我写了一个导出数据的视图,我的模型是这样的:
class Event(models.Model):
KIND_CHOICES = (('doing', 'doing'),
('done', 'done'),
('cancel', 'cancel'))
created_at = models.DateTimeField(auto_now_add=True)
created_by = models.ForeignKey('auth.User')
kind = models.CharField(max_length=20, choices=KIND_CHOICES)
事件是三种类型中的一种,每个用户每月可能有3~10个事件,首先我查询本月的事件:
events_this_month = Event.objects.filter(created_at__year=2013,
created_at__month=5)
然后找到所有用户:
users = User.objects.all()
我导出这样的数据:
for user in users:
# 1000 users with 5 events each
user_events = events_this_month.filter(created_by=user)
doing_count = user_events.filter(kind='doing').count()
done_count = user_events.filter(kind='done').count()
cancel_count = user.events.filter(kind='cancel').count()
append_to_csv([user.username, doing_count, done_count, cancel_count])
然后我尝试使用collections.Counter
,我认为这将减少计算SQL次数(实际上它从3000 +减少到1200):
for user in users:
user_events = events_this_month.filter(created_by=user)
counter = Counter(user_events.values_list('kind', flat=True))
doing_count = counter['doing']
done_count = counter['done']
cancel_count = counter['cancel']
...
哪种方式更好?
更像是惯用的惯用的方式更像是惯用?
答案 0 :(得分:1)
这未经过测试,但我们的想法是按user
进行分组,然后按kind
进行分组:
from django.db.models import Count
events_this_month = Event.objects.values('created_by', 'kind') \
.filter(created_at__year=2013, created_at__month=5) \
.annotate(cc=Count('kind'))
让我知道这是否有效,因为我没有测试过这个。