sqlalchemy redshift TypeError: unhashable type: 'list' version 1.3.11

After upgrading to 1.3.11, I am getting the following error when trying to reflect a AWS Redshift table: TypeError: unhashable type: 'list'. I am using the sqlalchemy-redshift dialect.

from sqlalchemy import create_engine, Table, MetaData
connect_args = dict(sslmode="verify-ca", keepalives_idle=200)
uri = 'redshift+psycopg2://<user>:<pass>@redshift-cluster:5439/my-db'
engine = create_engine(uri, connect_args=connect_args)
meta = MetaData()
t = Table('my_table', meta, schema='my_schema', autoload=True, autoload_with=engine)

If I downgrade to 1.3.10 then it works fine. The only data types in the table I am trying to reflect are bigint, timestamp, date, and varchar.

[zzzeek]

they (sqlalchemy-redshift) are hitting a hard change I had to make to fix #4955. so what is happening is that they have options like “sortkey” which they are reflecting at https://github.com/sqlalchemy-redshift/sqlalchemy-redshift/blob/master/sqlalchemy_redshift/dialect.py#L583 . these options get passed back into methods like get_coulmns(), for them it is at https://github.com/sqlalchemy-redshift/sqlalchemy-redshift/blob/master/sqlalchemy_redshift/dialect.py#L418, and the @reflection.cache they are using, which is optional, now caches on all the keyword arguments, including the list-based “sortkey” and others we see coming from table arguments.

so redshift has to either a. ensure table kwargs like sortkey and others are hashable, usually by making them into tuples or b. don’t use @reflection.cache.

they also should look into having a CI build that tests against the latest SQLAlchemy version.

[anthnyprschka]

hi @zzzeek i have similar issue with mysql where it returns unhashable type: 'bytearray' when i try to reflect a table. any ideas how to solve that?