Marat Sharafutdinov

Yes.. I started testing Python 3.8 and this exceptions appeared 2 weeks later. And they disappeared after ~75 seconds after they were first seen.

Installed dependencies: ``` sqlalchemy==1.2.18 alembic==1.0.7 psycopg2==2.7.7 ``` Here is the table definition (`test_partitioning/test.py`): ```python from sqlalchemy import Column, Integer, MetaData, Table, func from sqlalchemy.dialects.postgresql import TIMESTAMP metadata = MetaData() test...

After that I removed migration 0002 and created a couple of partitions manually: ```sql postgres=# create table test_y2019m02d23 partition of test for values from ('2019-02-23') to ('2019-02-24'); CREATE TABLE postgres=#...

@zzzeek thanks for help! Would it be possible to make alembic understand that table is sub-partition and automatically exclude it from generation (maybe as a future feature)? And one more...

@zzzeek ok, understand you. Just tested 1.3.0b3 again with partitioned table and found that partitioned table doesn't included in new migration but indices of that table are included: ```python ......

I guess it's because of the following: https://github.com/sqlalchemy/sqlalchemy/blob/53c7fc4690ead7a58947ad982c0a47e1af18e154/lib/sqlalchemy/dialects/postgresql/base.py#L3277-L3282 @zzzeek what is the reason for such behavior?

@zzzeek but I don't have any `postgresql_where` parameters specified.. So I should include all of partial indices in `include_object` hook algorithm also to make it work correctly?

@zzzeek you are right, PG partial indices are not related to the subject of issue, - my bad. Currently I've got to the following workaround for indices of partitioned tables...

@zzzeek I guess my workaround with `include_object` for PG partial indices isn't necessary with `sqlalchemy>=1.3.6` due to https://github.com/sqlalchemy/sqlalchemy/issues/4771 ?

@zzzeek it works for me now without workaround at `include_object`. But I still need to exclude partitioned tables from migrations within `include_object` (subject of the issue).