Concurrency Control

[duncangroenewald]

I am considering using Fluent for a client macOS application but need some form of concurrency control to make sure the user is not updating and overriding a newer copy of the record.

Typically this would be done using a TIMESTAMP property that gets updated by the database server when a record is updated. The client performing the update would need to perform the updated including the timestamp in the WHERE clause as explained below.

1. Fetch record (with timestamp)
2. Modify field
3. Save record (update record set field=xxx WHERE id=ID AND tstamp=fetchedtstamp)
4. Fetch the updated timestamp (updated by the server when the update is performed) and update the local model object so that any further updates have the correct timestamp

Is there any way to support this in Fluent currently ?

I don't believe the current Fluent @timestamp property is used to check that the record being updated in the database isn't a more recent version.

If another user modified the database record after the user has fetched the record then the update should fail and the model object should get the new version from the database.

[0xTim]

You can do an update when filtering with Fluent, but the timestamp doesn't do any checking, it's purely set when Fluent updates a model (it's done client side rather than database side).

In order to have behaviour where updates fail if a newer version exists you can either use a row lock or build the logic yourself

[duncangroenewald]

@0xTim - thanks - is there a way to extend the model safely to customise the update WHERE clause along the lines of "AND tstamp=currenttstamp" and to throw a specific exception if the update doesn't update exactly 1 record ?

I am not that familiar with the inner workings of the model to know if this is even feasible.

Not sure a row lock is practical for us.

[0xTim]

Unfortunately there's no mechanism to retune the number of rows in Fluent so would need to be an SQLKit query