python - SQLAlchemy not behaving correctly after table dropped -

i have test code creates tables , drops them each test case. however, tests fail after first 1 because relying on code uses sa.table() first , creates new tables if calling method errors nosuchtableerror. however, error not thrown after tables first dropped, though engine correctly reports not exist, , not created again.

i've reproduced behavior follows:

>>>import sqlalchemy sa >>>from sqlalchemy import * >>>m = metadata() 

the tables don't exist, calling sa.table here errors expected:

>>>t = sa.table('test', m, autoload_with=eng)  ... nosuchtableerror: test 

but if create table , drop it, sa.table not error expected:

>>>t = sa.table('test', m, column('t', string(2))) >>>m.create_all(eng) >>>eng.has_table('test') true >>>t = sa.table('test', m, autoload_with=eng) >>>eng.execute('drop table "test" cascade') <sqlalchemy.engine.result.resultproxy @ 0x106a06150> >>>eng.has_table('test') false >>>t = sa.table('test', m, autoload_with=eng) 

no error thrown after last call sa.table, though table not exist.

what need sa.table() correctly error after tables have been dropped? engine object passing knows tables not exist, there else need do, refreshing/reconnecting somehow, expected behavior?

turns out need refresh metadata object (create new one) each time call sa.table if expect schema change. solves problem.