mirror of
https://github.com/GreptimeTeam/greptimedb.git
synced 2026-09-07 22:18:57 +00:00
dennis zhuang
bb9b7e8778
* fix: re-scan stream-backed tables in recursive CTEs A recursive CTE re-executes its recursive term on every iteration, but DfTableProviderAdapter hands StreamScanAdapter a single-use stream built at planning time. The second iteration failed with "Stream already exhausted" for every table served through DataSource::get_stream — information_schema, pg_catalog, the computed entity-graph tables and numbers. Keep that stream for the first execution and open a new one over the same scan request for later executions. Closes #9037 Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * refactor: drop redundant binding in stream factory Signed-off-by: Dennis Zhuang <killme2008@gmail.com> --------- Signed-off-by: Dennis Zhuang <killme2008@gmail.com>
93 lines
2.7 KiB
SQL
93 lines
2.7 KiB
SQL
create table a(i integer, ts TIMESTAMP TIME INDEX);
|
|
|
|
insert into a values (42, 1);
|
|
|
|
with cte1 as (Select i as j from a) select * from cte1;
|
|
|
|
with cte1 as (Select i as j from a) select x from cte1 t1(x);
|
|
|
|
with cte1(xxx) as (Select i as j from a) select xxx from cte1;
|
|
|
|
with cte1(xxx) as (Select i as j from a) select x from cte1 t1(x);
|
|
|
|
with cte1 as (Select i as j from a), cte2 as (select ref.j as k from cte1 as ref), cte3 as (select ref2.j+1 as i from cte1 as ref2) select * from cte2 , cte3;
|
|
|
|
-- SQLNESS SORT_RESULT 3 1
|
|
with cte1 as (select i as j from a), cte2 as (select ref.j as k from cte1 as ref), cte3 as (select ref2.j+1 as i from cte1 as ref2) select * from cte2 union all select * FROM cte3 order by 1;
|
|
|
|
with cte1 as (select 42), cte1 as (select 42) select * FROM cte1;
|
|
|
|
-- reference to CTE before its actually defined, it's not supported by datafusion
|
|
with cte3 as (select ref2.j as i from cte1 as ref2), cte1 as (Select i as j from a), cte2 as (select ref.j+1 as k from cte1 as ref) select * from cte2 union all select * FROM cte3;
|
|
|
|
with cte1 as (Select i as j from a) select * from cte1 cte11, cte1 cte12;
|
|
|
|
with cte1 as (Select i as j from a) select * from cte1 where j = (select max(j) from cte1 as cte2);
|
|
|
|
with cte1(x, y) as (select 42 a, 84 b) select zzz, y from cte1 t1(zzz, y);
|
|
|
|
SELECT 1 UNION ALL (WITH cte AS (SELECT 42) SELECT * FROM cte) order by 1;
|
|
|
|
WITH RECURSIVE cte(d) AS (
|
|
SELECT 1
|
|
UNION ALL
|
|
(WITH c(d) AS (SELECT * FROM cte)
|
|
SELECT d + 1
|
|
FROM c
|
|
WHERE FALSE
|
|
)
|
|
)
|
|
SELECT max(d) FROM cte;
|
|
|
|
-- the recursive term is re-executed once per iteration, so its scans must be re-runnable
|
|
WITH RECURSIVE t AS (
|
|
SELECT 0 AS depth
|
|
UNION ALL
|
|
SELECT t.depth + 1 AS depth
|
|
FROM t, information_schema.tables
|
|
WHERE table_catalog = 'greptime'
|
|
AND table_schema = 'information_schema'
|
|
AND table_name = 'tables'
|
|
AND t.depth < 2
|
|
)
|
|
SELECT depth FROM t ORDER BY depth;
|
|
|
|
-- Nested aliases is not supported in datafusion
|
|
with cte (a) as (
|
|
select 1
|
|
)
|
|
select
|
|
a as alias1,
|
|
alias1 as alias2
|
|
from cte
|
|
where alias2 > 0;
|
|
|
|
drop table a;
|
|
|
|
|
|
CREATE TABLE grpc_latencies
|
|
(
|
|
ts TIMESTAMP TIME INDEX,
|
|
host VARCHAR(255),
|
|
latency FLOAT,
|
|
PRIMARY KEY (host),
|
|
);
|
|
|
|
INSERT INTO grpc_latencies
|
|
VALUES ('2023-10-01 10:00:00', 'host1', 120),
|
|
('2023-10-01 10:00:00', 'host2', 150),
|
|
('2023-10-01 10:00:05', 'host1', 130);
|
|
|
|
|
|
WITH latencies AS (SELECT ts,
|
|
host,
|
|
AVG(latency) RANGE '2s' AS avg_latency
|
|
FROM grpc_latencies ALIGN '2s' BY (host) FILL PREV
|
|
)
|
|
SELECT latencies.ts,
|
|
AVG(latencies.avg_latency)
|
|
FROM latencies
|
|
GROUP BY latencies.ts ORDER BY latencies.ts;
|
|
|
|
DROP TABLE grpc_latencies;
|