如何按DateTimeField排序并基于ForeignKey选择不同的元素？

Question

I am using PostgreSQL as database for my Django project. I have the following models:

class Food(models.Model):
    """Stores a food entry"""
    name = models.CharField(max_length=50)
    image_id = models.CharField(max_length=20)
    description = models.TextField()

class SurveyResult(models.Model):
    """Stores survey food result based on emoji and personality"""
    emoji = models.CharField(max_length=10)
    personality = models.CharField(max_length=50)
    food = models.ForeignKey(Food, on_delete=models.CASCADE)
    timestamp = models.DateTimeField(auto_now_add=True)
    rating = models.PositiveSmallIntegerField(default=0)

I need to write a query sorted by '-timestamp' , and just get distinct values for 'food' . This is my what I have tried:

SurveyResult.objects.order_by('-timestamp').distinct('food')

However, I am getting the following error:

ProgrammingError: SELECT DISTINCT ON expressions must match initial ORDER BY expressions LINE 1: SELECT DISTINCT ON ("recommender_surveyresult"."food_id") "r...

I have read several answers from several questions, and I found an answer that basically says to do something like this:

SurveyResult.objects.order_by('food', '-timestamp').distinct('food')

However, I am not getting what I want that is the latest SurveyResult objects that have a distinct food .

How can I achieve this?

Let's suppose I have the following table in my database:

# surveyresults
id emoji personality food timestamp            rating
1  :)    Famous US     1  2018-01-01 - 9:00      1
2  XD    Famous CN     3  2018-01-01 - 9:10      2
3  :)    Famous NZ     5  2018-01-01 - 9:15      3
4  :(    Famous FR     5  2018-01-01 - 9:35      4
5  XD    Famous US     1  2018-01-01 - 9:45      5
6  ;)    Famous JP     4  2018-01-01 - 9:55      5
7  :o    Famous SP     4  2018-01-01 - 10:00     5

The query should give me the folowing:

id emoji personality food timestamp            rating
7  :o    Famous SP     4  2018-01-01 - 10:00     5
5  XD    Famous US     1  2018-01-01 - 9:45      5
4  :(    Famous FR     5  2018-01-01 - 9:35      4
2  XD    Famous CN     3  2018-01-01 - 9:10      2

I would also like to know if it is possible to achieve this just with one query.

Answer 1

You can achieve this using subquery as below,

SurveyResult.objects.filter(
    id__in=SurveyResult.objects.order_by('food', '-timestamp').distinct('food')
).order_by('-timestamp')

Explanation:

So basically subquery involve evaluation of more than one query , in your case, we are evaluating two queries. The SQL for the above queryset would look like,

SELECT "id", "emoji", "personality", "food", "timestamp", "rating"
FROM "appname_surveyresult"
WHERE id IN (
    SELECT "id", DISTINCT ON ("food") 
    FROM "appname_surveyresult"
    ORDER BY "food" ASC, "timestamp" DESC 
)
ORDER BY "timestamp" DESC

So firstly the inner (just like mathematics) query evaluates whose result becomes the input to the outer query resulting in the latest of latest among the distinct . Read this for more.

Answer 2

You might consider adding a meta class to your SurveyResults model and then specify your order there

class SurveyResult(models.Model):
    ...

   class Meta:
        ordering = ('-timestamp',)