Split the plpython regression test into test cases arranged by topic, instead

of the previous monolithic setup-create-run sequence, that was apparently
inherited from a previous test infrastructure, but makes working with the
tests and adding new ones weird.

src/pl/plpython/sql/plpython_params.sql

@@ -0,0 +1,34 @@
+--
+-- Test named and nameless parameters
+--
+
+CREATE FUNCTION test_param_names0(integer, integer) RETURNS int AS $$
+return args[0] + args[1]
+$$ LANGUAGE plpythonu;
+
+CREATE FUNCTION test_param_names1(a0 integer, a1 text) RETURNS boolean AS $$
+assert a0 == args[0]
+assert a1 == args[1]
+return True
+$$ LANGUAGE plpythonu;
+
+CREATE FUNCTION test_param_names2(u users) RETURNS text AS $$
+assert u == args[0]
+return str(u)
+$$ LANGUAGE plpythonu;
+
+-- use deliberately wrong parameter names
+CREATE FUNCTION test_param_names3(a0 integer) RETURNS boolean AS $$
+try:
+	assert a1 == args[0]
+	return False
+except NameError, e:
+	assert e.args[0].find("a1") > -1
+	return True
+$$ LANGUAGE plpythonu;
+
+
+SELECT test_param_names0(2,7);
+SELECT test_param_names1(1,'text');
+SELECT test_param_names2(users) from users;
+SELECT test_param_names3(1);