feature request, modifying sequences

[julienaubert]

Typos such as mistakenly writing user.pk instead of e.g. user.profile.pk are sometimes not caught in tests (e.g. assert may pass as both have pk==1). I wrote a small snippet to make such mistakes more easily blow up in tests. Perhaps a modification of it could be considered for an optional feature?

def setup_modified_seq(connection):
    # detect cases where we are using the wrong pk (either in test or in code, e.g. user.pk vs user.profile.pk)
    if connection.vendor == 'postgresql':
        print('is postgres vnedor:', connection.vendor)
        def get_seqs(cursor):
            cursor.execute("SELECT c.relname FROM pg_class c WHERE c.relkind = 'S'")
            for seq_row in cursor.fetchall():
                seq = seq_row[0]
                yield seq
        cursor = connection.cursor()
        for count, seq in enumerate(get_seqs(cursor)):
            cursor.execute("ALTER SEQUENCE {} RESTART WITH %s;".format(seq), [(count+1)*1000])

and in hook def pytest_configure(config): adding:

    from django.db import connections
    for connection in connections.all():
        from pytest_django.db_reuse import _monkeypatch
        create_test_db = connection.creation.create_test_db
        def create_test_db_with_modified_sequences(self, *args, **kwargs):
            create_test_db(*args, **kwargs)
            setup_modified_seq(connection)
        _monkeypatch(connection.creation, 'create_test_db', create_test_db_with_modified_sequences)

[pelme]

Thanks for the suggestion. I am working on a refactor of how databases are setup, and this kind of extra configuration will be easy to achieve. I will keep this in mind and make an example out of this use case.

[blueyed]

Support for sequences has been added in #619.
Closing this therefore, also since db setup is more granular by now anyway.