PostgreSQL parser fails on GROUP BY ROLLUP(a), b with schema-qualified columns

[ang0858]

Database Type

postgresql

Database Version

PostgreSQL 15.3

Druid Version

1.2.23

JDK Version

Open JDK 21

Error SQL

SELECT
'day' AS stat_period,
TO_CHAR(dis_date, 'YYYY-MM-DD') AS stat_date,
CASE
WHEN GROUPING(dis_dept_code) = 1 THEN 'all'
ELSE dis_dept_code
END AS dept_code,
COUNT() FILTER (WHERE if_surg_flag = 0) AS surg_dis_cnt,
COUNT() FILTER (WHERE if_surg_flag = 0 and nrs_grd in('01','04')) AS surg_dis_hl_nurs_cnt
FROM dwd_inp_fact_discharge_reg
WHERE dis_date >= to_date('${start_time}', 'YYYY-MM-DD')
and dis_date <= to_date('${end_time}', 'YYYY-MM-DD')
AND dis_dept_code IS NOT NULL
GROUP BY rollup(dis_dept_code), dis_date
ORDER BY stat_date, dept_code;

Testcase Code

List stmts = SQLUtils.parseStatements(sql, dbType);

Stacktrace Info

No response

Error Info

com.alibaba.druid.sql.parser.ParserException: not supported.pos 562, line 8, column 34, token ,
at com.alibaba.druid.sql.parser.SQLStatementParser.parseStatementList(SQLStatementParser.java:654)
at com.alibaba.druid.sql.SQLUtils.parseStatements(SQLUtils.java:600)
at com.alibaba.druid.sql.SQLUtils.parseStatements(SQLUtils.java:621)
将group by子句下的rollup(...)放最后可解决

[hiSandog]

A useful next step might be to separate the repro into input, state/cache update, and final output for PostgreSQL. That would make it clearer whether “PostgreSQL parser fails on GROUP BY ROLLUP(a), b with schema-qualified columns” is failing during parsing/configuration, during internal state updates, or only at the user-visible result.

[fudianchn]

已定位并修复，详见 PR #6676。

根因：SQLSelectParser#parseGroupBy 把 ROLLUP/CUBE 当作"包装"形式消费（设 withRollUp/paren 并吃掉 ROLLUP(）；accept(RPAREN) 之后只有启用了 RewriteGroupByCubeRollupToFunction 的方言（ODPS）才会继续消费尾部的 , b，PostgreSQL 等标准 SQL 方言不会，导致剩余 , b 无人消费而抛 ParserException: not supported token ,。

修复：对未启用 rewrite 的方言，当 ROLLUP(...)/CUBE(...) 后跟逗号时，将其重组为函数式分组项并继续解析剩余分组项。仅新增一个 else if 分支，只在原本直接报错的场景生效；GROUP BY ROLLUP(a, b) 等既有用法保持不变。

回归测试 Issue6662 已覆盖（含本 issue 的原始 SQL），跨方言 GROUP BY 用例（PG/MySQL/ODPS/Oracle/Hive/Snowflake/DB2 等）全绿。