Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Django ORM. Joining subquery

I have a table which contains list of some web sites and a table with statistics of them.

class Site(models.Model):
    domain_name = models.CharField(
        max_length=256,
        unique=True,
    )


class Stats(models.Model):
    date = models.DateField()
    site = models.ForeignKey('Site')
    google_pr = models.PositiveIntegerField()

    class Meta:
        unique_together = ('site', 'date')

I want to see all sites and statistics for a concrete date. If a stats record for the date doesn't exist, then the selection must contain only site.

If I use:

Site.objects.filter(stats__date=my_date)

I will not get sites which have no records for my_date in stats table. Because in this case the SQL query will be like the following:

SELECT *
FROM site
LEFT OUTER JOIN stats ON site.id = stats.site_id
WHERE stats.date = 'my_date'

The query condition will exclude records with NULL-dates and sites without stats will be not included to the selection.

In my case I need join stats table, which has already been filtered by date:

SELECT *
FROM site
LEFT OUTER JOIN
  (SELECT *
   FROM stats
   WHERE stats.date = 'my-date') AS stats
ON site.id = stats.site_id

How can I translate this query to Django ORM?

Thanks.

like image 269
psln Avatar asked Apr 07 '14 01:04

psln


People also ask

Does Django ORM support subquery?

Django allows using SQL subqueries.

What is OuterRef Django?

to Django users. According to the documentation on models. OuterRef: It acts like an F expression except that the check to see if it refers to a valid field isn't made until the outer queryset is resolved.

How prefetch_ related works?

select_related() “follows” foreign-key relationships, selecting additional related-object data when it executes its query. prefetch_related() does a separate lookup for each relationship and does the “joining” in Python.


2 Answers

In Django v2.0 use FilteredRelation

Site.objects.annotate(
    t=FilteredRelation(
        'stats', condition=Q(stats__date='my-date')
).filter(t__google_pr__in=[...])
like image 150
Charmy Avatar answered Oct 09 '22 11:10

Charmy


I had a similar problem and wrote the following utility function for adding left outer join on a subqueryset using Django ORM.

The util is derived from a solution given to add custom left outer join to another table (not subquery) using Django ORM. Here is that solution: https://stackoverflow.com/a/37688104/2367394

Following is the util and all related code:

from django.db.models.fields.related import ForeignObject
from django.db.models.options import Options
from django.db.models.sql.where import ExtraWhere
from django.db.models.sql.datastructures import Join


class CustomJoin(Join):
    def __init__(self, subquery, subquery_params, parent_alias, table_alias, join_type, join_field, nullable):
        self.subquery_params = subquery_params
        super(CustomJoin, self).__init__(subquery, parent_alias, table_alias, join_type, join_field, nullable)

    def as_sql(self, compiler, connection):
        """
        Generates the full
        LEFT OUTER JOIN (somequery) alias ON alias.somecol = othertable.othercol, params
        clause for this join.
        """
        params = []
        sql = []
        alias_str = '' if self.table_alias == self.table_name else (' %s' % self.table_alias)
        params.extend(self.subquery_params)
        qn = compiler.quote_name_unless_alias
        qn2 = connection.ops.quote_name
        sql.append('%s (%s)%s ON (' % (self.join_type, self.table_name, alias_str))
        for index, (lhs_col, rhs_col) in enumerate(self.join_cols):
            if index != 0:
                sql.append(' AND ')
            sql.append('%s.%s = %s.%s' % (
                qn(self.parent_alias),
                qn2(lhs_col),
                qn(self.table_alias),
                qn2(rhs_col),
            ))
        extra_cond = self.join_field.get_extra_restriction(
            compiler.query.where_class, self.table_alias, self.parent_alias)
        if extra_cond:
            extra_sql, extra_params = compiler.compile(extra_cond)
            extra_sql = 'AND (%s)' % extra_sql
            params.extend(extra_params)
            sql.append('%s' % extra_sql)
        sql.append(')')
        return ' '.join(sql), params

def join_to(table, subquery, table_field, subquery_field, queryset, alias):
    """
    Add a join on `subquery` to `queryset` (having table `table`).
    """
    # here you can set complex clause for join
    def extra_join_cond(where_class, alias, related_alias):
        if (alias, related_alias) == ('[sys].[columns]',
                                    '[sys].[database_permissions]'):
            where = '[sys].[columns].[column_id] = ' \
                    '[sys].[database_permissions].[minor_id]'
            children = [ExtraWhere([where], ())]
            return where_class(children)
        return None
    foreign_object = ForeignObject(to=subquery, from_fields=[None], to_fields=[None], rel=None)
    foreign_object.opts = Options(table._meta)
    foreign_object.opts.model = table
    foreign_object.get_joining_columns = lambda: ((table_field, subquery_field),)
    foreign_object.get_extra_restriction = extra_join_cond
    subquery_sql, subquery_params = subquery.query.sql_with_params()
    join = CustomJoin(
        subquery_sql, subquery_params, table._meta.db_table,
        alias, "LEFT JOIN", foreign_object, True)

    queryset.query.join(join)

    # hook for set alias
    join.table_alias = alias
    queryset.query.external_aliases.add(alias)

    return queryset

join_to is the utility function you want to use. For your query you can use it in as follows:

sq = Stats.objects.filter(date=my_date)
q = Site.objects.filter()
q = join_to(Site, sq, 'id', 'site_id', q, 'stats')

And following statement would print a query similar to you example query (with subquery).

print q.query
like image 27
Debanshu Kundu Avatar answered Oct 09 '22 12:10

Debanshu Kundu