refresh_continuous_aggregate.mdx

title	refresh_continuous_aggregate()
description	Manually refresh a continuous aggregate
keywords	continuous aggregates refresh

import * as C from "@constants";

import SinceRelease from '@components/SinceRelease.astro';

import { Callout } from '@stainless-api/docs/components';

import ReturnsVoid from '@partials/_returns-void.mdx';

Refresh all buckets of a {C.CAGG} in the refresh window given by window_start and window_end.

A {C.CAGG} materializes aggregates in time buckets. For example, min, max, average over 1 day worth of data, and is determined by the time_bucket interval. Therefore, when refreshing the {C.CAGG}, only buckets that completely fit within the refresh window are refreshed. In other words, it is not possible to compute the aggregate over, for an incomplete bucket. Therefore, any buckets that do not fit within the given refresh window are excluded.

The function expects the window parameter values to have a time type that is compatible with the {C.CAGG}'s time bucket expression, for example, if the time bucket is specified in TIMESTAMP WITH TIME ZONE, then the start and end time should be a date or timestamp type. Note that a {C.CAGG} using the TIMESTAMP WITH TIME ZONE type aligns with the UTC time zone, so, if window_start and window_end is specified in the local time zone, any time zone shift relative UTC needs to be accounted for when refreshing to align with bucket boundaries.

To improve performance for {C.CAGG} refresh, see CREATE MATERIALIZED VIEW.

Samples

Refresh the {C.CAGG} conditions between 2020-01-01 and 2020-02-01 exclusive. This call uses the default settings shown in Options. Since {C.TIMESCALE_DB} 2.28.0, those defaults perform an incremental refresh in batches of 10 buckets (buckets_per_batch):

CALL refresh_continuous_aggregate('conditions', '2020-01-01', '2020-02-01');

Incrementally refresh the {C.CAGG} conditions between 2020-01-01 and 2020-02-01 exclusive, in batches of 5 buckets at a time, from newest data to oldest. This breaks a large refresh into smaller transactions, so locks are held for shorter periods and results become visible as each batch completes:

CALL refresh_continuous_aggregate(
  'conditions',
  '2020-01-01',
  '2020-02-01',
  options => '{"buckets_per_batch": 5, "refresh_newest_first": true}'::jsonb
);

Disable incremental refresh and refresh the entire window in a single atomic pass by setting buckets_per_batch to 0:

CALL refresh_continuous_aggregate(
  'conditions',
  '2020-01-01',
  '2020-02-01',
  options => '{"buckets_per_batch": 0}'::jsonb
);

Force the conditions {C.CAGG} to refresh between 2020-01-01 and 2020-02-01 exclusive, even if the data has already been refreshed.

CALL refresh_continuous_aggregate('conditions', '2020-01-01', '2020-02-01', force => TRUE);

Arguments

The syntax is:

CALL refresh_continuous_aggregate(
    continuous_aggregate = '<view_name>',
    window_start = <interval>,
    window_end = <interval>,
    force = true | false,
    options = '<jsonb_options>'
);

Name	Type	Default	Required	Description
continuous_aggregate	REGCLASS	-	✔	The {C.CAGG} to refresh.
window_start	INTERVAL, TIMESTAMPTZ, INTEGER	-	✔	Start of the window to refresh, has to be before window_end.
window_end	INTERVAL, TIMESTAMPTZ, INTEGER	-	✔	End of the window to refresh, has to be after window_start.
force	BOOLEAN	FALSE	-	Force refresh every bucket in the time range between window_start and window_end, even when the bucket has already been refreshed. This can be very expensive when a lot of data is refreshed.
options	JSONB	NULL	-	JSONB object with additional options. See Options.

You must specify the window_start and window_end parameters differently, depending on the type of the time column of the {C.HYPERTABLE}. For {C.HYPERTABLE}s with TIMESTAMP, TIMESTAMPTZ, and DATE time columns, set the refresh window as an INTERVAL type. For {C.HYPERTABLE}s with integer-based timestamps, set the refresh window as an INTEGER type.

A `NULL` value for `window_start` is equivalent to the lowest changed element in the raw hypertable of the CAgg. A `NULL` value for `window_end` is equivalent to the largest changed element in raw hypertable of the CAgg. As changed element tracking is performed after the initial CAgg refresh, running CAgg refresh without `window_start` and `window_end` covers the entire time range. Note that it's not guaranteed that all buckets will be updated: refreshes will not take place when buckets are materialized with no data changes or with changes that only occurred in the secondary table used in the JOIN.

Options

The options argument is a JSONB object that accepts the following keys:

Key	Type	Default	Description
buckets_per_batch	INTEGER	10	Number of buckets to refresh per batch. This value is multiplied by the {C.CAGG} bucket width to determine the size of each batch range. Set to 0 to refresh the whole window in a single atomic pass. Values less than 0 are not allowed.
max_batches_per_execution	INTEGER	0	Maximum number of batches to process in this call. Default 0 means no limit. Values less than 0 are not allowed.
refresh_newest_first	BOOLEAN	true	Control the order of incremental refreshes. Set to true to refresh from the newest data to the oldest, or false for oldest to newest.

Since {C.TIMESCALE_DB} 2.28.0, `refresh_continuous_aggregate()` processes the refresh window incrementally in batches by default (`buckets_per_batch` is `10`), matching the behavior of {C.CAGG} refresh policies. Each batch runs in its own transaction, so locks are released between batches and results become visible as each batch completes. To refresh the entire window in a single atomic pass instead, set `buckets_per_batch` to `0`.

Returns