duerwuyi opened a new issue, #37852:
URL: https://github.com/apache/shardingsphere/issues/37852
## Bug Report
### Which version of ShardingSphere did you use?
shardingsphere-proxy 5.5.2
postgres 17.6 as backend
### Expected behavior
This query should return distinct result:
```sql
select distinct
pg_catalog.round(case when (null::bool) then (ref_0.c38) else (ref_0.c37)
end
) as c_0
from
t24 as ref_0
where true::bool;
```
### Actual behavior
SELECT DISTINCT on a sharded table returns duplicate rows when executed
through ShardingSphere-Proxy. The same logical query executed on a single
PostgreSQL instance returns correct globally de-duplicated results.
From proxy logs, ShardingSphere rewrites the query into multiple per-shard
SELECT DISTINCT ... FROM t24_x combined by UNION ALL. This does not perform
global deduplication across shards, so duplicate values (including duplicate
NULLs) are returned.
This is a correctness bug: DISTINCT is a global set operation and cannot be
implemented as per-shard DISTINCT + UNION ALL without an additional global
distinct/aggregation step.
```sql
testdb=> select distinct
pg_catalog.round(case when (null::bool) then (ref_0.c38) else (ref_0.c37)
end
) as c_0
from
t24 as ref_0
where true::bool;
c_0
-----
0
-95
-94
(5 rows)
```
### Reason analyze (If you can)
log:
```log
[INFO ] 2026-01-26 11:25:33.802 [ShardingSphere-Command-456]
ShardingSphere-SQL - Actual SQL: ds_1 ::: select distinct
pg_catalog.round(case when (null::bool) then (ref_0.c38) else (ref_0.c37)
end
) as c_0
from
t24_1 as ref_0
where true::bool UNION ALL select distinct
pg_catalog.round(case when (null::bool) then (ref_0.c38) else (ref_0.c37)
end
) as c_0
from
t24_6 as ref_0
where true::bool UNION ALL select distinct
pg_catalog.round(case when (null::bool) then (ref_0.c38) else (ref_0.c37)
end
) as c_0
from
t24_11 as ref_0
where true::bool UNION ALL select distinct
pg_catalog.round(case when (null::bool) then (ref_0.c38) else (ref_0.c37)
end
) as c_0
from
t24_16 as ref_0
where true::bool
[INFO ] 2026-01-26 11:25:33.802 [ShardingSphere-Command-456]
ShardingSphere-SQL - Actual SQL: ds_2 ::: select distinct
pg_catalog.round(case when (null::bool) then (ref_0.c38) else (ref_0.c37)
end
) as c_0
from
t24_2 as ref_0
where true::bool UNION ALL select distinct
pg_catalog.round(case when (null::bool) then (ref_0.c38) else (ref_0.c37)
end
) as c_0
from
t24_7 as ref_0
where true::bool UNION ALL select distinct
pg_catalog.round(case when (null::bool) then (ref_0.c38) else (ref_0.c37)
end
) as c_0
from
t24_12 as ref_0
where true::bool UNION ALL select distinct
```
### Steps to reproduce the behavior, such as: SQL to execute, sharding rule
configuration, when exception occur etc.
[ss_postgres.zip](https://github.com/user-attachments/files/24228864/ss_postgres.zip)
run `docker compose up -d` in this file.
SQL to execute first:
```sql
CREATE SHARDING TABLE RULE t24 (
STORAGE_UNITS(ds_0,ds_1,ds_2,ds_3,ds_4),
SHARDING_COLUMN=colocated_key,TYPE(NAME="MOD",PROPERTIES(
"sharding-count"="21"))
);
DROP TABLE IF EXISTS t24;
create table t24 (
vkey int4 ,
pkey int4 ,
c36 numeric ,
c37 numeric ,
c38 numeric ,
c39 text ,
c40 text ,
c41 text ,
colocated_key int4
);
insert into t24 (vkey, pkey, c36, c37, c38, c39, c40, c41, colocated_key)
values
(3735, 3745000, -53.90, null::numeric, -56.84, null::text,
'RrHYP~C1vC^)PNa(DjV', null::text, 1826460456);
insert into t24 (vkey, pkey, c36, c37, c38, c39, c40, c41, colocated_key)
values
(3742, 3752000, -1456924258.14691, null::numeric, -29.97,
'*E]74/q<cIcNL7h#-l.eXx4YB', 'bEg^y;640^<', '4nT);kbc.Dz1<QvyN4mXWU',
451348812);
insert into t24 (vkey, pkey, c36, c37, c38, c39, c40, c41, colocated_key)
values
(3745, 3755000, null::numeric, -93.93, 58.15, null::text,
',AZS=$?X@mcMVWFu5~I`', '--M.OGWU:M7U;n:', -1539890837);
insert into t24 (vkey, pkey, c36, c37, c38, c39, c40, c41, colocated_key)
values
(3749, 3759000, -424786407.14788, -94.75, null::numeric, '*9E8bE',
't`{6}@WtG&A{bOFOc(5<;B', '+JARJ!#xMi8>#`2J', -65);
insert into t24 (vkey, pkey, c36, c37, c38, c39, c40, c41, colocated_key)
values
(3750, 3760000, -2114913156.2077, null::numeric, 1968229656.11327,
'Fq&QteE,#9GEj', 'H#Cngf', '=3VB,Q]qM00,rX=9;xZUOj ;H4oZU', 1652888866);
insert into t24 (vkey, pkey, c36, c37, c38, c39, c40, c41, colocated_key)
values
(3755, 3765000, null::numeric, 0.0, 58.74, 'GA4Ojj$EhN26',
'?Whaf@~(w}Y]Wld<MNm,8', '9:JRiOK&*|#:H!f-=RfY9C4wB`', 1144145842);
insert into t24 (vkey, pkey, c36, c37, c38, c39, c40, c41, colocated_key)
values
(3758, 3768000, 1136604594.30098, 0.0, null::numeric, '^:=akh+I',
null::text, '', 93677795);
insert into t24 (vkey, pkey, c36, c37, c38, c39, c40, c41, colocated_key)
values
(3765, 3775000, -705783199.6173, null::numeric, -1679695831.13212,
'&M-]P1r$#', 'LSC :Fi4W-jbu`', '*T ;5&qFEF5^;>>:/?qJW', 0);
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail:
[email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
