Bug/fix replicate table primary keys

[s7clarke10]

Resolving issue when a table has a primary key and unique key. Both unique and primary key columns were being identified as the primary key for the target table. Prioritising the primary key first, and unique key secondary if there is no primary key.

The code prior to the fix is overstating what the primary key is.

This change resolves #87

[mjsqu]

Recommended adding tests for the new logic

[mjsqu]

Can't add suggestions for files outside of the scope of the PR, however I recommend adding the following test here:
https://github.com/s7clarke10/pipelinewise-tap-mssql/blob/bug/fix_replicate_table_primary_keys/tests/test_tap_mssql.py#L652

class TestPrimaryKeyUniqueKey(unittest.TestCase):
    def setUp(self):
        self.conn = test_utils.get_test_connection()

        with connect_with_backoff(self.conn) as open_conn:
            with open_conn.cursor() as cursor:
                try:
                    cursor.execute("drop table uc_only_table")
                except:
                    pass
                try:
                    cursor.execute("drop table pk_only_table")
                except:
                    pass
                try:
                    cursor.execute("drop table pk_uc_table")
                except:
                    pass
                cursor.execute(
                    """
                    CREATE TABLE uc_only_table (
                      pk int,
                      uc_1 int,
                      uc_2 int,
                    CONSTRAINT constraint_uc_only_table UNIQUE(uc_1,uc_2)  )
                    """
                )
                cursor.execute(
                    """
                    CREATE TABLE pk_only_table (
                      pk int PRIMARY KEY,
                      uc_1 int,
                      uc_2 int,
                    )
                    """
                )
                cursor.execute(
                    """
                    CREATE TABLE pk_uc_table (
                      pk int PRIMARY KEY,
                      uc_1 int,
                      uc_2 int,
                    CONSTRAINT constraint_pk_uc_table UNIQUE(uc_1,uc_2)  )
                    """
                )

    def test_only_primary_key(self):
        catalog = test_utils.discover_catalog(self.conn, {})
        primary_keys = {}
        for c in catalog.streams:
            primary_keys[c.table] = (
                singer.metadata.to_map(c.metadata).get((), {}).get("table-key-properties")
            )

        self.assertEqual(primary_keys["uc_only_table"], ["uc_1","uc_2"])
        self.assertEqual(primary_keys["pk_only_table"], ["pk"])
        self.assertEqual(primary_keys["pk_uc_table"], ["pk"])

The test creates three tables for the scenarios you are addressing:

• uc_only_table - Unique Constraint Only
• pk_only_table - Primary Key Only
• pk_uc_table - Both Primary Key and Unique Constraints

Looking at the last three lines of the test, the assertions are used to test the expected primary key chosen in each scenario:

• uc_only_table - the uc columns
• pk_only_table - the pk column
• pk_uc_table - after your update this is just the primary key column - where prior behaviour was to combine pk and uc columns

To run this on your branch you can do:

docker run -p 1433:1433 -e "MSSQL_SA_PASSWORD=testDatabase1" -e "ACCEPT_EULA=Y" mcr.microsoft.com/mssql/server:2019-latest
poetry install
poetry run pytest -v -k TestPrimaryKey

(the pytest -k switch looks for any tests starting with the provided argument, so TestPrimaryKey matches just TestPrimaryKeyUniqueKey - the name of the added test)