automatic revison delects same changes when 'include_name' param of context.configure() set.

[gitNetw0rk]

Describe the bug

alembic revision --autogenerate -m "something1"keeps detecting previously migrated changes when applied:

def include_name(name, type_, parent_names):
if type_ == "schema":
return name in ["schema_one", "schema_two"]
else:
return True

context.configure(
# ...
include_schemas = True,
include_name = include_name
)

according to the documentation.

Expected behavior

to focus automatic script generation on selected schema, i.e. 'public':
def include_name(name, type_, parent_names):
if type_ == "schema":
return name in ["public"]
else:
return True

alembic revision --autogenerate -m "something2" should not detect changes detected at something1 moment that were already migrated with alembic upgrade head.

To Reproduce

class User(Base):
    __tablename__ = "users"

    ID: Mapped[pgUUID] = mapped_column(
        pgUUID(as_uuid=True),
        primary_key=True,
        nullable=False,
    )
    # later, add something i.e.:
    email: Mapped[str] = mapped_column(String, unique=True, nullable=False)

Error

...it just reports that new table detected and migration fails, because table exists in the database already

Versions.

• OS: Fedora Workstation 40
• Python: 3.12
• Alembic: 1.14.
• SQLAlchemy: 2.0.35
• Database: postgres image: pgvector/pgvector:0.8.0-pg17
• DBAPI:

Additional context

the solution to limit alembic down to selected db schema works as expected, however, it has bad side effect I was not able to debug any further.

Have a nice day!

[zzzeek]

this is not reproducible information and there is no general issue with the include_name callable, since you are using PostgreSQL and I see you mentioning the "public" schema, the recommended use for PostgreSQL is that "public" should never be named explicitly. best practices for PostgreSQL search_path are at https://docs.sqlalchemy.org/en/20/dialects/postgresql.html#remote-schema-table-introspection-and-postgresql-search-path , i.e. dont name "public" in your metadata models and ensure search_path does not contain other schemas inside of it.

for further assistance would need to see example models that you are using as well as what your search_path is set to.

[gitNetw0rk]

thanks @zzzeek for feedback, let me check it out on my side !