v52

latestOpenAPI 3.1.1Apache 2.0raw.githubusercontent.com2026-08-0354155395.9 KB

Table

Data

Merge insert (upsert) records into a table

Performs a merge insert (upsert) operation on table id. This operation updates existing rows based on a matching column and inserts new rows that don't match. It returns the number of rows inserted and updated.

For tables that have been declared but not yet created on storage (is_only_declared=true), this operation will create the table with the provided data (since there are no existing rows to merge with).

REST NAMESPACE ONLY REST namespace uses Arrow IPC stream as the request body. It passes in the MergeInsertIntoTableRequest information in the following way:

id: pass through path parameter of the same name
on: pass through query parameter of the same name
when_matched_update_all: pass through query parameter of the same name
when_matched_update_all_filt: pass through query parameter of the same name
when_not_matched_insert_all: pass through query parameter of the same name
when_not_matched_by_source_delete: pass through query parameter of the same name
when_not_matched_by_source_delete_filt: pass through query parameter of the same name

post/v1/table/{id}/merge_insert

Path parameters

idstring required

string identifier of an object in a namespace, following the Lance Namespace spec. When the value is equal to the delimiter, it represents the root namespace. For example, v1/namespace/$/list performs a ListNamespace on the root namespace.

Query parameters

delimiterstring

An optional delimiter of the string identifier, following the Lance Namespace spec. When not specified, the $ delimiter must be used.

branchstring

Optional branch to target. When not specified, the main branch is used. Used by branch-scoped operations that cannot carry a branch field in their request body (Arrow IPC stream and bodyless operations). Operations with a JSON request body carry branch as a body field instead.

onstring required

Lance field path to use for matching rows. Nested fields use dot-separated segments; use backtick-quoted segments for literal dots and double backticks inside quoted segments. Use canonical full paths for display and errors; leaf names alone only identify top-level fields; invalid or unresolved paths should return InvalidInput or TableColumnNotFound.

when_matched_update_allboolean

Update all columns when rows match

when_matched_update_all_filtstring

The row is updated (similar to UpdateAll) only for rows where the SQL expression evaluates to true. Field references must use Lance field path syntax: nested fields use dot-separated segments, literal dots require backtick-quoted segments, and backticks inside quoted segments are doubled.

when_not_matched_insert_allboolean

Insert all columns when rows don't match

when_not_matched_by_source_deleteboolean

Delete all rows from target table that don't match a row in the source table

when_not_matched_by_source_delete_filtstring

Delete rows from the target table if there is no match AND the SQL expression evaluates to true. Field references must use Lance field path syntax: nested fields use dot-separated segments, literal dots require backtick-quoted segments, and backticks inside quoted segments are doubled.

timeoutstring

Timeout for the operation (e.g., "30s", "5m")

use_indexboolean

Whether to use index for matching rows

Response

Result of merge insert operation

{"stackTrail":"components:schemas:MergeInsertIntoTableResponse:properties:context","oasType":"schema","type":"unknown","description":"Arbitrary context as key-value pairs.\nHow to use the context is custom to the specific implementation.\n\nOn a request, it carries caller-provided context to the implementation.\nOn a response, it carries implementation-provided context back to the caller.\n\nREST NAMESPACE ONLY\nContext entries are mapped to and from HTTP headers using the `header.` prefix:\n- On a request, any entry whose key starts with `header.` is sent as an HTTP\n request header with the prefix stripped. For example, the entry\n `{\"header.Authorization\": \"Bearer abc\"}` is sent as the request header\n `Authorization: Bearer abc`.\n- On a response, every HTTP response header is returned as an entry whose key is the\n header name prefixed with `header.`. For example, the response header\n `x-request-id: abc123` is returned as the entry `{\"header.x-request-id\": \"abc123\"}`.\n"}

transaction_idstring

Optional transaction identifier

num_updated_rowsinteger

Number of rows updated

num_inserted_rowsinteger

Number of rows inserted

num_deleted_rowsinteger

Number of rows deleted (typically 0 for merge insert)

versioninteger

The commit version associated with the operation

Example response

{
  "transaction_id": "transaction_id",
  "num_inserted_rows": 0,
  "num_updated_rows": 0,
  "context": {
    "key": "context"
  },
  "num_deleted_rows": 0,
  "version": 0
}