False change for unique constraint with case-sensitive names in 1.4.0

[ods]

PostgreSQL, a unique constraint with uppercase letters in name. _compare_indexes_and_uniques takes name as _constraint_sig.name (just the name without quotes) for constraint from connection, but as _constraint_sig.md_name_to_sql_name(context) (gives the name in double quotes) for constraint from metadata. That causes false changes detection with script that drops and creates the same constraint.

The following change solves the problem:

--- a/alembic/autogenerate/compare.py
+++ b/alembic/autogenerate/compare.py
@@ -541,7 +541,7 @@ def _compare_indexes_and_uniques(
     conn_uniques_by_name = dict((c.name, c) for c in conn_unique_constraints)
     conn_indexes_by_name = dict((c.name, c) for c in conn_indexes)
     conn_names = dict(
-        (c.name, c)
+        (c.md_name_to_sql_name(autogen_context), c)
         for c in conn_unique_constraints.union(conn_indexes)
         if c.name is not None
     )

[zzzeek]

Ok there's no test case for that so need to add one.

[zzzeek]

I can't reproduce [edit: got it, doesnt happen on SQLA master]. Can you provide a reproduction case? I think just a CREATE TABLE that is generating the false positives should be enough clue.

    def test_nothing_changed_uq_w_mixed_case_name(self):
        m1 = MetaData()
        m2 = MetaData()

        Table(
            "nothing_changed",
            m1,
            Column("id", Integer, primary_key=True),
            Column("x", Integer),
            UniqueConstraint("x", name="SomeConstraint"),
            mysql_engine="InnoDB",
        )

        Table(
            "nothing_changed",
            m2,
            Column("id", Integer, primary_key=True),
            Column("x", Integer),
            UniqueConstraint("x", name="SomeConstraint"),
            mysql_engine="InnoDB",
        )
        diffs = self._fixture(m1, m2)
        eq_(diffs, [])

in PG the table is created as:

CREATE TABLE nothing_changed (
	id SERIAL NOT NULL, 
	x INTEGER, 
	PRIMARY KEY (id), 
	CONSTRAINT "SomeConstraint" UNIQUE (x)
)

[sqla-tester]

Mike Bayer has proposed a fix for this issue in the master branch:

Handle mixed case identifiers correctly. https://gerrit.sqlalchemy.org/1709

[zzzeek]

that's just my test, will run it against multiple PG versions to confirm no issue so far.

[zzzeek]

oh, well duh, the whole issue that was fixed with that logic didn't happen in master :) k we're good thanks

[zzzeek]

different fix though. names aren't supposed to be quoted here.