Flávio Juvenal

Recent talks

0:24:57

Mixing reliability with Celery for delicious async tasks

DjangoCon US 2023 · December 14, 2023

Celery is a distributed system with several ways to lose tasks: publishing can fail, large arguments can overwhelm workers or RabbitMQ, database changes can commit without their follow-up task being queued, and workers can disappear before acknowledging work. Use publisher confirms when needed, pass references…

0:39:31

Why large Django projects need a data (prefetching) layer

DjangoCon US 2022 · November 24, 2022

Reusing Django REST Framework serializers across endpoints can hide the queryset work they require, causing N+1 queries, missing annotations, and logic failures as projects grow. Flávio Juvenal argues that the important design concern is reducing change amplification by keeping serializers and their data-fetching requirements aligned…

0:24:15

"Normalize until it hurts; denormalize until it works"...

DjangoCon US 2018 · November 8, 2018

Normalization reduces redundancy so each fact has one reliable source, preventing update, insertion, and deletion anomalies. Flávio Juvenal explains that Django developers can still introduce denormalization when extending models, while historical values such as order totals, shipment addresses, and published slugs should remain stored because…

Has appeared with

• Ernist W. Durbin III
• Filipe Ximenes
• Griffith Rees
• Katie McLaughlin
• Raphael Michel
• Sarah Heim
• Simon Willison
• Tim Allen

All talks