How do I annotate a Django queryset with a Regex capture group without using RawSQL so that I later can use that value for filtering and sorting?
For example, in PostgreSQL I could make the following query:
CREATE TABLE foo (id varchar(100));
INSERT INTO foo (id) VALUES ('disk1'), ('disk10'), ('disk2');
SELECT
"foo"."id",
CAST((regexp_matches("foo"."id", '^(.*\D)([0-9]*)$'))[2] AS integer) as grp2
FROM "foo"
ORDER BY "grp2"
From Django 1.8 onwards, you can use Func() expressions.
from django.db.models import Func
class EndNumeric(Func):
function = 'REGEXP_MATCHES'
template = "(%(function)s(%(expressions)s, '^(.*\D)([0-9]*)$'))[2]::integer"
qs = Foo.objects.annotate(
grp2=EndNumeric('id'),
).values('id', 'grp2').order_by('grp2')
Reference: Get sorted queryset by specified field with regex in django
You can use a custom Func class created to get it working, but I would like to implement in a better way, just like a normal function which could be used for further processing using other functions or annotations or etc. Like a "block" in the Django ORM ecosystem.
I would like to start with an "beta version" of the class which looks like this one:
from django.db.models.expressions import Func, Value
class RegexpMatches(Func):
function = 'REGEXP_MATCHES'
def __init__(self, source, regexp, flags=None, group=None, output_field=None, **extra):
template = '%(function)s(%(expressions)s)'
if group:
if not hasattr(regexp, 'resolve_expression'):
regexp = Value(regexp)
template = '({})[{}]'.format(template, str(group))
expressions = (source, regexp)
if flags:
if not hasattr(flags, 'resolve_expression'):
flags = Value(flags)
expressions += (flags,)
self.template = template
super().__init__(*expressions, output_field=output_field, **extra)
and a fully working example for an admin interface:
from django.contrib.admin import ModelAdmin, register
from django.db.models import IntegerField
from django.db.models.functions import Cast
from django.db.models.expressions import Func, Value
from .models import Foo
class RegexpMatches(Func):
function = 'REGEXP_MATCHES'
def __init__(self, source, regexp, flags=None, group=None, output_field=None, **extra):
template = '%(function)s(%(expressions)s)'
if group:
if not hasattr(regexp, 'resolve_expression'):
regexp = Value(regexp)
template = '({})[{}]'.format(template, str(group))
expressions = (source, regexp)
if flags:
if not hasattr(flags, 'resolve_expression'):
flags = Value(flags)
expressions += (flags,)
self.template = template
super().__init__(*expressions, output_field=output_field, **extra)
@register(Foo)
class Foo(ModelAdmin):
list_display = ['id', 'required_field', 'required_field_string']
def get_queryset(self, request):
qs = super().get_queryset(request)
return qs.annotate(
required_field=Cast(RegexpMatches('id', r'^(.*\D)([0-9]*)$', group=2), output_field=IntegerField()),
required_field_string=RegexpMatches('id', r'^(.*\D)([0-9]*)$', group=2)
)
def required_field(self, obj):
return obj.required_field
def required_field_string(self, obj):
return obj.required_field_string
As you see in I've added 2 annotations and one outputs like a number and the other one like a normal string (character), of course, we don't see it in the admin interface but it does in the SQL are executed:
SELECT "test_foo"."id" AS Col1,
((REGEXP_MATCHES("test_foo"."id", '^(.*\D)([0-9]*)$'))[2])::integer AS "required_field", (REGEXP_MATCHES("test_foo"."id", '^(.*\D)([0-9]*)$'))[2] AS "required_field_string"
FROM "test_foo"
And also a screenshot with an example for you :)
Github gist with a better source code formatting https://gist.github.com/phpdude/50675114aaed953b820e5559f8d22166
