Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-40853

SELECT INTO OUTFILE / LOAD DATA INFILE round-trip corrupts VECTOR column data

    XMLWordPrintable

Details

    • Bug
    • Status: Open (View Workflow)
    • Major
    • Resolution: Unresolved
    • 11.8, 12.3, 13.0, 13.1
    • 11.8, 12.3, 13.0
    • Vector search
    • None
    • Can result in unexpected behaviour

    Description

      VECTOR column does not survive a SELECT ... INTO OUTFILE followed by a LOAD DATA INFILE round-trip. Reloading the data either raises an error or, with LOAD DATA ... IGNORE, silently stores a zeroed-out vector instead of the original value.

      Test case

      CREATE TABLE t1 (id INT, v VECTOR(2) NOT NULL);
      INSERT INTO t1 VALUES (1, VEC_FROMTEXT('[1,2]'));
      SELECT HEX(v) FROM t1;
      SELECT * INTO OUTFILE '/tmp/vec_export.txt' FROM t1;
      CREATE TABLE t2 (id INT, v VECTOR(2) NOT NULL);
      LOAD DATA INFILE '/tmp/vec_export.txt' INTO TABLE t2;
       
      CREATE TABLE t3 (id INT, v VECTOR(2) NOT NULL);
      -- with IGNORE, the row is inserted, but with an incorrect, zeroed-out value instead of returning an error
      LOAD DATA INFILE '/tmp/vec_export.txt' IGNORE INTO TABLE t3;
      SELECT id, VEC_TOTEXT(v), HEX(v) FROM t3;
       
      #cleanup 
      DROP TABLE t1,t2,t3;
      

      CLI Output

      11.8.9-opt>CREATE TABLE t1 (id INT, v VECTOR(2) NOT NULL);
      Query OK, 0 rows affected (0.008 sec)
       
      11.8.9-opt>INSERT INTO t1 VALUES (1, VEC_FROMTEXT('[1,2]'));
      Query OK, 1 row affected (0.001 sec)
       
      11.8.9-opt>SELECT HEX(v) FROM t1;
      +------------------+
      | HEX(v)           |
      +------------------+
      | 0000803F00000040 |
      +------------------+
      1 row in set (0.001 sec)
       
      11.8.9-opt>SELECT * INTO OUTFILE '/tmp/vec_export.txt' FROM t1;
      Query OK, 1 row affected (0.000 sec)
       
      11.8.9-opt>
      11.8.9-opt>CREATE TABLE t2 (id INT, v VECTOR(2) NOT NULL);
      Query OK, 0 rows affected (0.005 sec)
       
      11.8.9-opt>LOAD DATA INFILE '/tmp/vec_export.txt' INTO TABLE t2;
      ERROR 1292 (22007): Incorrect vector value: '\0000\0000??\0000\0000\0000@' for column `test`.`t2`.`v` at row 1
      11.8.9-opt>CREATE TABLE t3 (id INT, v VECTOR(2) NOT NULL);
      Query OK, 0 rows affected (0.005 sec)
       
      11.8.9-opt>LOAD DATA INFILE '/tmp/vec_export.txt' IGNORE INTO TABLE t3;
      Query OK, 1 row affected, 1 warning (0.001 sec)
      Records: 1  Deleted: 0  Skipped: 0  Warnings: 1
       
      11.8.9-opt>SELECT id, VEC_TOTEXT(v), HEX(v) FROM t3;
      +------+---------------+------------------+
      | id   | VEC_TOTEXT(v) | HEX(v)           |
      +------+---------------+------------------+
      |    1 | [0,0]         | 0000000000000000 |
      +------+---------------+------------------+
      1 row in set (0.000 sec)
       
      11.8.9-opt>
      

      Attachments

        Activity

          People

            serg Sergei Golubchik
            ramesh Ramesh Sivaraman
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

            Dates

              Created:
              Updated:

              Time Tracking

                Estimated:
                Original Estimate - Not Specified
                Not Specified
                Remaining:
                Remaining Estimate - 0d
                0d
                Logged:
                Time Spent - 0.5h
                0.5h

                Git Integration

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