MERGE¶
Inserts, updates, and deletes values in a table based on values in a second table or a subquery. This can be useful if the second table is a change log that contains new rows (to be inserted), modified rows (to be updated), and/or marked rows (to be deleted) in the target table.
The command supports semantics for handling the following cases:
Values that match (for updates and deletes).
Values that do not match (for inserts).
Syntax¶
MERGE INTO <target_table> USING <source> ON <join_expr> { matchedClause | notMatchedClause } [ ... ]
Where:
matchedClause ::= WHEN MATCHED [ AND <case_predicate> ] THEN { UPDATE SET <col_name> = <expr> [ , <col_name2> = <expr2> ... ] | DELETE } [ ... ]notMatchedClause ::= WHEN NOT MATCHED [ AND <case_predicate> ] THEN INSERT [ ( <col_name> [ , ... ] ) ] VALUES ( <expr> [ , ... ] )
Parameters¶
target_table
Specifies the table to merge.
source
Specifies the table or subquery to join with the target table.
join_expr
Specifies the expression on which to join the target table and source.
matchedClause
(for updates or deletes)¶
WHEN MATCHED ... THEN UPDATE <col_name> = <expr> | DELETE
Specifies the action to perform when the values match.
AND case_predicate
Optionally specifies an expression which, when true, causes the matching case to be executed.
Default: No value (matching case is always executed)
SET col_name = expr
[ … ]Specifies the column within the target table to be updated or inserted and the corresponding expression for the new column value (can refer to both the target and source relations).
In a single
SET
subclause, you can specify multiple columns to update/delete.
notMatchedClause
(for inserts)¶
WHEN NOT MATCHED ... THEN INSERT
Specifies the action to perform when the values do not match.
AND case_predicate
Optionally specifies an expression which, when true, causes the not-matching case to be executed.
Default: No value (not-matching case is always executed)
( col_name [ , ... ] )
Optionally specifies one or more columns within the target table to be updated or inserted.
Default: No value (all columns within the target table are updated or inserted)
VALUES ( expr [ , ... ] )
Specifies the corresponding expressions for the inserted column values (must refer to the source relations).
Usage notes¶
A single MERGE statement can include multiple matching and not-matching clauses (i.e.
WHEN MATCHED ...
andWHEN NOT MATCHED ...
).Any matching or not-matching clause that omits the
AND
subclause (default behavior) must be the last of its clause type in the statement (e.g. aWHEN MATCHED ...
clause cannot be followed by aWHEN MATCHED AND ...
clause). Doing so results in an unreachable case, which returns an error.
Duplicate join behavior¶
Nondeterministic results for UPDATE and DELETE¶
When a merge joins a row in the target table against multiple rows in the source, the following join conditions produce nondeterministic results (i.e. the system is unable to determine the source value to use to update or delete the target row):
A target row is selected to be updated with multiple values (e.g.
WHEN MATCHED ... THEN UPDATE
).A target row is selected to be both updated and deleted (e.g.
WHEN MATCHED ... THEN UPDATE
,WHEN MATCHED ... THEN DELETE
).
In this situation, the outcome of the merge depends on the value specified for the ERROR_ON_NONDETERMINISTIC_MERGE session parameter:
If TRUE (default value), the merge returns an error.
If FALSE, one row from among the duplicates is selected to perform the update or delete; the row selected is not defined.
Deterministic results for UPDATE and DELETE¶
Deterministic merges always complete without error. A merge is deterministic if it meets the following conditions for each target row:
One or more source rows satisfy the
WHEN MATCHED ... THEN DELETE
clauses, and no other source rows satisfy anyWHEN MATCHED
clausesOR
Exactly one source row satisfies a
WHEN MATCHED ... THEN UPDATE
clause, and no other source rows satisfy anyWHEN MATCHED
clauses.
This makes MERGE semantically equivalent to the UPDATE and DELETE commands.
Note
To avoid errors when multiple rows in the data source (i.e. the source table or subquery) match the target table based on the ON condition, use GROUP BY in the source clause to ensure that each target row joins against one row (at most) in the source.
In the following example, assume src
includes multiple rows with the same k
value. It’s ambiguous which values (v
) will
be used to update rows in the target row with the same value of k
. By using MAX() and GROUP BY, the query clarifies exactly
which value of v
from src
is used:
MERGE INTO target USING (select k, max(v) as v from src group by k) AS b ON target.k = b.k
WHEN MATCHED THEN UPDATE SET target.v = b.v
WHEN NOT MATCHED THEN INSERT (k, v) VALUES (b.k, b.v);
Deterministic results for INSERT¶
Deterministic merges always complete without error.
If the MERGE contains a WHEN NOT MATCHED ... THEN INSERT
clause, and if there are no matching rows in the target, and if the
source contains duplicate values, then the target gets one copy of the row for each
copy in the source. (An example is included
below.)
Examples¶
Perform a simple merge:
Create and load the tables:
CREATE TABLE target_table (ID INTEGER, description VARCHAR); CREATE TABLE source_table (ID INTEGER, description VARCHAR);INSERT INTO target_table (ID, description) VALUES (10, 'To be updated (this is the old value)') ; INSERT INTO source_table (ID, description) VALUES (10, 'To be updated (this is the new value)') ;Execute the MERGE statement:
MERGE INTO target_table USING source_table ON target_table.id = source_table.id WHEN MATCHED THEN UPDATE SET target_table.description = source_table.description; +------------------------+ | number of rows updated | |------------------------| | 1 | +------------------------+Display the new value(s) in the target table (the source table is unchanged):
SELECT * FROM target_table; +----+---------------------------------------+ | ID | DESCRIPTION | |----+---------------------------------------| | 10 | To be updated (this is the new value) | +----+---------------------------------------+ SELECT * FROM source_table; +----+---------------------------------------+ | ID | DESCRIPTION | |----+---------------------------------------| | 10 | To be updated (this is the new value) | +----+---------------------------------------+
Perform a basic merge with a mix of operations (delete, update, insert):
MERGE INTO t1 USING t2 ON t1.t1Key = t2.t2Key WHEN MATCHED AND t2.marked = 1 THEN DELETE WHEN MATCHED AND t2.isNewStatus = 1 THEN UPDATE SET val = t2.newVal, status = t2.newStatus WHEN MATCHED THEN UPDATE SET val = t2.newVal WHEN NOT MATCHED THEN INSERT (val, status) VALUES (t2.newVal, t2.newStatus);
Perform a merge in which the source has duplicate values and the target has no matching values. Note that all copies of the source record are inserted into the target:
Truncate both tables and load new rows into the source table. Note that the rows include duplicates.
TRUNCATE TABLE source_table; TRUNCATE TABLE target_table; INSERT INTO source_table (ID, description) VALUES (50, 'This is a duplicate in the source and has no match in target'), (50, 'This is a duplicate in the source and has no match in target') ;Execute the MERGE statement:
MERGE INTO target_table USING source_table ON target_table.id = source_table.id WHEN MATCHED THEN UPDATE SET target_table.description = source_table.description WHEN NOT MATCHED THEN INSERT (ID, description) VALUES (source_table.id, source_table.description); +-------------------------+------------------------+ | number of rows inserted | number of rows updated | |-------------------------+------------------------| | 2 | 0 | +-------------------------+------------------------+Display the new value in the target table:
SELECT ID FROM target_table; +----+ | ID | |----| | 50 | | 50 | +----+
Merge records using joins that produce nondeterministic and deterministic results:
-- Setup for example. CREATE TABLE target_orig (k NUMBER, v NUMBER); INSERT INTO target_orig VALUES (0, 10); CREATE TABLE src (k NUMBER, v NUMBER); INSERT INTO src VALUES (0, 11), (0, 12), (0, 13); -- Multiple updates conflict with each other. -- If ERROR_ON_NONDETERMINISTIC_MERGE=true, returns an error; -- otherwise updates target.v with a value (e.g. 11, 12, or 13) from one of the duplicate rows (row not defined). CREATE OR REPLACE TABLE target CLONE target_orig; MERGE INTO target USING src ON target.k = src.k WHEN MATCHED THEN UPDATE SET target.v = src.v; -- Updates and deletes conflict with each other. -- If ERROR_ON_NONDETERMINISTIC_MERGE=true, returns an error; -- otherwise either deletes the row or updates target.v with a value (e.g. 12 or 13) from one of the duplicate rows (row not defined). CREATE OR REPLACE TABLE target CLONE target_orig; MERGE INTO target USING src ON target.k = src.k WHEN MATCHED AND src.v = 11 THEN DELETE WHEN MATCHED THEN UPDATE SET target.v = src.v; -- Multiple deletes do not conflict with each other; -- joined values that do not match any clause do not prevent the delete (src.v = 13). -- Merge succeeds and the target row is deleted. CREATE OR REPLACE TABLE target CLONE target_orig; MERGE INTO target USING src ON target.k = src.k WHEN MATCHED AND src.v <= 12 THEN DELETE; -- Joined values that do not match any clause do not prevent an update (src.v = 12, 13). -- Merge succeeds and the target row is set to target.v = 11. CREATE OR REPLACE TABLE target CLONE target_orig; MERGE INTO target USING src ON target.k = src.k WHEN MATCHED AND src.v = 11 THEN UPDATE SET target.v = src.v; -- Use GROUP BY in the source clause to ensure that each target row joins against one row -- in the source: CREATE OR REPLACE TABLE target CLONE target_orig; MERGE INTO target USING (select k, max(v) as v from src group by k) AS b ON target.k = b.k WHEN MATCHED THEN UPDATE SET target.v = b.v WHEN NOT MATCHED THEN INSERT (k, v) VALUES (b.k, b.v);
In the following example, the members
table stores the names, addresses, and current fees (members.fee
) paid to a
local gym. The signup
table stores each member’s signup date (signup.date
). The MERGE statement applies a standard
$40 fee to members who joined the gym more than 30 days ago, after the free trial expired:
MERGE INTO members m USING ( SELECT id, date FROM signup WHERE DATEDIFF(day, CURRENT_DATE(), signup.date::DATE) < -30) s ON m.id = s.id WHEN MATCHED THEN UPDATE SET m.fee = 40;