# Django Interview Questions: ORM, Middleware and DRF Deep Dive

> Django interview questions covering ORM optimization with select_related and prefetch_related, middleware architecture, and Django REST Framework serializer performance, permissions, and pagination patterns.

- Published: 2026-04-03
- Updated: 2026-04-03
- Author: SharpSkill
- Tags: django, python, interview, orm, middleware, drf, rest-api
- Reading time: 9 min

---

Django interview questions cover three pillars that separate senior candidates from the rest: ORM mastery, middleware architecture, and Django REST Framework (DRF) design patterns. This guide breaks down the exact questions hiring managers ask in 2026, with production-ready code examples using Django 5.2 LTS and DRF 3.17.

> **What interviewers actually test**
>
> Django interviews rarely ask about basic CRUD. The focus has shifted to QuerySet optimization (select_related vs prefetch_related), custom middleware design, and DRF serializer performance. Candidates who can explain N+1 queries and write efficient viewsets consistently outperform those who only know generic views.

## Django ORM Interview Questions: QuerySet Optimization

The most common Django ORM interview question targets the N+1 problem. Given a model with foreign key relationships, candidates must demonstrate when to use `select_related` versus `prefetch_related`.

```python
# models.py
from django.db import models

class Company(models.Model):
    name = models.CharField(max_length=200)
    founded_year = models.IntegerField()

class Developer(models.Model):
    name = models.CharField(max_length=200)
    company = models.ForeignKey(Company, on_delete=models.CASCADE, related_name="developers")
    skills = models.ManyToManyField("Skill", related_name="developers")

class Skill(models.Model):
    name = models.CharField(max_length=100)
    category = models.CharField(max_length=50)
```

The difference between the two methods comes down to the type of relationship being traversed.

```python
# queries.py — Correct approach for ForeignKey (single object)
developers = Developer.objects.select_related("company").all()
# Generates ONE SQL query with JOIN
# SELECT developer.*, company.* FROM developer INNER JOIN company ...

# Correct approach for ManyToMany (multiple objects)
developers = Developer.objects.prefetch_related("skills").all()
# Generates TWO SQL queries:
# 1. SELECT * FROM developer
# 2. SELECT * FROM skill INNER JOIN developer_skills WHERE developer_id IN (...)
```

`select_related` performs a SQL JOIN and works with ForeignKey and OneToOneField. `prefetch_related` executes a separate query and works with ManyToManyField and reverse ForeignKey relations. Mixing them up causes either unnecessary JOINs on large datasets or the dreaded N+1 pattern.

## Advanced ORM: Custom Managers and QuerySet Chaining

Interviewers often ask candidates to write a custom manager that encapsulates business logic. The goal is to verify that the candidate understands Django's manager pattern and can write reusable query interfaces.

```python
# managers.py
from django.db import models
from django.utils import timezone

class ActiveDeveloperQuerySet(models.QuerySet):
    def active(self):
        """Filter developers who logged in within the last 30 days."""
        cutoff = timezone.now() - timezone.timedelta(days=30)
        return self.filter(last_login__gte=cutoff)

    def senior(self):
        """Filter developers with 5+ years of experience."""
        return self.filter(years_experience__gte=5)

    def by_skill(self, skill_name):
        """Filter developers by a specific skill."""
        return self.filter(skills__name=skill_name)

class ActiveDeveloperManager(models.Manager):
    def get_queryset(self):
        return ActiveDeveloperQuerySet(self.model, using=self._db)

    def active(self):
        return self.get_queryset().active()
```

The key follow-up question: "Why use a custom QuerySet instead of just a Manager?" The answer is chaining. Custom QuerySet methods can be chained together, while Manager methods cannot be chained after the first call.

```python
# usage.py — QuerySet chaining in action
# This works because each method returns a QuerySet
senior_python_devs = (
    Developer.active_objects  # custom manager
    .active()                 # ActiveDeveloperQuerySet method
    .senior()                 # chains another QuerySet method
    .by_skill("Python")       # chains a third method
    .select_related("company")  # standard QuerySet method still works
)
```

> **Django 5.2 LTS: Composite Primary Keys**
>
> Django 5.2 introduced CompositePrimaryKey, a long-awaited feature for models that need multi-column primary keys. Interview questions about this feature are becoming more common, especially for candidates working with legacy databases or data warehouse schemas.

## Django Middleware Interview Questions: Request-Response Pipeline

Middleware questions test a candidate's understanding of Django's request-response lifecycle. The standard question: "Explain the order in which middleware processes a request and a response."

The answer follows a strict pattern. During a request, middleware classes execute top-to-bottom as defined in `MIDDLEWARE`. During a response, they execute bottom-to-top. This onion-layer architecture means the first middleware in the list wraps everything.

```python
# middleware.py
import time
import logging
from django.http import JsonResponse

logger = logging.getLogger(__name__)

class RequestTimingMiddleware:
    """Logs the time taken to process each request."""

    def __init__(self, get_response):
        self.get_response = get_response  # next middleware or view

    def __call__(self, request):
        start_time = time.monotonic()

        response = self.get_response(request)  # passes to next layer

        duration_ms = (time.monotonic() - start_time) * 1000
        logger.info(
            "method=%s path=%s status=%d duration=%.2fms",
            request.method,
            request.path,
            response.status_code,
            duration_ms,
        )
        return response
```

A common follow-up: "How would you short-circuit the middleware chain?" Returning an HttpResponse from `__call__` before calling `self.get_response(request)` stops the chain entirely. The remaining middleware and the view never execute.

```python
# middleware.py — Rate limiting with short-circuit
from django.core.cache import cache

class RateLimitMiddleware:
    """Blocks requests exceeding 100 per minute per IP."""

    def __init__(self, get_response):
        self.get_response = get_response

    def __call__(self, request):
        ip = request.META.get("REMOTE_ADDR")
        cache_key = f"rate_limit:{ip}"
        request_count = cache.get(cache_key, 0)

        if request_count >= 100:
            return JsonResponse(  # short-circuits — view never executes
                {"error": "Rate limit exceeded. Try again in 60 seconds."},
                status=429,
            )

        cache.set(cache_key, request_count + 1, timeout=60)
        return self.get_response(request)
```

## Middleware Hooks: process_view, process_exception, and process_template_response

Beyond `__call__`, Django middleware supports three optional hook methods. Interviewers use these to gauge depth of knowledge.

```python
# middleware.py — Full middleware with all hooks
class AuditMiddleware:
    def __init__(self, get_response):
        self.get_response = get_response

    def __call__(self, request):
        return self.get_response(request)

    def process_view(self, request, view_func, view_args, view_kwargs):
        """Called after URL resolution, before the view executes."""
        logger.info("Calling view: %s", view_func.__name__)
        return None  # returning None continues normal processing

    def process_exception(self, request, exception):
        """Called only if the view raises an exception."""
        logger.error("View exception: %s", exception, exc_info=True)
        return None  # returning None lets Django's default handling proceed

    def process_template_response(self, request, response):
        """Called if the response has a render() method (TemplateResponse)."""
        response.context_data["audit_timestamp"] = time.time()
        return response  # must return a response with render()
```

`process_view` fires after URL resolution but before the view. Returning `None` continues execution; returning an `HttpResponse` short-circuits. `process_exception` fires only on unhandled exceptions. `process_template_response` fires only for `TemplateResponse` objects, not regular `HttpResponse`.

## DRF Interview Questions: Serializer Performance

[Django REST Framework](https://www.django-rest-framework.org/) serializer questions focus on nested serialization and the performance implications of different approaches. The most asked question: "How do you handle nested relationships without causing N+1 queries?"

```python
# serializers.py
from rest_framework import serializers
from .models import Developer, Company, Skill

class SkillSerializer(serializers.ModelSerializer):
    class Meta:
        model = Skill
        fields = ["name", "category"]

class CompanySerializer(serializers.ModelSerializer):
    class Meta:
        model = Company
        fields = ["name", "founded_year"]

class DeveloperSerializer(serializers.ModelSerializer):
    company = CompanySerializer(read_only=True)   # nested FK
    skills = SkillSerializer(many=True, read_only=True)  # nested M2M

    class Meta:
        model = Developer
        fields = ["id", "name", "company", "skills"]
```

The serializer alone does not solve performance. The ViewSet must optimize the queryset.

```python
# views.py
from rest_framework import viewsets
from .models import Developer
from .serializers import DeveloperSerializer

class DeveloperViewSet(viewsets.ReadOnlyModelViewSet):
    serializer_class = DeveloperSerializer

    def get_queryset(self):
        return (
            Developer.objects
            .select_related("company")       # JOIN for FK
            .prefetch_related("skills")      # separate query for M2M
            .order_by("-id")
        )
```

Without `select_related` and `prefetch_related` in the ViewSet, each serialized developer triggers individual queries for its company and skills. On a list of 50 developers, that means 1 + 50 + 50 = 101 queries instead of 3.

## DRF Custom Permissions and Authentication

A frequent DRF interview question: "Write a custom permission that restricts access based on the user's role and the object's owner."

```python
# permissions.py
from rest_framework.permissions import BasePermission

class IsOwnerOrAdmin(BasePermission):
    """
    Object-level permission:
    - Admin users can access any object
    - Regular users can only access objects they own
    """
    message = "Access restricted to the object owner or admin users."

    def has_object_permission(self, request, view, obj):
        if request.user.is_staff:
            return True
        # Assumes the model has an 'owner' field pointing to User
        return obj.owner == request.user
```

```python
# views.py — Applying custom permissions
from rest_framework import viewsets, permissions
from .permissions import IsOwnerOrAdmin

class ProjectViewSet(viewsets.ModelViewSet):
    permission_classes = [permissions.IsAuthenticated, IsOwnerOrAdmin]

    def get_queryset(self):
        # Non-admin users only see their own projects
        if self.request.user.is_staff:
            return Project.objects.all()
        return Project.objects.filter(owner=self.request.user)

    def perform_create(self, serializer):
        serializer.save(owner=self.request.user)  # auto-assign owner
```

The nuance interviewers look for: `has_permission` runs on every request (list-level), while `has_object_permission` runs only when `get_object()` is called (detail-level). Forgetting to override `get_queryset` for list views creates a security gap where users can see objects they do not own.

> **Common DRF security mistake**
>
> Relying solely on has_object_permission without filtering the queryset leaves the list endpoint unprotected. Always combine object-level permissions with a filtered get_queryset to enforce access control on both list and detail views.

## DRF Throttling and Pagination Patterns

Throttling and pagination are standard follow-up questions. Interviewers want to see that candidates understand production-ready API configuration.

```python
# settings.py — Production DRF configuration
REST_FRAMEWORK = {
    "DEFAULT_THROTTLE_CLASSES": [
        "rest_framework.throttling.AnonRateThrottle",
        "rest_framework.throttling.UserRateThrottle",
    ],
    "DEFAULT_THROTTLE_RATES": {
        "anon": "20/minute",    # unauthenticated users
        "user": "200/minute",   # authenticated users
    },
    "DEFAULT_PAGINATION_CLASS": "rest_framework.pagination.CursorPagination",
    "PAGE_SIZE": 25,
}
```

```python
# pagination.py — Custom cursor pagination for consistent ordering
from rest_framework.pagination import CursorPagination

class CreatedAtCursorPagination(CursorPagination):
    page_size = 25
    ordering = "-created_at"       # must be a unique, sequential field
    cursor_query_param = "cursor"  # ?cursor=abc123
```

Cursor-based pagination outperforms offset pagination on large datasets because it does not require counting total rows. The trade-off: clients cannot jump to arbitrary pages. This is the expected answer when interviewers ask "Why would you choose cursor pagination over page number pagination?"

## Conclusion

Key takeaways for Django interview preparation:

- **ORM optimization**: Always match the query method to the relationship type. `select_related` for ForeignKey/OneToOne, `prefetch_related` for ManyToMany and reverse FK. Custom QuerySets enable chainable, reusable business logic.
- **Middleware architecture**: The request flows top-to-bottom, the response bottom-to-top. Short-circuiting by returning a response before `get_response()` is a fundamental pattern for rate limiting, auth checks, and request validation.
- **DRF serializer performance**: Nested serializers require explicit queryset optimization in the ViewSet. Without `select_related`/`prefetch_related`, serialization causes N+1 queries at scale.
- **DRF permissions**: Combine `has_object_permission` with a filtered `get_queryset` to enforce access control on both list and detail endpoints. Object-level permissions alone leave list views unprotected.
- **Throttling and pagination**: Cursor pagination scales better than offset on large tables. Configure throttle rates separately for anonymous and authenticated users.

Practice these patterns with [Django ORM interview questions](/technologies/django/interview-questions/django-orm-advanced) and [Django middleware questions](/technologies/django/interview-questions/django-middleware) on SharpSkill to build confidence before the real interview.

---

Source: SharpSkill (https://sharpskill.dev), tech interview preparation for your real stack.
HTML version of this page: https://sharpskill.dev/en/blog/django/django-interview-orm-middleware-drf