Autogenerate creates indexes separately on new tables

[ziima]

I noticed autogenerate to create indexes on new tables separately.

Autogenerate output (batch mode on) for a new table looks like:

def upgrade():
    op.create_table(
        "my_table",
        # ... columns ...
        # ... constraints ...
    )
    with op.batch_alter_table("my_table", schema=None) as batch_op:
        batch_op.create_index(batch_op.f("my_table_status_key"), ["status"], unique=False)

While it seems that having the Index as part of the create_table works the same:

def upgrade():
    op.create_table(
        "my_table",
        # ... columns ...
        # ... constraints ...
        Index(op.f("my_table_status_key"), "status", unique=False)
    )

Is there a reason for autogenerate to separate creation of indexes, or is it just an unoptimized part of autogenerate?
Similar difference also applies to dropping tables.

[zzzeek]

there's no reason it looks like an implementation artifact that I agree is not ideal, and I would support fixing it for this case.