Introduce dynamic scope_path regexps Instead of relying on scope_path's being "one slash deep", we should instead allow for either: 1. scope_paths that have a pre-defined depth 2. scope_paths that can be arbitrarily deep We achieve 1 by setting an entities scope_logic to another logic module. We then recursively call getScopeDepth until we get to the topmost entity (that is, an unscoped entity). A little different is the solution to 2, since some entities can have an arbitrarily deep scope (such as Documents), we need to have some way of signaling this to getScopePattern. A clean solution is to return None, rather than a number. If None is returned, the SCOPE_PATH_ARG_PATTERN is returned as regexp instead, which will match an arbitrarily deeply nested scope. The solution for 2 requires that we return None somewhere in the scope_logic chain, the most straight forward method to do so is to override getScopeDepth anywhere such a scope is needed and make it return None. A more elegant solution however, is to set the scope_logic to that module in all entities that require it. Patch by: Sverre Rabbelier

     1

from django.db.backends.postgresql.introspection import DatabaseIntrospection as PostgresDatabaseIntrospection

     2

     3

class DatabaseIntrospection(PostgresDatabaseIntrospection):

     4

     5

    def get_relations(self, cursor, table_name):

     6

        """

     7

        Returns a dictionary of {field_index: (field_index_other_table, other_table)}

     8

        representing all relationships to the given table. Indexes are 0-based.

     9

        """

    10

        cursor.execute("""

    11

            SELECT con.conkey, con.confkey, c2.relname

    12

            FROM pg_constraint con, pg_class c1, pg_class c2

    13

            WHERE c1.oid = con.conrelid

    14

                AND c2.oid = con.confrelid

    15

                AND c1.relname = %s

    16

                AND con.contype = 'f'""", [table_name])

    17

        relations = {}

    18

        for row in cursor.fetchall():

    19

            # row[0] and row[1] are single-item lists, so grab the single item.

    20

            relations[row[0][0] - 1] = (row[1][0] - 1, row[2])

    21

        return relations