GROUP BY and Select MAX from each group in Django, 2 queries

gistfile1.txt

'''
given a Model with:

   category    = models.CharField(max_length=32, choices=CATEGORY_CHOICES)
   pubdate     = models.DateTimeField(default=datetime.now)
   <other fields>

Fetch the item from each category with the latest pubdate.

''' 

model_max_set = Model.objects.values('category').annotate(max_pubdate=Max('pubdate')).order_by()

q_statement = Q()
for pair in model_max_set:
    q_statement |= (Q(category__exact=pair['category']) & Q(pubdate=pair['max_pubdate']))

model_set = Model.objects.filter(q_statement)

Thanks @ns5d, exactly what I needed. Worked perfectly

God bless you..:)

You can use sub-queries to achieve the same thing in a single query as follows:

    sq = Item.objects.filter(category=OuterRef('category')).order_by('-pubdate')  # deferred execution
    Item.objects.filter(pk=Subquery(sq.values('pk')[:1]))

Alternatively, If you make Category a model like:

   category    = models.ForeignKey(Category, on_delete=models.SET_NULL, related_name='items')
   pubdate     = models.DateTimeField(default=datetime.now)

You can then do it in one query irrespective of database backend as follows:

    Item.objects.annotate(_sel=Max('category__items__pubdate')).filter(pubdate=F('_sel'))

I haven't tested which of them is faster, but both execute a single query on the database backend and work with every backend supported by Django, including SQLite.

Thanks, that's awesome.

Thanks, exactly what I needed :)

Awesome! Thanks a lot!

@michel4j Took me some time to adapt to my needs, but in the end it worked! Amazing solution!

It is very slow for large tables. Use the version provided by @ns5d

Thanks @ryanpitts for this trick.

MyModel.objects.all().order_by('id', '-date').distinct('id')

Works only on PostGres?

@Elawphant In theory, if Django supports the database, and the database supports GROUP BY and DISTINCT (these are basic operations), then everything should work.

Thank you Michel. Nice solution if PostGres isn't an option.

Thanks a lot. This worked for sqlite where distinct isn't an option.

Any Idea...???

class AllLogout(models.Model):

user = models.ForeignKey(Account,on_delete= models.CASCADE)
login_time = models.DateTimeField(null=True,)
logout_time = models.DateTimeField(null=True)


def __str__(self):
    return str(self.user) + ': ' + str(self.logout_time)

this is my model, i want output like this
one day user login&logout multiple times but i need day1 user login,logout time again day 2 same user login, logout time
i tried with this but not working

data1 = Account.objects.filter(Role='Agent').values_list('id', flat=True)

res = []

for agentname in data1:
agentdata = AllLogout.objects.filter(user_id=agentname).order_by('login_time').distinct()
res.append(agentdata1)

output like this:
agent1 login 2021-10-20 17:09:50.957295 logout 2021-10-21 21:45:54.955938
agent 3 login 2021-10-20 12:57:57.923346 logout 2021-10-20 12:58:20.246713
day 2
agent1 login 2021-10-21 12:58:40.147530 logout 2021-10-21 12:59:02.528687
agent 3 login 2021-10-21 21:45:54.858950 logout 2021-10-21 14:01:22.189149