Use default behaviour inside include_object hook

[Enhaloed]

I'm currently trying to add some custom configuration to the env.py file. I'm trying to use the include_object hook in run_migrations_online()

My problem is that I want to add some detection for things that aren't handled by default, but not deal with all cases. If I perform my logic for include/exclude then arbitrarily return True then almost every column turns up in the migrations and similarly if I return False then it excludes changes that otherwise would've been detected by default.

When I added some logic to the compare_server_default hook I could use return None to allow sqlalchemy to use the default implementation but I can't find any similar usage for include_object.

Ideally I'm looking for a way for my code to do something like this:

def my_include_object(object, name, type_, reflected, compare_to):
    if (my custom logic condition to include the object):
        return True
    else:
        # Do the generic implementation of include_object

def run_migrations_online() -> None:
    context.configure(
            include_object=my_include_object
        )

Thanks for any help/suggestions!

[zzzeek]

the return value of True is the default. include_object is used for omitting things from autogenerate that normally would be considered for changes.

if the columns are turning up in migrations that would mean alembic sees something different about them.

To solve your problem we would need to know what it is you are actually trying to achieve, such as, what are these "things that arent handled by default". that sounds like a bigger problem

[Enhaloed]

Ah.. I think perhaps the docs could do with a little extra clarification with this? From reading and rereading it still reads as if it can become a "be all" include or exclude function that gets sent every object rather than specifically an omission. Regardless, thank you for the clarification. It was a different issue with the configs causing all of the columns to be included but it made me think that include_object was intentionally being sent every object.

Regarding what I'm trying to handle - I'm using mysql types such as sqlalchemy.dialects.mysql.BIGINT sqlalchemy.dialects.mysql.DECIMAL etc. so I can use unsigned. Changes to unsigned don't get included in autogenerated migrations, so I was hoping to try a hook and see if this information was available on the objects

[zzzeek]

I'm not really sure why signed/unsigned wouldnt be detected, however, if you want to add your own check for that, you would use the compare_type hook.

[zzzeek]

From reading and rereading it still reads as if it can become a "be all" include or exclude function that gets sent every object rather than specifically an omission.

that is accurate. every object is sent to it, and returning True means, "consider this object in the autogenration sweep", which means, "we will compare this object to the corresponding metadata/database object, and if different, generate a migration". it does not imply that the object will be forced to be compared as different, and therefore needs a change of some kind in the migration, that doesnt actually make any sense, what would autogenerate produce if it doesnt actually know what is different?

[Enhaloed]

Thanks again for the info, very much appreciated. In case someone else comes across this, this is what seemed to work for detecting unsigned changes (or adaptable to any mysql type specific options if needed)

def _compare_type(context, inspected_column, metadata_column, inspected_type, metadata_type):
    inspected_type_unsigned = getattr(inspected_type, 'unsigned', False)
    metadata_type_unsigned = getattr(metadata_type, 'unsigned', False)
    if not inspected_type_unsigned == metadata_type_unsigned:
        return True

    return None

&

context.configure(
    ...
    compare_type=_compare_type
    ...
)