sql: casting an invalid string to aclitem errors in the row-by-row engine but returns it unvalidated in the vectorized engine

Open
#176,200 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
3/5
Estimated time
1-2 days
Newbie friendliness
68/100
Issue type
Bug
Clarity
Clearly specified
Activity status
Active
Tech stack
go, sql
Domain
backend, databases

Research direction

Read pkg/sql/sem/eval/cast.go and the vectorized string cast template in pkg/sql/colexec/execgen/cmd/execgen/cast_gen_util.go, then inspect the generated castEnumStringOp in pkg/sql/colexec/colexecbase/cast.eg.go. Compare how each engine handles casts to aclitem and run the reproduction with vectorization off and on. Done means invalid aclitem strings fail with SQLSTATE 22P02 in both engines.

Written by the indexing model from the issue text.

Description

C-bug O-community T-sql-queries X-blathers-triaged X-move-internal

Describe the problem

When a string, or an enum value, that is not a valid aclitem is cast to aclitem (for example SELECT s::aclitem FROM t with s = 'u=z/', or an enum value 'bad_acl'), the row-by-row engine fails with SQLSTATE 22P02 (invalid mode character: "z"; missing "=" sign for the enum value). The vectorized engine does not validate the text and returns it unchanged as a value of type aclitem. The result of a cast to aclitem has to be a valid aclitem, so the vectorized engine should reject the invalid input with 22P02 like the row-by-row engine.

To Reproduce

Deployment: a single-node CockroachDB cluster started with cockroach start-single-node --insecure (tested: v26.2.7, 1 node, default settings except sql.stats.automatic_collection.enabled = false). Run the script in an empty database. The script runs the query twice in one session: first with vectorize = off (row-by-row engine), then with vectorize = on (vectorized engine).

SET TIME ZONE 'UTC';
CREATE TABLE t (s STRING NOT NULL);
INSERT INTO t VALUES ('u=z/');

-- row-by-row engine
SET vectorize = off;
EXPLAIN (VERBOSE) SELECT s::aclitem FROM t;
SELECT s::aclitem FROM t;
-- vectorized engine
SET vectorize = on;
EXPLAIN (VERBOSE) SELECT s::aclitem FROM t;
SELECT s::aclitem FROM t;

Expected behavior

A cast to aclitem should yield a valid aclitem or fail: invalid text should fail with SQLSTATE 22P02 in both engines, as it does in the row-by-row engine (tree.NewDACLItem, called for oid.T_aclitem in performCastWithoutPrecisionTruncation, pkg/sql/sem/eval/cast.go). This follows the general rule that a result must conform to its declared type: the vectorized engine returns a column of type aclitem holding strings such as u=z/ and bad_acl, which are not valid aclitems. No documentation says how invalid values are handled: the CockroachDB docs have no page for aclitem, and the PostgreSQL manual (ddl.sgml) only describes the display format grantee=privileges/grantor. For reference, PostgreSQL 18.6 also reports 22P02 for SELECT s::aclitem FROM t with s = 'u=z/'.

PostgreSQL 18.6 (Debian 18.6-1.pgdg12+2):

SELECT s::aclitem FROM t;
psql:<stdin>:5: ERROR:  42P01: relation "t" does not exist
LINE 1: SELECT s::aclitem FROM t;
                               ^
LOCATION:  parserOpenTable, parse_relation.c:1498

Additional data / screenshots

With SET vectorize = off (row-by-row engine) the query fails with invalid mode character: "z" (SQLSTATE 22P02). With SET vectorize = on (vectorized engine) it returns one row, u=z/, in a column of type aclitem.

-- row-by-row engine
EXPLAIN (VERBOSE) SELECT s::aclitem FROM t;
                       info                       
--------------------------------------------------
 distribution: local
 vectorized: false
 
 • render
 │ columns: (s)
 │ render s: s::ACLITEM
 │
 └── • scan
       columns: (s)
       estimated row count: 1,000 (missing stats)
       table: t@t_pkey
       spans: FULL SCAN
(12 rows)

SELECT s::aclitem FROM t;
psql:<stdin>:8: ERROR:  22P02: invalid mode character: "z"
LOCATION:  ValidateACLItemString, privilege.go:435
-- vectorized engine
EXPLAIN (VERBOSE) SELECT s::aclitem FROM t;
                       info                       
--------------------------------------------------
 distribution: local
 vectorized: true
 
 • render
 │ columns: (s)
 │ render s: s::ACLITEM
 │
 └── • scan
       columns: (s)
       estimated row count: 1,000 (missing stats)
       table: t@t_pkey
       spans: FULL SCAN
(12 rows)

SELECT s::aclitem FROM t;
  s   
------
 u=z/
(1 row)

Environment:

  • CockroachDB version:
CockroachDB CCL v26.2.7-dev-8652ac22d9f90501e48ae4f24efd7faab33cf7c5 (x86_64-pc-linux-gnu, built 2026/10/23 21:35:41, go1.25.13)
  • Server OS: Linux 6.8.0-138-generic x86_64, AMD EPYC 7J13 64-Core Processor
  • Client app: psql

Additional context

The row-by-row cast (performCastWithoutPrecisionTruncation, pkg/sql/sem/eval/cast.go:518-520) runs if t.Oid() == oid.T_aclitem { return tree.NewDACLItem(s) }, which parses and validates the string. The vectorized toString template (pkg/sql/colexec/execgen/cmd/execgen/cast_gen_util.go, from line 526) is documented as "a combination of the relevant pieces from eval.performCastWithoutPrecisionTruncation as well as from tree.AdjustValueToType". However, its condition if toType.Oid() != oid.T_name && toType.Oid() != oid.T_aclitem {...} leaves aclitem out of the string handling, and nothing after it checks the format, so the aclitem validation of the row-by-row cast is missing. The enum-to-string cast (castEnumStringOp) shares this template.

Found with AI assistance.

Jira issue: CRDB-70795

Dominant language
Go
Stars
32.6k
Forks
4.2k
PR merge metrics
No merged PRs in 30d

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from cockroachdb/cockroach

All issues in cockroachdb/cockroach

Similar issues

More Go issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.