Why does not Django release fetched db objects?

问题:

I have very simple django app:

models.py:

class Product(models.Model):
        name = models.CharField(max_length=1000, default="")
        desc = models.TextField(default="")

views.py:

from django.http import HttpResponse
from models import Product

def fetch(request):
        for p in Product.objects.all()[:300000]:
                pass
        return HttpResponse("done")

I've loaded 300k sample records in MySQL database, turned debug off in settings.py and tried executed fetch view - after it completes, django still sits on 700Mb of RAM

I understand that it needs memory to fetch all these 300k objects, but why on earth it keeps them after view functions exits?

Again, I'm with DEBUG=False, tried this with django dev web server and also with uwsgi and its the same weird behavior.

P.S. Verified with Django 1.4 and 1.5.4 on both python2.6/2.7 Linux 64-bit

回答1:

This is not really anything to do with Django. Generally, Python does not return memory to the operating system until it needs to.

See the effbot's explanation for more detail.

回答2:

This line is the culprit

for p in Product.objects.all()[:300000]:

The slicing forces the QuerySet to evaluate, ie. hit the database, and it then returns a list() containing the slice of objects, which you then can iterate over. And Django's QuerySet cache keeps it in memory since you might want to iterate over the same "QuerySet" again.

You could optimize this by using an iterator.

It's in the docs here