I think I get why Cloudflare’s CTO is asking this question. personally started using Django for backend development, boy, that thing is robust and cool, esp…
I think I get why Cloudflare’s CTO is asking this question. personally started using Django for backend development, boy, that thing is robust and cool, esp…
I use Django and there are some things that annoy me, like the fact we still need Django Rest Framework for a decent API experience, but at the same time, Django and DRF probably overuse inheritance and abstractions, to a point that every now and then I have to look up their source code to understand what’s exactly happening.
Lack of type annotations is another aspect Django really needs to catch up compared to other frameworks, and the whole dynamic nature of Django makes it difficult to have type stubs from third party packages working reliably, with lots of false negatives and false positives. So sometimes I still have runtime errors due to typing that could have been prevented in a different framework/ORM.
So I understand why some people are adopting different and more lightweight frameworks in new projects. The admin panel can be nice, but I’m not sure I’d choose Django myself to start a new project today.
Working with django the last 10 years. I mostly agree, but I think the inheritance thing results from the DRY mantra. Its efficient and prioritizes being able to easily change central things in the framework over readability. DRF is a decent standard, but the lack of further development and innovation (“considered feature complete”) is somewhat frustrating at times. For the typing thing just enable pydantic. :)
Does Django support async now? I remember that it was supposed to come with Django 4.
Yes it does, to some extent. It’s a work in progress. https://docs.djangoproject.com/en/6.0/topics/async/
That’s the other part that’s a bit problematic. There is some async support, but ORM calls, to my knowledge, are still blocking. There has been some effort put into it, but I’m not sure they’ll manage to reimplement it. Other parts of the framework and third party middlewares are blocking too, which makes writing async Django a bit of a minefield.
There needs to be a non negligible amount of work to make sure a given request remains asynchronous across all layers today.
That’s a pity.
One of the biggest things I miss about django is its ORM. There is simply nothing better in my opinion. It allows me to think about the application I want to write, not about all the database nonsense and how to join tables, make unions, or aggregate stuff, etc. I have to reevaluate if async is that important to me or not. Every other solution without a similar ORM feels inferior.
Possibly SQLalchemy? But yeah, it’s quite good, and there are ways you can bypass the Django ORM and write your own SQL in the cases where it’s suboptimal. I’ve only had to do that a very small number of times during the 20-plus years I’ve been using Django (I was an early adopter, left Zope like a .38 round leaves a gun).
It also helps that I have some decent SQL optimization skills: when I was doing consulting, I triggered the resignations of two client DBAs because they insisted that there was no way to significantly improve a query’s performance, then I did it. Stupid hill to die on, all they had to have done was to have said “OK, let’s prototype it and see” instead of getting all dogmatic about it. Ego’s a dangerous thing in software development.
SQLalchemy is horrible, IMO. You have to read so much of its documentation to get a grip on things. And the backwards compatible changes they made while moving to (I think) SQLAlchemy2 make it terribly difficult to tell apart the API. And type hints were an abomination to deal with when they tried adding them (it’s better now).
And don’t get me started on migrations with alembic.
I wish the Django ORM could be extracted from Django and used in other projects, but then I also just want to go back to Django.
While your point is valid, it’s a bit much to fault Django for the deficiencies of third-party middleware.