IBX-12078: Adjusted index of content id and version to be unique

[bnowak]

🎫 Issue	IBX-12078

Related PRs:

• https://github.com/ibexa/installer/pull/212

Description:

The ibexa_content_version table allowed multiple rows with the same (contentobject_id, version) pair - nothing on the database level prevented duplicate content versions (e.g. created by concurrent version-creating requests). The missing key was discovered while working on IBX-12029.

Changes:

• Replaced the non-unique ibexa_content_version_idx_ver index on (contentobject_id, version) with the uq_ibexa_content_version_coid_version unique constraint, so duplicate version rows are now rejected by the database itself.
• Made the contentobject_id column non-nullable - it is part of the new unique key, and a version row without a content id is not a valid state.
• Added an integration test asserting that DoctrineDatabase::insertVersion() throws UniqueConstraintViolationException when inserting a duplicate (contentobject_id, version) pair.

Migration scripts for existing installations are handled in the related installer PR. Databases already containing duplicates must be cleaned up before applying the migration - see the Documentation section below for detection and cleanup snippets.

For QA:

SQL snippet to reproduce duplicates in db:

INSERT INTO ibexa_content_version
    (contentobject_id, created, creator_id, initial_language_id, language_mask,
     modified, status, user_id, version, workflow_event_pos)
SELECT
    src.contentobject_id,
    src.created,
    src.creator_id,
    src.initial_language_id,
    src.language_mask,
    src.modified,
    src.status,
    src.user_id,
    src.version,
    src.workflow_event_pos
FROM (
    SELECT contentobject_id, created, creator_id, initial_language_id, language_mask,
           modified, status, user_id, version, workflow_event_pos
    FROM ibexa_content_version
    ORDER BY MD5(CONCAT(id, CURRENT_TIMESTAMP))
    LIMIT 10 -- <<< ADJUST: number of rows to duplicate
) AS src;

Documentation:

SQL snippet to list all duplicates in the database. Clients can use that to locate content versions that need fixes (duplicates removal):

SELECT contentobject_id, version, COUNT(*) AS occurrences
FROM ibexa_content_version
GROUP BY contentobject_id, version
HAVING COUNT(*) > 1
ORDER BY contentobject_id, version;

SQL snippets as proposed solution for client to automatically removing of duplicates:

-- 1) Remove scheduler entries referencing version rows that are about to be deleted.
--    Must run BEFORE the deduplication delete, while the doomed ids are still resolvable.
--    * This is optional step to remove orphan rows which would be kept after execution of 2nd SQL below.
--       However, theoretically there shouldn't be duplications here, as `ibexa_scheduler_scheduled_entry`
--       contains unique key (`content_id`, `version_number`, `action`) which should prevent duplicates.
DELETE FROM ibexa_scheduler_scheduled_entry
WHERE version_id IN (
    SELECT doomed.id
    FROM (
        SELECT v.id
        FROM ibexa_content_version v
        INNER JOIN ibexa_content_version keeper
            ON keeper.contentobject_id = v.contentobject_id
           AND keeper.version = v.version
           AND keeper.id < v.id
    ) AS doomed
);

-- 2) Deduplication SQL: for every (contentobject_id, version) group keep the lowest id, delete every other row.
DELETE FROM ibexa_content_version
WHERE id IN (
    SELECT doomed.id
    FROM (
        SELECT v.id
        FROM ibexa_content_version v
        INNER JOIN ibexa_content_version keeper
            ON keeper.contentobject_id = v.contentobject_id
           AND keeper.version = v.version
           AND keeper.id < v.id
    ) AS doomed
);

[bnowak]

there are some ending with uq right?

@ViniTou I didn't find uq ending/suffixes, but starting/prefixes in you data intelligence layer package 😉
please see #780 (comment) and continue there if you have anything to add here - it's the same topic

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud