Group by limit 1 in django

Question

I have the following models in Django (simplified for brevity):

class DistinctWord(models.Model):
    ...

class Word(models.Model):
    distinct_word = models.F         


        

Answer 1

How about (given your queryset of uw): [obj.words.first() for obj in uw]

Answer 2

Let:

uw # be a given queryset of UserWord's
dw # be a queryset of DistinctWords (will be derived from `uw`)
w  # be a queryset of Words needed (will be derived from `dw`)

Each UserWord has a DistinctWord, and each DistinctWord has many Word's (loosely denoted as uw>dw<w).

Here is my answer:

dw_id=uw.values_list('distinct_word_id', flat=True) # 1: get dw ids from uw
dw=DistinctWord.objects.filter(id__in=dw_id)        # 2: get dw's
w_first_id=dw.annotate(first_word=Min('words')).values_list('first_word', flat=True) 
                                                    # 3: find id of first word 
w=Word.objects.filter(id__in=w_first_id)            # 4: get first words

In summary: lines 1 and 2 get dw and should be just 1 trip to the database

line 3 uses annotate followed by values_list to find the id of first related Word

Line 4 brings the actual Word objects from the id's generated in the previous step. Lines 3 and 4 should be another trip to the database since annotate is not a terminal statement.

Thus 2 trips to the database (not tested).

Answer 3

You can do this using the Subquery API:

from django.db.models.expressions import Subquery, OuterRef

first_word = Word.objects.filter(
    distinct_word=OuterRef('distinct_word')
).order_by('pk').values('pk')[:1]

UserWord.objects.filter(
     # whatever filters...
).annotate(
     first_word=Subquery(first_word)
)

This will result in SQL that looks something like:

SELECT user_word.*,
       (SELECT word.id 
          FROM word 
         WHERE word.distinct_word_id = user_word.distinct_word_id
       ) AS first_word
  FROM user_word
 WHERE ...

This will probably not perform as well as a JOIN with a DISTINCT ON in postgres, and may not perform as well as a JOIN with a GROUP BY, as it will need to execute the subquery for each row.