array_agg function
View as MarkdownThe array_agg(value) function aggregates values (including nulls) as an array.
The input values to the aggregate can be filtered.
Syntax
array_agg ( <value>
[ORDER BY <col_ref> [ASC | DESC] [NULLS FIRST | NULLS LAST] [, ...]]
)
[FILTER (WHERE <filter_clause>)]
| Syntax element | Description |
|---|---|
<value>
|
The values you want aggregated. |
ORDER BY <col_ref> [ASC | DESC] [NULLS FIRST | NULLS LAST] [, …]
|
Optional. Specifies the ordering of values within the aggregation. If not specified, incoming rows are not guaranteed any order. |
FILTER (WHERE <filter_clause>)
|
Optional. Specifies which rows are sent to the aggregate function. Rows for which the <filter_clause> evaluates to true contribute to the aggregation. See Aggregate function filters for details.
|
Signatures
| Parameter | Type | Description |
|---|---|---|
| value | any | The values you want aggregated. |
Return value
array_agg returns the aggregated values as an array.
Any ORDER BY applied before the aggregate function is evaluated, such as in
a feeding subquery, is ignored. Unless ORDER BY is included in the aggregate
function call itself, the order in which the values are aggregated is
unspecified.
Details
Usage in dataflows
While array_agg is available in Materialize, materializing array_agg(values)
is considered an incremental view maintenance anti-pattern. Any change to the data
underlying the function call will require the function to be recomputed entirely,
discarding the benefits of maintaining incremental updates.
Instead, we recommend that you materialize all components required for the
array_agg function call and create a non-materialized view using array_agg
on top of that. That pattern is illustrated in the following statements:
CREATE MATERIALIZED VIEW foo_view AS SELECT * FROM foo;
CREATE VIEW bar AS SELECT array_agg(foo_view.bar) FROM foo_view;
Examples
SELECT
title,
ARRAY_AGG (
first_name || ' ' || last_name
ORDER BY
last_name
) actors
FROM
film
GROUP BY
title;