TASK: Add migration for workspaceName constraints

[dlubitz]

This migration rewrites all events containing a workspaceName, which doesn't match the new workspaceName constraints and rewrites them with (shortend) a md5 hash of themself.

Run the migration with:

./flow migrateevents:migratepayloadtovalidworkspacenames

If the migration found some workspaceNames or baseWorkspaceNames to migrate, please run a projection replay of the workspace projection.

./flow cr:projectionreplay --projection workspace

This sould just be neccesary if you update your Neos 9 from <= beta10 to a newer version.

Fixes #5201

[bwaidelich]

+1 by reading. Thanks a lot for cleaning after me 🫣

[mhsdesign]

Migration was not necessary.

as desired :)

[mhsdesign]

Hä but no,

Invalid workspace name "Test-wy99x" given.

i dont understand? It didnt work after all?

[dlubitz]

🤔 The migration just handles to long workspaceNames

[kitsunet]

Wait what is the problem with that? uppercase?

[dlubitz]

Yes, the new constraint only allows lowercase.

    private const PATTERN = '/^[a-z][a-z0-9\-]{0,' . (self::MAX_LENGTH - 1) . '}$/';

[mhsdesign]

jup bastis change only allows lowercase workspace names... i guess Test-wy99x should be transformed to test-wy99x right? Or a hash as well? live and user-admin are safe btw

[dlubitz]

I'd go for lowercase everything.

[dlubitz]

🙈 Also, it's not allowed to start with a number anymore ... and the md5 hash could be a number.

[dlubitz]

Wait.... but the regex in the migration should find your Test-wy99x ... but mysql Regex is case insensitive

[bwaidelich]

I read through this more in-depth now and I wonder why you used SUBSTR(..., 1, ...) – if the original value was uppercase it would still be invalid afterwards.
My original suggestion was:

CONCAT('w', MD5(JSON_UNQUOTE(JSON_EXTRACT(payload, '$.workspaceName')))))

(note the "w"-prefix)

[dlubitz]

The md5 hash is 32 chars long, but the workspaceName can just be 30 chars long. So I had to shorten it. To reduce the chance of collision, I removed the prefix to be able to keep 30 chars of the hash instead of 29.

But yeah...

[bwaidelich]

Ah, right, sorry

LEFT(CONCAT('w', MD5(JSON_UNQUOTE(JSON_EXTRACT(payload, '$.workspaceName'))))), 30)

could work

[bwaidelich]

Ah, I hadn't realized: SUBSTR(..., 1, x) is the same as LEFT(..., x) (since it's not zero-based like in PHP).
Still, the first character of an MD5 could be a digit

[dlubitz]

Bääähhh... what about "baseWorkspaceName"? This also needs to be migrated, right? @bwaidelich

And after the migration a projection replay is needed for workspace projection.