MDEV-22269 GROUPING function by jaeheonshim · Pull Request #5314 · MariaDB/server

jaeheonshim · 2026-07-01T02:43:13Z

MDEV-22269: Add support for the GROUPING() function (SQL:2008 feature T431).

jaeheonshim · 2026-07-01T02:45:07Z

Note: GROUPING is currently broken and segfaults if nested inside a function (e.g.: select name, size, bin(grouping(name, size)), sum(quantity) from t group by name, size with rollup;). This is because in JOIN::rollup_init, we only check the outer-level select list Items, so Item_func_grouping::setup_rollup doesn't get called on any nested GROUPING functions. Working on a fix.

gemini-code-assist

Code Review

This pull request implements the SQL GROUPING() function, introducing the Item_func_grouping class, adding parser support for the GROUPING keyword, and integrating it with rollup processing. The feedback highlights critical safety improvements: initializing member variables in the Item_func_grouping constructor to prevent undefined behavior, limiting the number of arguments to 64 to avoid bit-shifting overflow, and adding a null-pointer check along with using bitwise OR operations in val_int() for safer bitmask construction.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

mariadb-RexJohnston · 2026-08-03T23:12:04Z

I created https://jira.mariadb.org/browse/MDEV-40611
It'll need fixing before you can uncomment this in your tests

# Test with view and GROUPING() in having clause
#CREATE VIEW v AS SELECT (SELECT  MAX(a) FROM t1) as field1 FROM t1
#GROUP BY field1 WITH ROLLUP HAVING GROUPING(field1)=0;
#SELECT * FROM v;
#DROP VIEW v;

gkodinov

Thanks for your work. This is a preliminary review.

Couple of formal things to fix:

rebase and resolve the merge conflicts please
squash the two separate commits into a single one (or state the reason for having tw o commits instead of one)
fix the P_S digest failures, e.g.

perfschema.start_server_low_digest_sql_length w17 [ fail ]
        Test ended at 2026-08-10 23:46:51
CURRENT_TEST: perfschema.start_server_low_digest_sql_length
--- /home/buildbot/amd64-msan-clang-20/build/mysql-test/suite/perfschema/r/start_server_low_digest_sql_length.result	2026-08-10 23:25:32.000000000 +0000
+++ /home/buildbot/amd64-msan-clang-20/build/mysql-test/suite/perfschema/r/start_server_low_digest_sql_length.reject	2026-08-10 23:46:51.017149385 +0000
@@ -8,5 +8,5 @@
 ####################################
 SELECT event_name, digest, digest_text, sql_text FROM events_statements_history_long;
 event_name	digest	digest_text	sql_text
-statement/sql/select	bc85fb4e8188bed1c842f764ec9fb06a	SELECT ? + ? + 	SELECT ...
-statement/sql/truncate	208418d9203df0b1b2cadb4d7196ea77	TRUNCATE TABLE 	truncat...
+statement/sql/select	2490fb0602b2ad29d2adc5923c6be917	SELECT ? + ? + 	SELECT ...
+statement/sql/truncate	8460b8b5a83d4357fcb94a1ccbb54817	TRUNCATE TABLE 	truncat...

gkodinov

Either fix everything that there is to fix in this one or keep it in "draft" until you do please.
There's still two commits, unresolved conflicts and spurious errors in buildbot.

jaeheonshim · 2026-08-14T19:41:59Z

Sorry about that, I've resolved the merge conflicts and fixed the digest failures. the reason I have two commits is purely for attribution purposes: the first commit contains only test cases pulled from mysql, and the second commit contains my actual contributions

gkodinov

I'd expand a little bit on the commit comments. Otherwise, LGTM.

Tests for the GROUPING() function taken from mysql-test/t/olap.test

SQL:2008 feature T431

mariadb-RexJohnston · 2026-08-24T21:51:25Z

+      empty grouping set. Rollup level is set to 0 so that GROUPING() behaves
+      correctly.
+    */
+    rollup_set_level(0);


you also need a complimentary join->rollup_set_level(0); in the send_row_on_empty_set() branch of do_select()
Add this to test

CREATE TABLE t1 (a INT PRIMARY KEY); INSERT INTO t1 VALUES (1); SELECT a, GROUPING(a) FROM t1 WHERE a=1 AND rand() > 2 GROUP BY a WITH ROLLUP;

mariadb-RexJohnston · 2026-08-24T21:54:41Z

+  if (Item_func::fix_fields(thd, ref))
+    return true;
+
+  // longlong is 64 bits, so we can represent a maximum of 64 arguments


actually, longlong is 63 bits + sign bit. hint. hint.

does the sign bit not count? MySQL does this, so that if you have all 64 arguments activated the value of GROUPING is -1

Does it work with 64 args? I can see it working for 63.

It works for 64, added a test to confirm this

mariadb-RexJohnston · 2026-08-24T22:05:20Z

+
+longlong Item_func_grouping::val_int()
+{
+  if (!min_rollup_levels)


Is this the right thing to do here? We are trying to extract a value out a function that isn't resolved.

mariadb-RexJohnston · 2026-08-24T22:09:28Z

+
+bool Item_func_grouping::resolve_args(THD *thd, ORDER *group_list)
+{
+  if (!(min_rollup_levels= thd->alloc<uint>(arg_count)))


This is allocated on execution memory. Is the first allocation freed before, say, second execution of a prepared statement?

I'm not really sure. I thought allocating on the mem_root arena was standard for stuff like this, from just looking around the code.

mariadb-RexJohnston · 2026-08-24T22:14:28Z

+
+# Test with prepared statements
+PREPARE ps FROM "SELECT a FROM t1 GROUP BY a WITH ROLLUP HAVING GROUPING(a)=0";
+EXECUTE ps;


+EXECUTE ps;

mariadb-RexJohnston · 2026-08-24T22:15:32Z

+EXECUTE ps;
+PREPARE ps FROM
+"SELECT a FROM t1 GROUP BY a WITH ROLLUP HAVING GROUPING(a)=1";
+EXECUTE ps;


+EXECUTE PS;
+DEALLOCATE PREPARE ps;

Or for a single execution -EXECUTE IMMEDIATE .. that doesn't need deallocation.

mariadb-RexJohnston · 2026-08-24T22:18:29Z

+SELECT EXISTS (SELECT t0.x FROM t0 a WHERE 1=0 GROUP BY a.x WITH ROLLUP) AS e
+  FROM t0;
+DROP TABLE t0;
+


please add a condition pushdown test, like this

SELECT * FROM (SELECT a, GROUPING(a) g FROM t1 GROUP BY a WITH ROLLUP) dt WHERE dt.g=1;

gemini-code-assist Bot reviewed Jul 1, 2026

View reviewed changes

Comment thread sql/item_func.h Outdated

Comment thread sql/item_func.cc Outdated

Comment thread sql/item_func.cc

gkodinov added the External Contribution All PRs from entities outside of MariaDB Foundation, Corporation, Codership agreements. label Jul 2, 2026

jaeheonshim force-pushed the MDEV-22269-grouping branch from 257f8c9 to b1721fd Compare July 5, 2026 23:10

grooverdan added the GSoC label Jul 5, 2026

grooverdan assigned grooverdan and mariadb-RexJohnston Jul 5, 2026

jaeheonshim force-pushed the MDEV-22269-grouping branch 5 times, most recently from f635cc0 to 7011078 Compare August 10, 2026 23:07

jaeheonshim marked this pull request as ready for review August 10, 2026 23:10

jaeheonshim force-pushed the MDEV-22269-grouping branch 2 times, most recently from ca4106a to 597b361 Compare August 10, 2026 23:20

gkodinov requested changes Aug 11, 2026

View reviewed changes

jaeheonshim force-pushed the MDEV-22269-grouping branch from 597b361 to 4703752 Compare August 12, 2026 20:15

gkodinov requested changes Aug 13, 2026

View reviewed changes

jaeheonshim force-pushed the MDEV-22269-grouping branch from 4703752 to 13342cc Compare August 14, 2026 07:47

gkodinov approved these changes Aug 17, 2026

View reviewed changes

grooverdan requested a review from mariadb-RexJohnston August 18, 2026 07:14

jaeheonshim added 2 commits August 25, 2026 06:00

MDEV-22269 Add GROUPING() tests from MySQL

03b4c24

Tests for the GROUPING() function taken from mysql-test/t/olap.test

MDEV-22269 Add support for the GROUPING() function

3003500

SQL:2008 feature T431

mariadb-RexJohnston force-pushed the MDEV-22269-grouping branch from 13342cc to 3003500 Compare August 24, 2026 18:00

mariadb-RexJohnston requested changes Aug 24, 2026

View reviewed changes

Add test that uses GROUPING with max number of arguments

f5e1b40

grooverdan reviewed Aug 28, 2026

View reviewed changes

Comment thread sql/item_func.cc Outdated

Fix ER_TOO_MANY_ARGS

14d4b56

jaeheonshim force-pushed the MDEV-22269-grouping branch from 2da9b58 to 14d4b56 Compare August 28, 2026 20:13

Replaced min_rollup_levels check in val_int with a dbug_assert

e9a75eb

Uh oh!

Uh oh!

Conversation

jaeheonshim commented Jul 1, 2026

Uh oh!

jaeheonshim commented Jul 1, 2026

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

Uh oh!

Uh oh!

mariadb-RexJohnston commented Aug 3, 2026

Uh oh!

gkodinov left a comment

Choose a reason for hiding this comment

Uh oh!

gkodinov left a comment

Choose a reason for hiding this comment

Uh oh!

jaeheonshim commented Aug 14, 2026

Uh oh!

gkodinov left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

4 participants