Inheritance of defaults

[olvidio]

I have some iuses with pyrseas: a) I use schemas and user with char '-' in name, so i have to enclose it into double quote. b) i use extensive inheritance inter schemas

I modified table.py and ini.py to correct some iuses, but:

• I know to very little of python and less of pyrseas. So perhaps there os a better solution.
• I see in git hub that version I've installed is no the last or development, and honestly i not work with git hub. I send you this two files.

basicly:

ini.py

line 82 (aprox) qualsch = qualsch.strip('"')

line 431 return "ALTER %s %s OWNER TO \"%s\"" % ( self.objtype, self.identifier(), owner or self.owner)

table.py

line 420:

def inhquery(): """reglass no put schema 'public' in name of parent. Generate new query: """ return """SELECT inhrelid::regclass AS sub, quote_ident(nspname) || '.' || quote_ident(relname) as parent, inhseqno FROM pg_inherits INNER JOIN pg_class ON(inhparent = pg_class.oid) INNER JOIN pg_namespace ON (pg_class.relnamespace = pg_namespace.oid) ORDER BY 1, 3"""

line 704:

add defaults in inherits cols (overwrite, i don't kown how to chek if exists)

        if incol.inherited and incol.default:
            stmts.append(base + "ALTER COLUMN %s SET DEFAULT %s" % (incol.name, incol.default))

table.py.txt init.py.txt

[jmafc]

With regard to the quotes in schemas, a month ago I committed a change to fix issue #186 which is very similar to your suggested change (uses a local undelim function that is more precise than str.strip and it applies it to both schemas and objects). The fix is available on both the master and r0.8 branches.

[jmafc]

As far as the second issue in __init__.py is concerned, I'll have to check and test your suggested fix.

[jmafc]

I'll take a look at the suggested changes to table.py later and let you know. I'll leave them as they are for now, but it's preferable (although it may be a bit more work on your side) to post each issue separately, unless they're closely related.

[jmafc]

Concerning the ALTER OWNER issue, I added the following test to tests/dbobject/test_owner.py:

@@ -103,3 +103,15 @@ class OwnerToSqlTestCase(InputMapToSqlTestCase):
             'owner': 'someuser'}})
         sql = self.to_sql(inmap, [CREATE_TABLE])
         assert sql[0] == "ALTER TABLE sd.t1 OWNER TO someuser"
+
+    def test_change_table_owner_delim(self):
+        "Change the owner of a table with delimited identifiers"
+        inmap = self.std_map()
+        inmap.update({'schema a-schema': {'table a-table': {
+            'columns': [{'c1': {'type': 'integer'}}, {'c2': {'type': 'text'}}],
+            'owner': 'someuser'}}})
+        sql = self.to_sql(inmap, ["CREATE SCHEMA \"a-schema\"",
+                                  "CREATE TABLE \"a-schema\".\"a-table\" ("
+                                  "c1 integer, c2 text)"])
+        assert sql[0] == "ALTER TABLE \"a-schema\".\"a-table\" OWNER TO " \
+                          "someuser"

The test passes without any change to pyrseas/dbobject/__init__.py. Therefore, I don't see any need to take further action (but I'll submit the test since it's useful).

[jmafc]

I have also added two tests to tests/dbobject/test_table.py based on the existing test_map_inherit and test_table_inheritance but using delimited identifiers. The tests pass without any changes to Table.inhquery. I suspect this is due to the undelim function in split_schema_obj which handles both schema and object names, rather than only your suggested change which only did a strip on the qualsch variable.

[olvidio]

OK. Thanks. I close this issue and re-open separate concepts

[jmafc]

OK, but the only issue which I consider as open is the very last one, i.e., the one about column defaults in inherited columns, for which I still have to build a test case. I could reopen if necessary (and change its title).

[jmafc]

@olvidio I'm considering reopening this issue to deal with the inherited defaults, but I need some feedback. Please take a look at the test starting at https://github.com/perseas/Pyrseas/blob/8b787fe51ed7e226dc3805bf19c2e67b0a9e895a/tests/dbobject/test_table.py#L404 In the test, column c1 is created in table t1 and inherited in table t2, simply as an integer. If I were to add, for example, NOT NULL and DEFAULT 1 to the column, by using ALTER TABLE on t1, t2 would automatically inherit those properties. In other words, I don't think it's necessary to issue an ALTER TABLE t2 SET DEFAULT 1, as it seems you're suggesting in your proposed change.