Confused with alembic + SQLAlchemy Base.metadata

[kernelguardian]

For using alembic autogenerate feature it is mentioned in the documentation to import the Base from models and add the Base.metadata attribute to a target variable. But when I tried importing the model class directly say myModelClass and assigning that to the target variable instead of Base.metadata , i found no difference. Later I tried importing models in other files and they were also detected by alembic without adding them to a list. what exactly does alembic do with the Base.metadata attribute.

[CaselIT]

I suggest reading the fist sections of this documentation page https://alembic.sqlalchemy.org/en/latest/autogenerate.html#auto-generating-migrations

TLDR: it compares what's in the metadata with was is in the db and writes the difference in the generated revision

[CaselIT]

I think Base is just referring to the example. what alembic looks for is one (or more) metadata instances, and usually when using the orm with declarative meta the base class of all models is called Base in the examples. But any metadata is fine there

[kernelguardian]

So say I have two models modelA in fileA.py and modelB in fileB.py. Do I have to assign target_metadata = [modelA, modelB] like this or target_metadata = Base or target_metadata = modelA.

[CaselIT]

Please note that with metadata we are talking about https://docs.sqlalchemy.org/en/14/core/metadata.html instances, not models.

And Base in the example is defined like this https://docs.sqlalchemy.org/en/14/orm/mapping_styles.html#declarative-mapping

[kernelguardian]

Sorry I made a mistake there ,
So say I have two models modelA in fileA.py and modelB in fileB.py. Do I have to assign target_metadata = [modelA.Base.metadata, modelB.Base.metadata] like this or target_metadata = Base.metadata or target_metadata = modelA.Base.metadata. does this makes sense?

So say I have two models modelA in fileA.py and modelB in fileB.py. Do I have to assign target_metadata = [modelA, modelB] like this or target_metadata = Base or target_metadata = modelA.

[CaselIT]

Assuming that the models are defined as

class modelA(Base):
   ...
...
class modelB(Base):
   ...

you would use Base.metadata