Django Unique返回的记录多于计数 [英] Django distinct returns more records than count

问题描述

以下显示了不同的计数为2247

The following shows the distinct count is 2247

In [6]: VirtualMachineResources.objects.all().values('machine', 'cluster')
   ...: .distinct().count()                                               
Out[6]: 2247

但是当我循环遍历它时，它返回的内容比应有的多：

But then when I loop through it, it returned way more than it should:

In [4]: a = [] 
   ...: for resource in VirtualMachineResources.objects.all().values('mach
   ...: ine', 'cluster').distinct(): 
   ...:     if resource['cluster']: 
   ...:         a.append(resource['cluster']) 
   ...:          
   ...:                                                                   

In [5]: len(a)                                                            
Out[5]: 96953

给定记录的簇字段为空，当我遍历查询集时，我可以看到同一台机器的很多重复而没有提示er值，但我只期望一次。

Given the records have cluster field being empty, when I loop through the queryset, I can see a lot of repeats of the same machine without cluster value, but I would expect only once.

for resource in VirtualMachineResources.objects.all().values('machine', 'cluster').distinct(): 
    print(resource['machine'], resource['cluster'])

打印...

server1
server1
server1

这是一个postgres数据库。有任何想法吗？
在Google和这似乎有关联吗？

It's a postgres database. Any ideas? Having a few more digs on google and this seems related?

更新：
创建了Django问题此处

推荐答案

您已经定义了 VirtualMachineResources.Meta.ordering 会混淆ORM，因为使用 DISTINCT 时，按列排序必须出现在 SELECT 子句中。使用 .count（）时，顺序会被清除。

You have a defined VirtualMachineResources.Meta.ordering that confuses the ORM as the ordered by columns must appear in the SELECT clause when using DISTINCT. The ordering happens to be cleared when using .count().

长存储短时，添加 .order_by（）将您的 Meta.ordering 放到您要遍历的查询集上，应该很好。没有很好的方法来生成 DISTINCT 来排除Django ORM上的排序字段，因为这将需要复杂的子查询下推，如＃24218 。

Long storing short, add a .order_by() to drop your Meta.ordering on the queryset you are iterating over and you should be good to go. There's no good way to generate a DISTINCT that excludes ordering fields on Django's ORM yet as that would require a complex subquery pushdown as detailed in #24218.

顺便说一句，请避免再次使用Django的票跟踪器层支持渠道，当您没有尽快收到您对问题的答复时。

By the way, please avoid Django's ticket tracker as a second tier support channel when you are not receiving a reply to your question as fast as you'd like.

这篇关于Django Unique返回的记录多于计数的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！