FOR statement
Applies to: Databricks Runtime 16.3 and later
This feature is in Public Preview.
Repeat the execution of a list of statements for each row returned by query.
This statement may only be used within a compound statement.
Syntax
[ label : ] FOR [ variable_name AS ] query
DO
{ stmt ; } [...]
END FOR [ label ]
Parameters
-
An optional label for the loop which is unique amongst all labels for statements within which the
FOR
statement is contained. If an end label is specified, it must match the beginning label. The label can be used to LEAVE or ITERATE the loop. To qualify loop column references, use thevariable_name
, not thelabel
. -
An optional name you can use as a qualifier when referencing the columns in the cursor.
-
stmt
A SQL statement
Notes
If the query operates on a table that is also modified within the loop's body, the semantics depend on the data source.
For Delta tables, the query will remain unaffected.
Databricks does not guarantee the full execution of the query if the FOR
loop completes prematurely due to a LEAVE
statement or an error condition.
When exceptions or side-effects occur during the execution of the query, Databricks does not guarantee at which point in time within the loop these occur.
Often FOR
loops can be replaced with relational queries, which are typically more efficient.
Examples
-- sum up all odd numbers from 1 through 10
> BEGIN
DECLARE sum INT DEFAULT 0;
sumNumbers: FOR row AS SELECT num FROM range(1, 20) AS t(num) DO
IF num > 10 THEN
LEAVE sumNumbers;
ELSEIF num % 2 = 0 THEN
ITERATE sumNumbers;
END IF;
SET sum = sum + row.num;
END FOR sumNumbers;
VALUES (sum);
END;
25
-- Compare with the much more efficient relational computation:
> SELECT sum(num) FROM range(1, 10) AS t(num) WHERE num % 2 = 1;
25