistent number of values in each column.

In any event, you must write all the batch's row data before calling the corresponding end methods.