window
grouping expression
Applies to: Databricks SQL Databricks Runtime
Creates a hopping based sliding-window over a timestamp expression.
Syntax
window(expr, width [, slide [, start] ] )
Arguments
expr
: ATIMESTAMP
expression specifying the subject of the window.width
: ASTRING
literal representing the width of the window as anINTERVAL DAY TO SECOND
literal.slide
: An optionalSTRING
literal representing an offset from midnight to start, expressed as anINTERVAL HOUR TO SECOND
literal.start
: An optionalSTRING
literal representing the start of the next window expressed as anINTERVAL DAY TO SECOND
literal.
Returns
Returns a set of groupings which can be operated on with aggregate functions.
The GROUP BY
column name is window
. It is of type STRUCT<start TIMESTAMP, end TIMESTAMP>
slide
must be less than or equal to width
.
start
must be less than slide
.
If slide
< width
the rows in each groups overlap.
By default slide
equals width
so expr
are partitioned into groups.
The windowing starts at 1970-01-01 00:00:00 UTC + start
. The default for start
is '0 SECONDS'
’
Examples
> SELECT window, min(val), max(val), count(val)
FROM VALUES (TIMESTAMP'2020-08-01 12:20:21', 17),
(TIMESTAMP'2020-08-01 12:20:22', 12),
(TIMESTAMP'2020-08-01 12:23:10', 8),
(TIMESTAMP'2020-08-01 12:25:05', 11),
(TIMESTAMP'2020-08-01 12:28:59', 15),
(TIMESTAMP'2020-08-01 12:30:01', 23),
(TIMESTAMP'2020-08-01 12:30:15', 2),
(TIMESTAMP'2020-08-01 12:35:22', 16) AS S(stamp, val)
GROUP BY window(stamp, '2 MINUTES 30 SECONDS', '30 SECONDS', '15 SECONDS');
{2020-08-01 12:19:15, 2020-08-01 12:21:45} 12 17 2
{2020-08-01 12:18:15, 2020-08-01 12:20:45} 12 17 2
{2020-08-01 12:20:15, 2020-08-01 12:22:45} 12 17 2
{2020-08-01 12:19:45, 2020-08-01 12:22:15} 12 17 2
{2020-08-01 12:18:45, 2020-08-01 12:21:15} 12 17 2
{2020-08-01 12:21:45, 2020-08-01 12:24:15} 8 8 1
{2020-08-01 12:22:45, 2020-08-01 12:25:15} 8 11 2
{2020-08-01 12:21:15, 2020-08-01 12:23:45} 8 8 1
{2020-08-01 12:22:15, 2020-08-01 12:24:45} 8 8 1
{2020-08-01 12:20:45, 2020-08-01 12:23:15} 8 8 1
{2020-08-01 12:23:45, 2020-08-01 12:26:15} 11 11 1
{2020-08-01 12:23:15, 2020-08-01 12:25:45} 11 11 1
{2020-08-01 12:24:45, 2020-08-01 12:27:15} 11 11 1
{2020-08-01 12:24:15, 2020-08-01 12:26:45} 11 11 1
{2020-08-01 12:27:15, 2020-08-01 12:29:45} 15 15 1
{2020-08-01 12:27:45, 2020-08-01 12:30:15} 15 23 2
{2020-08-01 12:28:45, 2020-08-01 12:31:15} 2 23 3
{2020-08-01 12:26:45, 2020-08-01 12:29:15} 15 15 1
{2020-08-01 12:28:15, 2020-08-01 12:30:45} 2 23 3
{2020-08-01 12:29:45, 2020-08-01 12:32:15} 2 23 2