get_columns() does not reflect Identity columns

[suparngp]

CockroachDB supports GENERATED ALWAYS AS IDENTITY columns, and information_schema.columns correctly reports is_identity = 'YES' and identity_generation = 'ALWAYS' for these columns.

However, the dialect's get_columns() does not include identity information in the returned column dict. It returns default: nextval('..._seq') but identity is absent from the result.

This causes Alembic autogenerate to emit a spurious AlterColumnOp on every run for any Identity column — it sees Identity(always=True) in the model but no Identity in the reflected column, and tries to add it.

Reproduction

from sqlalchemy import create_engine, inspect, text

engine = create_engine("cockroachdb://...")

# Create a table with an Identity column
with engine.begin() as conn:
    conn.execute(text("CREATE TABLE IF NOT EXISTS identity_test (id BIGINT GENERATED ALWAYS AS IDENTITY PRIMARY KEY)"))

# Reflect it
inspector = inspect(engine)
cols = inspector.get_columns("identity_test")
for col in cols:
    print(col)
    # Expected: {..., 'identity': {'always': True, ...}, ...}
    # Actual:   {'name': 'id', 'type': INTEGER(), 'nullable': False, 'default': "nextval('public.identity_test_id_seq'::REGCLASS)", 'autoincrement': True, ...}
    # 'identity' key is absent

# Cleanup
with engine.begin() as conn:
    conn.execute(text("DROP TABLE IF EXISTS identity_test"))

Expected: The identity key should be populated when the column uses GENERATED ALWAYS AS IDENTITY.

Actual: The identity key is absent. The underlying sequence is reported as a plain default.

[gordthompson]

Hi @suparngp . Thanks for this. We'll check it out and report back here.