Quest

Example Django project for The Temple of Django Database Performance by Andrew Brookins.

This tree is the code/ submodule of the book repo, on branch temple-book. Listings in the book include:: files from here.

Versions

The runnable app targets Python 3.12 and Django 4.2. That is the last LTS whose ORM APIs still match the book (select_related, prefetch_related, iterator, JSONField, GIN indexes, concurrent index operations) without rewriting the listings.

Django 3.0 (the original pin) does not install on Apple Silicon and does not run against Postgres 15+.

Setup (Docker)

Docker is the default. Postgres, Redis, and Django run as Compose services. This is also the path later updates will use when the profiling chapter grows a local Grafana and OpenTelemetry stack.

From this directory:

make build
make dev          # http://127.0.0.1:8000/
make test
make superuser

If port 8000 is already taken:

QUEST_WEB_PORT=8001 make dev

Postgres is published on host port 15432 (psql -h 127.0.0.1 -p 15432 -U quest). Redis stays on the Compose network so it does not collide with a host redis-server.

make logs         # follow Django
make down         # stop the stack
make generate-events NUM=500000 USER_ID=1

Frontend asset watch is opt-in: docker compose --profile frontend up.

Without Docker

Postgres 16 and Redis on the host, plus a Python 3.12 venv. On macOS:

brew install postgresql@16 redis
brew services start postgresql@16
brew services start redis

If brew services does not listen:

pg_ctl -D /opt/homebrew/var/postgresql@16 start
redis-server --daemonize yes

Then:

make local-setup    # .venv, role/database `quest` / password `test`, migrate
make local-test
make local-run      # http://127.0.0.1:8000/  (PORT=8001 if needed)
make local-superuser

The N+1 demo

GET /analytics is intentionally N+1 (Event.objects.all(), template walks event.user.profile.account). GET /analytics_select_related is the book's fix. Do not add select_related to all_events or the chapter demo disappears.

Weather annotation

process_events writes hourly cloud cover from Open-Meteo. No API key.

generate_events does not add coordinates (a 500k run plus weather would blow the free 10,000 calls/day cap). For a small weather demo, with the Compose stack up:

docker compose exec web python manage.py generate_events --num 5 --user-id 1 --with-location
docker compose exec web python manage.py process_events

Generating data for performance problems

make generate-events NUM=500000 USER_ID=1

USER_ID is a Django user id (your superuser is usually 1). The web container must already be running (make dev).