[MDEV-21445] Strange/inconsistent behavior of IN condition when mixing numbers and strings - Jira

Details

Type: Bug
Status: Closed (View Workflow)
Priority: Critical
Resolution: Fixed
Affects Version/s: 10.2.16, 10.3.21, 5.5(EOL), 10.1(EOL), 10.2(EOL), 10.3(EOL), 10.4(EOL), 10.5
Fix Version/s: 10.3.36, 10.4.26, 10.5.17, 10.6.9, 10.7.5, 10.8.4, 10.9.2
Component/s: Data types
Labels:
- in
- upstream
- where
Environment:
Tested on Debian with 10.3.21 and Gentoo with 10.2.16.

Description

Hi,

today we encountered one strange behavior in an older project. Below are simple SQL statements that quickly show where the problem is.

When combining a number and a number as a string in an IN WHERE condition, it sometimes returns extra unwanted records.

As you can see, condition id = 1234 OR id = "97716021308405775" returns expected 2 records, but equivalent id IN(1234,"97716021308405775") returns 4 records. This unexpected behavior is independent on storage engine - it works the same with InnoDB, MYISAM or TokuDB.

We'll fix our ORM layer to use the IN condition correctly (with either numbers or strings only), but maybe it's a bug in MariaDB that should be fixed. Maybe it's just old, familiar, unexpected behavior that can't be corrected?

DROP TABLE IF EXISTS test;
CREATE TABLE `test` (
`id` bigint(20) unsigned NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB;

INSERT IGNORE INTO `test` VALUES(1234),(97716021308405775),(97716021308405770),(97716021308405780);

– OK
SELECT * FROM `test` WHERE `id` IN(97716021308405775);
SELECT * FROM `test` WHERE `id` IN("97716021308405775");
SELECT * FROM `test` WHERE `id` IN("1234",97716021308405775);
SELECT * FROM `test` WHERE `id` IN("1234","97716021308405775");
SELECT * FROM `test` WHERE `id` = 1234 OR id = "97716021308405775";

– NOT OK (returns 4 records instead of 2)
SELECT * FROM `test` WHERE `id` IN(1234,"97716021308405775");
SELECT * FROM `test` WHERE `id` IN("97716021308405775",1234);

Attachments

- Sort By Name
- Sort By Date
- Ascending
- Descending
- Thumbnails
- List
- Download All

MDEV-21445.diff
64 kB
2021-04-26 02:17
MDEV-21445-REVIEW.txt
3 kB
2021-04-26 02:20

Issue Links

is blocked by

MDEV-22976 CAST(JSON_EXTRACT() AS DECIMAL) does not handle boolean values

Closed

MDEV-29041 Redundant truncation warning on CAST(string_column AS DECIMAL)

Closed

is duplicated by

MDEV-25492 BETWEEN clause returns incorrect results on quoted 64-bit ints

Closed

relates to

MDEV-25871 compare error bigint to varchar

Closed

MDEV-29259 Comparison semantic of int = string changes with creation of an index

Confirmed

Activity

Ascending order - Click to sort in descending order

Ján Regeš created issue - 2020-01-08 17:37

Elena Stepanova added a comment - 2020-01-13 00:48

Thanks for the report. Reproducible as described on 5.5-10.5 and MySQL 5.6, 8.0.

Elena Stepanova added a comment - 2020-01-13 00:48 Thanks for the report. Reproducible as described on 5.5-10.5 and MySQL 5.6, 8.0.

Elena Stepanova made changes - 2020-01-13 00:48

Field	Original Value	New Value
Component/s		Data types [ 13906 ]
Component/s	N/A [ 14411 ]
Fix Version/s		10.1 [ 16100 ]
Fix Version/s		10.2 [ 14601 ]
Fix Version/s		10.3 [ 22126 ]
Fix Version/s		10.4 [ 22408 ]
Affects Version/s		5.5 [ 15800 ]
Affects Version/s		10.1 [ 16100 ]
Affects Version/s		10.2 [ 14601 ]
Affects Version/s		10.3 [ 22126 ]
Affects Version/s		10.4 [ 22408 ]
Assignee		Alexander Barkov [ bar ]
Labels	in where	in upstream where

Elena Stepanova made changes - 2020-01-13 00:48

Status

Open [ 1 ]

Confirmed [ 10101 ]

Alexander Barkov added a comment - 2020-01-14 07:32 - edited

The problem happens because convert_const_to_int() is applied inconsistently:

When this query is performed:
SELECT * FROM t1 WHERE id = 1234 OR id = "97716021308405775";
the string constant "97716021308405775" gets converted to a longlong constant 97716021308405775 using convert_const_to_int(). The further comparision is done in longlong format.
When this query is performed:
SELECT * FROM test WHERE id IN("1234","97716021308405775");
all string constants get converted to longlong constants using convert_const_to_int() and comparison is further done in longlong format.
When this query is performed:
SELECT * FROM test WHERE id IN(1234,"97716021308405775");
conversion of the string constant "97716021308405775" to longlong does not happen because pairs (id,1234) and (id,"97716021308405775") are compared using different data types. The latter is performed using DOUBLE format, which leads to precision loss and therefore extra rows are returned.

Alexander Barkov added a comment - 2020-01-14 07:32 - edited The problem happens because convert_const_to_int() is applied inconsistently: When this query is performed: SELECT * FROM t1 WHERE id = 1234 OR id = "97716021308405775" ; the string constant "97716021308405775" gets converted to a longlong constant 97716021308405775 using convert_const_to_int() . The further comparision is done in longlong format. When this query is performed: SELECT * FROM test WHERE id IN ( "1234" , "97716021308405775" ); all string constants get converted to longlong constants using convert_const_to_int() and comparison is further done in longlong format. When this query is performed: SELECT * FROM test WHERE id IN (1234, "97716021308405775" ); conversion of the string constant "97716021308405775" to longlong does not happen because pairs (id,1234) and (id,"97716021308405775") are compared using different data types. The latter is performed using DOUBLE format, which leads to precision loss and therefore extra rows are returned.

Alexander Barkov made changes - 2020-06-18 10:52

Affects Version/s

10.5 [ 23123 ]

Alexander Barkov made changes - 2020-06-18 10:52

Fix Version/s

10.5 [ 23123 ]

Alexander Barkov added a comment - 2020-06-18 11:02

The same problem happen with a simple CASE:

CREATE OR REPLACE TABLE t1 (id bigint(20) unsigned NOT NULL PRIMARY KEY);

INSERT INTO t1 VALUES(1234),(97716021308405775),(97716021308405770),(97716021308405780);

SELECT

id,

  CASE id

    WHEN '97716021308405770' THEN '70'

    WHEN '97716021308405775' THEN '75'

    WHEN '97716021308405780' THEN '80'

  END AS c1

FROM t1;

+-------------------+------+

| id                | c1   |

+-------------------+------+

|              1234 | NULL |

| 97716021308405770 | 70   |

| 97716021308405775 | 70   |

| 97716021308405780 | 70   |

+-------------------+------+

Alexander Barkov added a comment - 2020-06-18 11:02 The same problem happen with a simple CASE: CREATE OR REPLACE TABLE t1 (id bigint (20) unsigned NOT NULL PRIMARY KEY ); INSERT INTO t1 VALUES (1234),(97716021308405775),(97716021308405770),(97716021308405780); SELECT id, CASE id WHEN '97716021308405770' THEN '70' WHEN '97716021308405775' THEN '75' WHEN '97716021308405780' THEN '80' END AS c1 FROM t1; +-------------------+------+ | id | c1 | +-------------------+------+ | 1234 | NULL | | 97716021308405770 | 70 | | 97716021308405775 | 70 | | 97716021308405780 | 70 | +-------------------+------+

Alexander Barkov added a comment - 2020-06-18 11:03

BETWEEN works fine:

CREATE OR REPLACE TABLE t1 (id bigint(20) unsigned NOT NULL PRIMARY KEY);

INSERT INTO t1 VALUES(1234),(97716021308405775),(97716021308405770),(97716021308405780);

SELECT id, id BETWEEN "97716021308405775" AND "97716021308405775" AS c1 FROM t1;

+-------------------+----+

| id                | c1 |

+-------------------+----+

|              1234 |  0 |

| 97716021308405770 |  0 |

| 97716021308405775 |  1 |

| 97716021308405780 |  0 |

+-------------------+----+

Searched CASE works fine:

CREATE OR REPLACE TABLE t1 (id bigint(20) unsigned NOT NULL PRIMARY KEY);

INSERT INTO t1 VALUES(1234),(97716021308405775),(97716021308405770),(97716021308405780);

SELECT

id,

  CASE

    WHEN id='97716021308405770' THEN '70'

    WHEN id='97716021308405775' THEN '75'

    WHEN id='97716021308405780' THEN '80'

  END AS c1

FROM t1;

+-------------------+------+

| id                | c1   |

+-------------------+------+

|              1234 | NULL |

| 97716021308405770 | 70   |

| 97716021308405775 | 75   |

| 97716021308405780 | 80   |

+-------------------+------+

Alexander Barkov added a comment - 2020-06-18 11:03 BETWEEN works fine: CREATE OR REPLACE TABLE t1 (id bigint (20) unsigned NOT NULL PRIMARY KEY ); INSERT INTO t1 VALUES (1234),(97716021308405775),(97716021308405770),(97716021308405780); SELECT id, id BETWEEN "97716021308405775" AND "97716021308405775" AS c1 FROM t1; +-------------------+----+ | id | c1 | +-------------------+----+ | 1234 | 0 | | 97716021308405770 | 0 | | 97716021308405775 | 1 | | 97716021308405780 | 0 | +-------------------+----+ Searched CASE works fine: CREATE OR REPLACE TABLE t1 (id bigint (20) unsigned NOT NULL PRIMARY KEY ); INSERT INTO t1 VALUES (1234),(97716021308405775),(97716021308405770),(97716021308405780); SELECT id, CASE WHEN id= '97716021308405770' THEN '70' WHEN id= '97716021308405775' THEN '75' WHEN id= '97716021308405780' THEN '80' END AS c1 FROM t1; +-------------------+------+ | id | c1 | +-------------------+------+ | 1234 | NULL | | 97716021308405770 | 70 | | 97716021308405775 | 75 | | 97716021308405780 | 80 | +-------------------+------+

Alexander Barkov made changes - 2020-06-22 09:50

Link

This issue is blocked by ~~MDEV-22976~~ [ ~~MDEV-22976~~ ]

Alexander Barkov added a comment - 2020-08-10 06:34 - edited

A possible solution would be to compare pairs (int,string) as DECIMAL rather than DOUBLE.

@serg suggests trying this starting from 10.2.

We should suppress warnings when decimal2string() runs outside of the DECIMAL range, e.g. :

SELECT '1e100'=1;

as the comparison to INT will work correctly anyway.

Alexander Barkov added a comment - 2020-08-10 06:34 - edited A possible solution would be to compare pairs (int,string) as DECIMAL rather than DOUBLE. @serg suggests trying this starting from 10.2. We should suppress warnings when decimal2string() runs outside of the DECIMAL range, e.g. : SELECT '1e100' =1; as the comparison to INT will work correctly anyway.

Alexander Barkov added a comment - 2020-08-10 06:57

The problem is also repeatable with engines MyISAM and HEAP.

CREATE OR REPLACE TABLE t1 (

  id bigint(20) unsigned NOT NULL,

  PRIMARY KEY (id)

) ENGINE=MyISAM;

INSERT IGNORE INTO t1 VALUES(1234),(97716021308405775),(97716021308405770),(97716021308405780);

SELECT * FROM t1 WHERE id IN(97716021308405775);

SELECT * FROM t1 WHERE id IN('97716021308405775');

SELECT * FROM t1 WHERE id IN('1234',97716021308405775);

SELECT * FROM t1 WHERE id IN('1234','97716021308405775');

SELECT * FROM t1 WHERE id = 1234 OR id = '97716021308405775';

-- NOT OK (returns 4 records instead of 2)

SELECT * FROM t1 WHERE id IN(1234,'97716021308405775');

SELECT * FROM t1 WHERE id IN('97716021308405775',1234);

Alexander Barkov added a comment - 2020-08-10 06:57 The problem is also repeatable with engines MyISAM and HEAP. CREATE OR REPLACE TABLE t1 ( id bigint (20) unsigned NOT NULL , PRIMARY KEY (id) ) ENGINE=MyISAM; INSERT IGNORE INTO t1 VALUES (1234),(97716021308405775),(97716021308405770),(97716021308405780); SELECT * FROM t1 WHERE id IN (97716021308405775); SELECT * FROM t1 WHERE id IN ( '97716021308405775' ); SELECT * FROM t1 WHERE id IN ( '1234' ,97716021308405775); SELECT * FROM t1 WHERE id IN ( '1234' , '97716021308405775' ); SELECT * FROM t1 WHERE id = 1234 OR id = '97716021308405775' ; -- NOT OK (returns 4 records instead of 2) SELECT * FROM t1 WHERE id IN (1234, '97716021308405775' ); SELECT * FROM t1 WHERE id IN ( '97716021308405775' ,1234);

Alexander Barkov made changes - 2020-08-18 09:22

Status

Confirmed [ 10101 ]

In Progress [ 3 ]

Alexander Barkov added a comment - 2020-08-25 12:01

serg, please review a patch for 10.3:

https://github.com/MariaDB/server/commit/d19f380d62c3145258c229d102b829c2f1c77223

Thanks.

Alexander Barkov added a comment - 2020-08-25 12:01 serg , please review a patch for 10.3: https://github.com/MariaDB/server/commit/d19f380d62c3145258c229d102b829c2f1c77223 Thanks.

Alexander Barkov made changes - 2020-08-25 12:01

Assignee	Alexander Barkov [ bar ]	Sergei Golubchik [ serg ]
Status	In Progress [ 3 ]	In Review [ 10002 ]

Sergei Golubchik made changes - 2020-10-12 13:17

Assignee	Sergei Golubchik [ serg ]	Alexander Barkov [ bar ]
Status	In Review [ 10002 ]	Stalled [ 10000 ]

Julien Fritsch made changes - 2020-11-06 16:03

Fix Version/s

10.1 [ 16100 ]

Alexander Barkov made changes - 2021-04-26 02:17

Attachment		MDEV-21445.diff [ 57500 ]
Attachment		MDEV-21445-REVIEW.txt [ 57501 ]

Alexander Barkov made changes - 2021-04-26 02:18

Attachment

~~MDEV-21445~~-REVIEW.txt [ 57501 ]

Alexander Barkov made changes - 2021-04-26 02:20

Attachment

MDEV-21445-REVIEW.txt [ 57502 ]

Sergei Golubchik made changes - 2021-04-27 11:33

Link

This issue relates to ~~MDEV-25492~~ [ ~~MDEV-25492~~ ]

Alexander Barkov made changes - 2021-04-27 11:36

Link

This issue relates to ~~MDEV-25492~~ [ ~~MDEV-25492~~ ]

Alexander Barkov made changes - 2021-04-27 11:36

Link

This issue relates to ~~MDEV-25492~~ [ ~~MDEV-25492~~ ]

Ralf Gebhardt made changes - 2021-05-11 09:29

Assignee

Alexander Barkov [ bar ]

Oleksandr Byelkin [ sanja ]

Alice Sherepa made changes - 2021-06-08 12:32

Link

This issue relates to ~~MDEV-25871~~ [ ~~MDEV-25871~~ ]

Oleksandr Byelkin made changes - 2021-09-21 07:53

Labels

in upstream where

in need_feedback upstream where

Elena Stepanova added a comment - 2021-09-21 08:38

sanja:

Probably this is the review: https://lists.launchpad.net/maria-developers/msg12411.html

Elena Stepanova added a comment - 2021-09-21 08:38 sanja : Probably this is the review: https://lists.launchpad.net/maria-developers/msg12411.html

Sergei Golubchik made changes - 2021-10-25 10:04

Labels

in need_feedback upstream where

in upstream where

Julien Fritsch made changes - 2021-11-03 14:50

Assignee

Oleksandr Byelkin [ sanja ]

Alexander Barkov [ bar ]

Sergei Golubchik made changes - 2021-12-06 21:35

Workflow

MariaDB v3 [ 102851 ]

MariaDB v4 [ 143623 ]

Alexander Barkov made changes - 2022-07-06 06:01

Link

This issue is blocked by ~~MDEV-29041~~ [ ~~MDEV-29041~~ ]

Alexander Barkov made changes - 2022-07-06 12:17

Fix Version/s		10.3.36 [ 27513 ]
Fix Version/s		10.4.26 [ 27511 ]
Fix Version/s		10.5.17 [ 27509 ]
Fix Version/s		10.6.9 [ 27507 ]
Fix Version/s		10.7.5 [ 27505 ]
Fix Version/s		10.8.4 [ 27503 ]
Fix Version/s		10.9.2 [ 27115 ]
Fix Version/s	10.2 [ 14601 ]
Fix Version/s	10.3 [ 22126 ]
Fix Version/s	10.4 [ 22408 ]
Fix Version/s	10.5 [ 23123 ]
Resolution		Fixed [ 1 ]
Status	Stalled [ 10000 ]	Closed [ 6 ]

Alexander Barkov made changes - 2022-07-06 13:22

Link

This issue is duplicated by ~~MDEV-25492~~ [ ~~MDEV-25492~~ ]

Alexander Barkov made changes - 2022-07-06 13:22

Link

This issue relates to ~~MDEV-25492~~ [ ~~MDEV-25492~~ ]

Alexander Barkov made changes - 2022-08-18 05:44

Link

This issue relates to MDEV-29259 [ MDEV-29259 ]

People

Assignee:: Alexander Barkov

Reporter:: Ján Regeš

Votes:: 0 Vote for this issue

Watchers:: 8 Start watching this issue

Dates

Created:: 2020-01-08 17:37

Updated:: 2022-08-18 05:44

Resolved:: 2022-07-06 12:17

Git Integration

Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.

MariaDB Server

Details

Description

Attachments

Attachments

Issue Links

Activity

People

Dates

Git Integration