Change checkpoint_data and restore_connection_on_error to subtransaction

checkpoint_data and restore_connection_on_error used to commit() and
rollback() the db connection. This causes additional I/O and load.

Changed them to use 'SAVEPOINT name' and 'ROLLBACK TO name' to get a
faster method for handling errors within a tranaction.

One thing to note is that postgresql (unlike SQL std) doesn't
overwrite an older savepoint with he same name. It keeps all
savepoints but only rolls back to the newest one with a given name.
This could be a resource issue. I left a commented out release
statement in case somebody runs into an issue due to too many
savepoints. I expect it to slow down the import but....

Author: rouilj

roundup/backends/back_postgresql.py

@@ -203,21 +203,33 @@ def open_connection(self):
             self._add_fts_table()
             self.commit()
 
-    def checkpoint_data(self):
-        """Commit the state of the database. Allows recovery/retry
-           of operation in exception handler because postgres
-           requires a rollback in case of error generating exception
+    def checkpoint_data(self, savepoint="importing"):
+        """Create a subtransaction savepoint. Allows recovery/retry
+           of operation in exception handler because
+           postgres requires a rollback in case of error
+           generating exception.  Used with
+           restore_connecion_on_error to handle uniqueness
+           conflict in import_table().
         """
-        self.commit()
-
-    def restore_connection_on_error(self):
-        """Postgres leaves a cursor in an unusable state after
-           an error. Rollback the transaction to recover and
-           permit a retry of the failed statement. Used with
-           checkpoint_data to handle uniqueness conflict in
-           import_table()
+        # Savepoints take resources. Postgres keeps all
+        # savepoints (rather than overwriting) until a
+        # commit(). If an import fails because of a resource
+        # issue with savepoints, uncomment this line. I
+        # expect it will slow down the import but it should
+        # eliminate any issue with stored savepoints and
+        # resource use.
+        #
+        # self.sql('RELEASE SAVEPOINT %s' % savepoint)
+        self.sql('SAVEPOINT %s' % savepoint)
+
+    def restore_connection_on_error(self, savepoint="importing"):
+        """Postgres leaves a connection/cursor in an unusable state
+           after an error. Rollback the transaction to a
+           previous savepoint and permit a retry of the
+           failed statement. Used with checkpoint_data to
+           handle uniqueness conflict in import_table().
         """
-        self.rollback()
+        self.sql('ROLLBACK TO %s' % savepoint)
 
     def create_version_2_tables(self):
         # OTK store