Creating Raw SQL Migrations in Django
I am using Django, which has great support for PostgreSQL-specific functions in it’s ORM. I wanted to use a SearchVectorField
inside my application and PostgreSQL recommends creating a trigger for to populate that column. Well that was easy enough but I didn’t want to have a snippet of SQL laying around to have to remember to put that in the database manually. Django has a great migration system – so why not use that? Can we manage a raw SQL entry in a Django migration?
It seems that you can, but no where spells it out step-by-step. I combined a few posts and Django documentation together and figured out how to create the trigger manually. Note that I didn’t make a backwards migration for it – I assumed if I didn’t need this after all, I will remove it myself or write another migration.
First step is to create an empty migration. This is simple: ./manage makemigration <myapp> --empty
. It will tell you the file of the empty migration file.
Open that file in your editor, and it will look something like the file below. The entry in the dependencies
line will be different based on your app name and what your previous migration was.
from django.db import connection from django.db import migrations class Migration(migrations.Migration): dependencies = [ ('myapp, '0012_auto_20190122_1540'), ] operations = [ ]
Now we need to write a function that will actually run the SQL that we need. Just put that function above the Migration
class in the file. Here was mine.
def make_trigger(apps,schema_editor): with connection.cursor() as cursor: cursor.execute("""CREATE TRIGGER caretaker_tsvectorupdate BEFORE INSERT OR UPDATE ON myapp_myobject FOR EACH ROW EXECUTE procedure tsvector_update_trigger(search_vector, 'pg_catalog.english', search_txt); """)
The cursor
will execute our command to create our trigger. Of course, you can put any SQL command you want in there.
Now you just need to tell the migration to run that function. That turns out to be a one-liner in the Operations
list. The final migration file looked list this:
from django.db import connection from django.db import migrations def make_trigger(apps,schema_editor): with connection.cursor() as cursor: cursor.execute("""CREATE TRIGGER caretaker_tsvectorupdate BEFORE INSERT OR UPDATE ON units_caretaker FOR EACH ROW EXECUTE procedure tsvector_update_trigger(search_vector, 'pg_catalog.english', search_txt); """) class Migration(migrations.Migration): dependencies = [ ('units', '0012_auto_20190122_1540'), ] operations = [ migrations.RunPython(make_trigger) ]
Now just run the migration: ./manage.py migrate myapp
and now your SQL will be ran in your existing table.
Hi, I’m newbie in Django and I’m trying to understand Django philosophy for databases consistency.
When dealing about consistency in databases, I understand (or understood?) that it refers to constraints, cascades, triggers, etc. at database level, hence a solid database diagram is (or was) very important. But I don’t feel that with Django, it seems that its philosophy is to bring all that consistency to every app in a Django project, and to leave database with relations only.
I could (hardly) accept that if it was not because while I was using the database API, no rule (defined in models) applied on every record saved into database, then danger alarms sounded.
Here a paradox: if it were necessary to establish the consistency at db level yet, why would I need define rules in models? is it not redundant?
Now I was asking to myself, is the concurrency not dealt by database itself if it has consistency? why would I need an extra package like “django-concurrency”?
I wish to know your expert opinion about these subjects. Thanks in advance.