It's the sample code for a presentation comparing different methods of loading bulk data into a Postgres database using Python -- sometimes with Django, sometimes without

Once you create the table (in some form of Postgres), the good stuff is in website/demo3/scripts. You can run the scripts with

python manage.py runscript <script-name>

(without the .py extension). Some of the scripts don't use any part of Django, but I'm still using the django-extensions "runscript" command just for consistency

• The obvious speed champion is d_make_bulk_file -- if you don't mind the complexity
• I'm surprised at how much faster it is to create a psycopg2 connection and use that, rather than using the one you get by from django.db import connection

• psycopg3 is out, and it supports prepared statements under certain conditions.
  • You can't use prepared statements with PgBouncer (maybe you could fake it with transactions). I'd hate to lose PgBouncer -- that might be the cheapest peformance win in history (if you have a lot of connections)
  • There are syntax and behavioral changes in psycopg3, but you're probably going to have to deal with that sooner or later