yamltodb drops database-wide objects when using a schema list

[rhunwicks]

I have a database with some schemas in it. Some of them (postgis) create user-defined CAST functions. I also have the plpythonu language installed.

I have a yaml file that defines a new schema and I want to add the new schema without affecting the other schemas in the database so I run:

yamltodb.py -n example_schema pyrdev example.yaml

This gives me the CREATE SCHEMA, CREATE TABLE, etc. that I am expecting to create the new schema, but it also gives:

DROP CAST (postgis.box3d AS box);
DROP CAST (postgis.box3d AS postgis.box2d);
DROP CAST (postgis.geometry AS bytea);
DROP CAST (postgis.geometry AS text);
DROP CAST (postgis.geometry AS postgis.geography);
DROP CAST (postgis.geography AS postgis.geography);
DROP CAST (postgis.box2d AS postgis.box3d);
DROP CAST (postgis.geometry AS postgis.box3d);
DROP CAST (postgis.chip AS postgis.geometry);
DROP CAST (postgis.geography AS postgis.geometry);
DROP CAST (text AS postgis.geometry);
DROP CAST (postgis.box3d_extent AS postgis.box2d);
DROP CAST (postgis.box2d AS postgis.geometry);
DROP CAST (postgis.geometry AS box);
DROP CAST (postgis.box3d_extent AS postgis.box3d);
DROP CAST (bytea AS postgis.geometry);
DROP CAST (postgis.geometry AS postgis.box2d);
DROP CAST (postgis.box3d AS postgis.geometry);
DROP CAST (postgis.box3d_extent AS postgis.geometry);
DROP LANGUAGE plpythonu;

I think that when yamltodb is run with a schema list parameter it should ignore objects defined at the database-wide level.

In order for round-tripping to work as expected I would also expect dbtoyaml not to include database-wide objects when running with a schema list parameter.

[rhunwicks]

Currently yamltodb attempts to drop the database-wide objects when using a schema list, even if they are defined in the yaml input file

[jmafc]

I agree that yamltodb -n shouldn't be dropping database-wide objects such as casts and languages. However, I don't think dbtoyaml -n should exclude such database-wide objects, since the dbtoyaml output would then be incomplete, e.g., a function in the selected schema may be using the language in question, or a view in that schema may use a cast.

[jmafc]

Could you please try the following fix:

diff --git a/pyrseas/database.py b/pyrseas/database.py
index 03d314f..3b86ee4 100644
--- a/pyrseas/database.py
+++ b/pyrseas/database.py
@@ -106,6 +106,9 @@ class Database(object):
         for sch in self.db.schemas.keys():
             if sch not in schemas:
                 del self.db.schemas[sch]
+        # exclude database-wide objects
+        self.db.languages = LanguageDict()
+        self.db.casts = CastDict()

     def from_catalog(self):
         """Populate the database objects by querying the catalogs

[jmafc]

Misunderstood cause of problem in #27.