Opened 11 years ago

Closed 9 years ago

#22022 closed New feature (wontfix)

Allow field subclasses to require additional SQL

Reported by: Marc Tamlyn Owned by: nobody
Component: Database layer (models, ORM) Version: dev
Severity: Normal Keywords:
Cc: Triage Stage: Accepted
Has patch: yes Needs documentation: no
Needs tests: no Patch needs improvement: yes
Easy pickings: no UI/UX: no

Description (last modified by Federico Capoano)

Problem

In postgres, a number of data types require a specific extension to be installed. We have one instance of this in core at present, which is the postgis extension. Our documentation recommends[1] (in postgres 9.1+) that you explicitly create the extension in your primary database by hand, and then we customise the _create_test_db method on the postgis database backend to ensure the test database has the relevant extensions installed.

Other third party apps[2] recommend adding the extension to the template to ensure it's present in tests.

I wish to avoid adding a custom backend to d.c.postgres if at all possible, especially as using it in conjunction with gis should be permissible. Consequently, I'd like if possible for model fields to have pre_create_sql to allow each individual field to create the extensions as needed. The same mechanism could also be used for a user defined custom field to create functions which could then be used as a custom transform object.

This feature will be useful for:
1) one part of the de-hack-gis project
2) hstore
3) citext, trigrams, generating uuids database side and other possible future features of d.c.postgres

I've written a draft patch[3]. It probably needs some work (hence patch needs improvement) but an initial review of the names and code layout would be useful.

[1] https://docs.djangoproject.com/en/1.6/ref/contrib/gis/install/postgis/#creating-a-spatial-database-with-postgis-2-0-and-postgresql-9-1
[2] http://djangonauts.github.io/django-hstore/#_running_tests
[3] https://github.com/django/django/pull/2266

Change History (3)

comment:1 by Russell Keith-Magee, 11 years ago

Triage Stage: UnreviewedAccepted

In principle, sounds like a reasonable proposition to me. There's clearly some discussion happening on the pull request, but I suspect that this will only shape the final form of the feature, not the base functionality.

comment:2 by Federico Capoano, 10 years ago

Description: modified (diff)

comment:3 by Tim Graham, 9 years ago

Resolution: wontfix
Status: newclosed

Marc on the pull request, "I'm going to give up on this concept in favour of manual Operations."

Note: See TracTickets for help on using tickets.
Back to Top