mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-12-02 04:22:03 +00:00
dea1706d4e
* Fix GROUP BY AggregateFunction
finalizeChunk() was unconditionally converting AggregateFunction to the
underlying type, however this should be done only if the aggregate was
applied.
So pass names of aggregates as an argument to the finalizeChunk()
Fuzzer report [1]:
Logical error: 'Bad cast from type DB::ColumnArray to DB::ColumnAggregateFunction'. Received signal 6 Received signal Aborted (6)
For the following query:
SELECT
arraySort(groupArrayArray(grp_simple)),
grp_aggreg,
arraySort(groupArrayArray(grp_simple)),
b,
arraySort(groupArrayArray(grp_simple)) AS grs
FROM data_02294
GROUP BY
a,
grp_aggreg,
b
SETTINGS optimize_aggregation_in_order = 1
[1]: https://s3.amazonaws.com/clickhouse-test-reports/37050/323ae98202d80fc4b311be1e7308ef2ac39e6063/fuzzer_astfuzzerdebug,actions//fuzzer.log
v2: fix conflicts in src/Interpreters/InterpreterSelectQuery.cpp
v3: Fix header for GROUP BY AggregateFunction WITH TOTALS
v4: Add sanity check into finalizeBlock()
v5: Use typeid_cast<&> to get more sensible error in case of bad cast (as suggested by @nickitat)
v6: Fix positions passed to finalizeChunk()
Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
* Core/ColumnNumbers.h: remove unused <string>
Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
* Optimize finalizeChunk()/finalizeBlock()
v2: s/ByPosition/Mask/ s/by_position/mask/
Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
25 lines
1.5 KiB
SQL
25 lines
1.5 KiB
SQL
drop table if exists data_02295;
|
|
|
|
create table data_02295 (
|
|
-- the order of "a" and "b" is important here
|
|
-- (since finalizeChunk() accepts positions and they may be wrong)
|
|
b Int64,
|
|
a Int64,
|
|
grp_aggreg AggregateFunction(groupArrayArray, Array(UInt64))
|
|
) engine = MergeTree() order by a;
|
|
insert into data_02295 select 0 b, intDiv(number, 2) a, groupArrayArrayState([toUInt64(number)]) from numbers(4) group by a, b;
|
|
|
|
-- { echoOn }
|
|
SELECT grp_aggreg FROM data_02295 GROUP BY a, grp_aggreg SETTINGS optimize_aggregation_in_order = 0 FORMAT JSONEachRow;
|
|
SELECT grp_aggreg FROM data_02295 GROUP BY a, grp_aggreg SETTINGS optimize_aggregation_in_order = 1 FORMAT JSONEachRow;
|
|
SELECT grp_aggreg FROM data_02295 GROUP BY a, grp_aggreg WITH TOTALS SETTINGS optimize_aggregation_in_order = 0 FORMAT JSONEachRow;
|
|
SELECT grp_aggreg FROM data_02295 GROUP BY a, grp_aggreg WITH TOTALS SETTINGS optimize_aggregation_in_order = 1 FORMAT JSONEachRow;
|
|
-- regression for incorrect positions passed to finalizeChunk()
|
|
SELECT a, min(b), max(b) FROM data_02295 GROUP BY a ORDER BY a, count() SETTINGS optimize_aggregation_in_order = 1;
|
|
SELECT a, min(b), max(b) FROM data_02295 GROUP BY a ORDER BY a, count() SETTINGS optimize_aggregation_in_order = 1, max_threads = 1;
|
|
SELECT a, min(b), max(b) FROM data_02295 GROUP BY a WITH TOTALS ORDER BY a, count() SETTINGS optimize_aggregation_in_order = 1;
|
|
SELECT a, min(b), max(b) FROM data_02295 GROUP BY a WITH TOTALS ORDER BY a, count() SETTINGS optimize_aggregation_in_order = 1, max_threads = 1;
|
|
-- { echoOff }
|
|
|
|
drop table data_02295;
|