Schema Tool: Ignore some database fields

[mpdude]

When working with legacy databases, I often have some columns in the database that have no corresponding field in the entity class (for various reasons probably not relevant here).

This makes the schema tool almost useless as it always tries to drop or otherwise change these columns. You'll always have to carefully cherry-pick the "real" changes.

I know that the schema tool can be configured to ignore certain tables, but that does (to my knowledge) not work on the field level.

So, my idea was to add entity configuration meta-data to denote these "ignored" fields. But as I am writing this, adding something like the FilterSchemaAssetsExpression in \Doctrine\DBAL\Configuration for the field level seems like another possible solution (yes, that's another repo).

What do lead devs think? Would you support such a PR?

[Ocramius]

@mpdude I had this scenario multiple times, but this drags in an additional family of problems, as this would then leak also to FKs, constraints, indexes, sequences.

Not sure we want to dig into this rabbit hole, but you can try. What would you imagine the filters to look like?

[mpdude]

Oh, good point!

If it were a config setting in the DBAL layer, simply pretending that the fields would not be there – are FKs, indexes etc. an issue at all?

Sure, if you get this config wrong by hiding a field that is actually needed, you might be surprised why the schema tool tries to add it. But, either way, chances are somebody gets hurt.

As to how filters could look like, I probably don't get the point... Maybe a regex? :-)

[Ocramius]

Maybe a regex? :-)

Callback?

[dbu]

Note that it is possible to make the Doctrine Schema tool not see database fields with the onSchemaColumnDefinition event, and also for indexes with onSchemaIndexDefinition.

I wrote a blogpost with code examples: https://www.liip.ch/en/blog/doctrine-and-generated-columns

The listener is rather heavy though, so an explicit configuration would be easier to understand and involve less effort.

[Brewal]

What I did which works for me, is to use the postGenerateSchema event to manually add the columns, indexes and foreign keys of the fields you want to ignore to the schema. Here is the gist :

https://gist.github.com/Brewal/4a623208e7cd60c4dfb5ab9ab56bcb2e