Allow CAggs with variable sized bucked with origin/offset (#7005)

On 2.15.x we added complete support of CAggs using time bucket with
origin and/or offset, but we restrict the creating when the bucked size
is variable due to some uncertaing regarding monthly buckets.

When bucketing by month we always align with the beginning of the month
even defining an origin with day component. So to be consistent with the
current implementation we'll not change this behavior and allow it to be
used in Continuous Aggregates.

Disable-check: force-changelog-file

(cherry picked from commit af8ca2d)

tsl/src/continuous_aggs/common.c

@@ -967,47 +967,13 @@ cagg_validate_query(const Query *query, const bool finalized, const char *cagg_s
 				(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
 				 errmsg("cannot create continuous aggregate on hypertable with row security")));
 
-	/* Test for broken time_bucket configurations (variable with with offset and origin). We need to
-	 * check only time based buckets since integer based buckets are always fixed. */
-	bool time_offset_or_origin_set =
-		(bucket_info.bf->bucket_time_offset != NULL) ||
-		(TIMESTAMP_NOT_FINITE(bucket_info.bf->bucket_time_origin) == false);
-
 	/* At this point, we should have a valid bucket function. Otherwise, we have errored out before.
 	 */
 	Ensure(OidIsValid(bucket_info.bf->bucket_function), "unable to find valid bucket function");
 
 	/* Ignore time_bucket_ng in this check, since offset and origin were allowed in the past */
 	FuncInfo *func_info = ts_func_cache_get_bucketing_func(bucket_info.bf->bucket_function);
 	Ensure(func_info != NULL, "bucket function is not found in function cache");
-	bool is_time_bucket_ng = func_info->origin == ORIGIN_TIMESCALE_EXPERIMENTAL;
-
-	/*
-	 * Some time_bucket variants using variable-sized buckets and custom origin/offset values are
-	 * not behaving correctly. To prevent misaligned buckets, these variants are blocked at the
-	 * moment. This restriction can be removed as soon as time_bucket behaves correctly.
-	 *
-	 * 		--- Align with default origin ('midnight on January 1, 2000')
-	 * 		test2=# SELECT time_bucket('1 month', '2000-01-01 01:05:00 UTC'::timestamptz,
-	 *         timezone=>'UTC'); time_bucket
-	 *		------------------------
-	 *		2000-01-01 00:00:00+00
-	 *
-	 *		--- Using a custom origin
-	 *		test2=# SELECT time_bucket('1 month', '2000-01-01 01:05:00 UTC'::timestamptz,
-	 *         origin=>'2000-01-01 01:05:00 UTC'::timestamptz, timezone=>'UTC'); time_bucket
-	 *		------------------------
-	 *		2000-01-01 00:00:00+00              <--- Should be 2000-01-01 01:05:00+00
-	 *		(1 row)
-	 */
-	if (time_bucket_info_has_fixed_width(&bucket_info) == false && time_offset_or_origin_set &&
-		!is_time_bucket_ng)
-	{
-		ereport(ERROR,
-				(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-				 errmsg("cannot create continuous aggregate with variable-width bucket using "
-						"offset or origin.")));
-	}
 
 	/* hierarchical cagg validations */
 	if (is_hierarchical)

tsl/test/expected/cagg_ddl-13.out

@@ -1811,36 +1811,29 @@ SELECT * FROM cashflows;
 -- 3. test named ts
 -- 4. test named bucket width
 -- named origin
--- Currently not supported due to a bug in time_bucket (see comment in cagg_validate_query)
-\set ON_ERROR_STOP 0
 CREATE MATERIALIZED VIEW cagg_named_origin WITH
 (timescaledb.continuous, timescaledb.materialized_only=false) AS
 SELECT time_bucket('1h', time, 'UTC', origin => '2001-01-03 01:23:45') AS bucket,
 avg(amount) as avg_amount
 FROM transactions GROUP BY 1 WITH NO DATA;
-ERROR:  cannot create continuous aggregate with variable-width bucket using offset or origin.
 -- named timezone
 CREATE MATERIALIZED VIEW cagg_named_tz_origin WITH
 (timescaledb.continuous, timescaledb.materialized_only=false) AS
 SELECT time_bucket('1h', time, timezone => 'UTC', origin => '2001-01-03 01:23:45') AS bucket,
 avg(amount) as avg_amount
 FROM transactions GROUP BY 1 WITH NO DATA;
-ERROR:  cannot create continuous aggregate with variable-width bucket using offset or origin.
 -- named ts
 CREATE MATERIALIZED VIEW cagg_named_ts_tz_origin WITH
 (timescaledb.continuous, timescaledb.materialized_only=false) AS
 SELECT time_bucket('1h', ts => time, timezone => 'UTC', origin => '2001-01-03 01:23:45') AS bucket,
 avg(amount) as avg_amount
 FROM transactions GROUP BY 1 WITH NO DATA;
-ERROR:  cannot create continuous aggregate with variable-width bucket using offset or origin.
 -- named bucket width
 CREATE MATERIALIZED VIEW cagg_named_all WITH
 (timescaledb.continuous, timescaledb.materialized_only=false) AS
 SELECT time_bucket(bucket_width => '1h', ts => time, timezone => 'UTC', origin => '2001-01-03 01:23:45') AS bucket,
 avg(amount) as avg_amount
 FROM transactions GROUP BY 1 WITH NO DATA;
-ERROR:  cannot create continuous aggregate with variable-width bucket using offset or origin.
-\set ON_ERROR_STOP 1
 -- Refreshing from the beginning (NULL) of a CAGG with variable time bucket and
 -- using an INTERVAL for the end timestamp (issue #5534)
 CREATE MATERIALIZED VIEW transactions_montly
@@ -1923,7 +1916,7 @@ CREATE TABLE conditions (
 SELECT create_hypertable('conditions', 'time');
     create_hypertable     
 --------------------------
- (48,public,conditions,t)
+ (52,public,conditions,t)
 (1 row)
 
 INSERT INTO conditions VALUES ( '2018-01-01 09:20:00-08', 'SFO', 55);
@@ -1953,10 +1946,10 @@ WITH NO DATA;
  bucket   | timestamp with time zone |           |          |         | plain    | 
  avg      | double precision         |           |          |         | plain    | 
 View definition:
- SELECT _materialized_hypertable_49.location,
-    _materialized_hypertable_49.bucket,
-    _materialized_hypertable_49.avg
-   FROM _timescaledb_internal._materialized_hypertable_49;
+ SELECT _materialized_hypertable_53.location,
+    _materialized_hypertable_53.bucket,
+    _materialized_hypertable_53.avg
+   FROM _timescaledb_internal._materialized_hypertable_53;
 
 -- Should return NO ROWS
 SELECT * FROM conditions_daily ORDER BY bucket, avg;
@@ -1973,17 +1966,17 @@ ALTER MATERIALIZED VIEW conditions_daily SET (timescaledb.materialized_only=fals
  bucket   | timestamp with time zone |           |          |         | plain    | 
  avg      | double precision         |           |          |         | plain    | 
 View definition:
- SELECT _materialized_hypertable_49.location,
-    _materialized_hypertable_49.bucket,
-    _materialized_hypertable_49.avg
-   FROM _timescaledb_internal._materialized_hypertable_49
-  WHERE _materialized_hypertable_49.bucket < COALESCE(_timescaledb_functions.to_timestamp(_timescaledb_functions.cagg_watermark(49)), '-infinity'::timestamp with time zone)
+ SELECT _materialized_hypertable_53.location,
+    _materialized_hypertable_53.bucket,
+    _materialized_hypertable_53.avg
+   FROM _timescaledb_internal._materialized_hypertable_53
+  WHERE _materialized_hypertable_53.bucket < COALESCE(_timescaledb_functions.to_timestamp(_timescaledb_functions.cagg_watermark(53)), '-infinity'::timestamp with time zone)
 UNION ALL
  SELECT conditions.location,
     time_bucket('@ 1 day'::interval, conditions."time") AS bucket,
     avg(conditions.temperature) AS avg
    FROM conditions
-  WHERE conditions."time" >= COALESCE(_timescaledb_functions.to_timestamp(_timescaledb_functions.cagg_watermark(49)), '-infinity'::timestamp with time zone)
+  WHERE conditions."time" >= COALESCE(_timescaledb_functions.to_timestamp(_timescaledb_functions.cagg_watermark(53)), '-infinity'::timestamp with time zone)
   GROUP BY conditions.location, (time_bucket('@ 1 day'::interval, conditions."time"));
 
 -- Should return ROWS because now it is realtime
@@ -2008,10 +2001,10 @@ ALTER MATERIALIZED VIEW conditions_daily SET (timescaledb.materialized_only=true
  bucket   | timestamp with time zone |           |          |         | plain    | 
  avg      | double precision         |           |          |         | plain    | 
 View definition:
- SELECT _materialized_hypertable_49.location,
-    _materialized_hypertable_49.bucket,
-    _materialized_hypertable_49.avg
-   FROM _timescaledb_internal._materialized_hypertable_49;
+ SELECT _materialized_hypertable_53.location,
+    _materialized_hypertable_53.bucket,
+    _materialized_hypertable_53.avg
+   FROM _timescaledb_internal._materialized_hypertable_53;
 
 CALL refresh_continuous_aggregate('conditions_daily', NULL, NULL);
 SELECT * FROM conditions_daily ORDER BY bucket, avg;
@@ -2088,15 +2081,15 @@ NOTICE:  refreshing continuous aggregate "cagg1"
 ALTER MATERIALIZED VIEW cagg1 SET (timescaledb.compress);
 NOTICE:  defaulting compress_orderby to time_bucket
 WARNING:  there was some uncertainty picking the default segment by for the hypertable: You do not have any indexes on columns that can be used for segment_by and thus we are not using segment_by for compression. Please make sure you are not missing any indexes
-NOTICE:  default segment by for hypertable "_materialized_hypertable_52" is set to ""
+NOTICE:  default segment by for hypertable "_materialized_hypertable_56" is set to ""
 SELECT count(compress_chunk(ch)) FROM show_chunks('cagg1') ch;
  count 
 -------
      1
 (1 row)
 
 DROP MATERIALIZED VIEW cagg1;
-NOTICE:  drop cascades to table _timescaledb_internal._hyper_52_68_chunk
+NOTICE:  drop cascades to table _timescaledb_internal._hyper_56_68_chunk
 SELECT * FROM _timescaledb_catalog.compression_settings;
  relid | segmentby | orderby | orderby_desc | orderby_nullsfirst 
 -------+-----------+---------+--------------+--------------------

tsl/test/expected/cagg_ddl-14.out

@@ -1811,36 +1811,29 @@ SELECT * FROM cashflows;
 -- 3. test named ts
 -- 4. test named bucket width
 -- named origin
--- Currently not supported due to a bug in time_bucket (see comment in cagg_validate_query)
-\set ON_ERROR_STOP 0
 CREATE MATERIALIZED VIEW cagg_named_origin WITH
 (timescaledb.continuous, timescaledb.materialized_only=false) AS
 SELECT time_bucket('1h', time, 'UTC', origin => '2001-01-03 01:23:45') AS bucket,
 avg(amount) as avg_amount
 FROM transactions GROUP BY 1 WITH NO DATA;
-ERROR:  cannot create continuous aggregate with variable-width bucket using offset or origin.
 -- named timezone
 CREATE MATERIALIZED VIEW cagg_named_tz_origin WITH
 (timescaledb.continuous, timescaledb.materialized_only=false) AS
 SELECT time_bucket('1h', time, timezone => 'UTC', origin => '2001-01-03 01:23:45') AS bucket,
 avg(amount) as avg_amount
 FROM transactions GROUP BY 1 WITH NO DATA;
-ERROR:  cannot create continuous aggregate with variable-width bucket using offset or origin.
 -- named ts
 CREATE MATERIALIZED VIEW cagg_named_ts_tz_origin WITH
 (timescaledb.continuous, timescaledb.materialized_only=false) AS
 SELECT time_bucket('1h', ts => time, timezone => 'UTC', origin => '2001-01-03 01:23:45') AS bucket,
 avg(amount) as avg_amount
 FROM transactions GROUP BY 1 WITH NO DATA;
-ERROR:  cannot create continuous aggregate with variable-width bucket using offset or origin.
 -- named bucket width
 CREATE MATERIALIZED VIEW cagg_named_all WITH
 (timescaledb.continuous, timescaledb.materialized_only=false) AS
 SELECT time_bucket(bucket_width => '1h', ts => time, timezone => 'UTC', origin => '2001-01-03 01:23:45') AS bucket,
 avg(amount) as avg_amount
 FROM transactions GROUP BY 1 WITH NO DATA;
-ERROR:  cannot create continuous aggregate with variable-width bucket using offset or origin.
-\set ON_ERROR_STOP 1
 -- Refreshing from the beginning (NULL) of a CAGG with variable time bucket and
 -- using an INTERVAL for the end timestamp (issue #5534)
 CREATE MATERIALIZED VIEW transactions_montly
@@ -1923,7 +1916,7 @@ CREATE TABLE conditions (
 SELECT create_hypertable('conditions', 'time');
     create_hypertable     
 --------------------------
- (48,public,conditions,t)
+ (52,public,conditions,t)
 (1 row)
 
 INSERT INTO conditions VALUES ( '2018-01-01 09:20:00-08', 'SFO', 55);
@@ -1953,10 +1946,10 @@ WITH NO DATA;
  bucket   | timestamp with time zone |           |          |         | plain    | 
  avg      | double precision         |           |          |         | plain    | 
 View definition:
- SELECT _materialized_hypertable_49.location,
-    _materialized_hypertable_49.bucket,
-    _materialized_hypertable_49.avg
-   FROM _timescaledb_internal._materialized_hypertable_49;
+ SELECT _materialized_hypertable_53.location,
+    _materialized_hypertable_53.bucket,
+    _materialized_hypertable_53.avg
+   FROM _timescaledb_internal._materialized_hypertable_53;
 
 -- Should return NO ROWS
 SELECT * FROM conditions_daily ORDER BY bucket, avg;
@@ -1973,17 +1966,17 @@ ALTER MATERIALIZED VIEW conditions_daily SET (timescaledb.materialized_only=fals
  bucket   | timestamp with time zone |           |          |         | plain    | 
  avg      | double precision         |           |          |         | plain    | 
 View definition:
- SELECT _materialized_hypertable_49.location,
-    _materialized_hypertable_49.bucket,
-    _materialized_hypertable_49.avg
-   FROM _timescaledb_internal._materialized_hypertable_49
-  WHERE _materialized_hypertable_49.bucket < COALESCE(_timescaledb_functions.to_timestamp(_timescaledb_functions.cagg_watermark(49)), '-infinity'::timestamp with time zone)
+ SELECT _materialized_hypertable_53.location,
+    _materialized_hypertable_53.bucket,
+    _materialized_hypertable_53.avg
+   FROM _timescaledb_internal._materialized_hypertable_53
+  WHERE _materialized_hypertable_53.bucket < COALESCE(_timescaledb_functions.to_timestamp(_timescaledb_functions.cagg_watermark(53)), '-infinity'::timestamp with time zone)
 UNION ALL
  SELECT conditions.location,
     time_bucket('@ 1 day'::interval, conditions."time") AS bucket,
     avg(conditions.temperature) AS avg
    FROM conditions
-  WHERE conditions."time" >= COALESCE(_timescaledb_functions.to_timestamp(_timescaledb_functions.cagg_watermark(49)), '-infinity'::timestamp with time zone)
+  WHERE conditions."time" >= COALESCE(_timescaledb_functions.to_timestamp(_timescaledb_functions.cagg_watermark(53)), '-infinity'::timestamp with time zone)
   GROUP BY conditions.location, (time_bucket('@ 1 day'::interval, conditions."time"));
 
 -- Should return ROWS because now it is realtime
@@ -2008,10 +2001,10 @@ ALTER MATERIALIZED VIEW conditions_daily SET (timescaledb.materialized_only=true
  bucket   | timestamp with time zone |           |          |         | plain    | 
  avg      | double precision         |           |          |         | plain    | 
 View definition:
- SELECT _materialized_hypertable_49.location,
-    _materialized_hypertable_49.bucket,
-    _materialized_hypertable_49.avg
-   FROM _timescaledb_internal._materialized_hypertable_49;
+ SELECT _materialized_hypertable_53.location,
+    _materialized_hypertable_53.bucket,
+    _materialized_hypertable_53.avg
+   FROM _timescaledb_internal._materialized_hypertable_53;
 
 CALL refresh_continuous_aggregate('conditions_daily', NULL, NULL);
 SELECT * FROM conditions_daily ORDER BY bucket, avg;
@@ -2088,15 +2081,15 @@ NOTICE:  refreshing continuous aggregate "cagg1"
 ALTER MATERIALIZED VIEW cagg1 SET (timescaledb.compress);
 NOTICE:  defaulting compress_orderby to time_bucket
 WARNING:  there was some uncertainty picking the default segment by for the hypertable: You do not have any indexes on columns that can be used for segment_by and thus we are not using segment_by for compression. Please make sure you are not missing any indexes
-NOTICE:  default segment by for hypertable "_materialized_hypertable_52" is set to ""
+NOTICE:  default segment by for hypertable "_materialized_hypertable_56" is set to ""
 SELECT count(compress_chunk(ch)) FROM show_chunks('cagg1') ch;
  count 
 -------
      1
 (1 row)
 
 DROP MATERIALIZED VIEW cagg1;
-NOTICE:  drop cascades to table _timescaledb_internal._hyper_52_68_chunk
+NOTICE:  drop cascades to table _timescaledb_internal._hyper_56_68_chunk
 SELECT * FROM _timescaledb_catalog.compression_settings;
  relid | segmentby | orderby | orderby_desc | orderby_nullsfirst 
 -------+-----------+---------+--------------+--------------------