Give anonymizer metadata responsibility to the AnonymizerRegistry class

[pounard]

Preparation work for later, which will allow dynamic anonymizers to be created differently than from a class name. This allows dynamic anonymizers creation from configuration.

This will allow packs to be written without PHP code.

In order to allow this:

1. we move all anonymizer metadata code into the registry itself, it's not the anonymizer responsability anymore (since some will be dynamic, we cannot know them beforehand in order to create them then ask for metadata, this is a chicken and egg problem),
2. we move the responsibility to create instances to the registry as well, which then also acts as a factory.

This is a first step, tests won't pass right away since I didn't fix them all, it's a PoC for discussion.

[SimonMellerin]

except my 2 comments, seems good to me.

Just one question, are you sure that if we use the same anonymization twice (on 2 different tables and/or columns, it is instantiated only once ? (ie one sample table will be created and used)

[pounard]

@SimonMellerin

Just one question, are you sure that if we use the same anonymization twice (on 2 different tables and/or columns, it is instantiated only once ? (ie one sample table will be created and used)

Yes, a new instance is being created every time, if not, then there is a bug, but not in this code, probably more in the Anonymizator, but we would already have found it in that case.

[SimonMellerin]

ok for me, if we put aside the fact that the tests don't pass 😅

[pounard]

ok for me, if we put aside the fact that the tests don't pass 😅

Yolo, the only way of living.