Comprehensive Django ORM Guide: Model Definition, Custom Table Names, Indexes, Relationships, Query Optimization, and Validation

The article introduces Django's ORM capabilities, explaining that a model can create database tables, operate on them, and perform data validation.

Custom Table Name

By default Django generates a table name as app_label_classname (e.g., app01_user ). To override this, define db_table inside the inner Meta class:

class User(models.Model):
    user = models.CharField(max_length=32)
    pwd = models.CharField(max_length=64)
    class Meta:
        db_table = "user"

Indexes

Single‑column indexes can be added with db_index=True on a field. Composite (joint) indexes are defined via index_together in Meta :

class User(models.Model):
    user = models.CharField(max_length=32)
    pwd = models.CharField(max_length=64)
    class Meta:
        db_table = "user"
        index_together = [("user", "pwd")]

Only queries that include the leftmost column of a composite index can use the index (e.g., WHERE user='zhangsan' or WHERE user='zhangsan' AND pwd='111' ).

Unique Composite Index

Use unique_together = (("name", "pwd"),) to enforce uniqueness of a column pair.

One‑to‑Many Relationship

class UserType(models.Model):
    name = models.CharField(max_length=32)

class User(models.Model):
    user = models.CharField(max_length=32, db_index=True)
    pwd = models.CharField(max_length=64, db_index=True)
    ut = models.ForeignKey(
        to="UserType",
        to_field="id",
        on_delete=models.CASCADE,
        related_name="cc",
        limit_choices_to={"id__gt": 2},
    )

When a UserType instance is deleted, Django 1.10+ cascades the delete to related User rows if on_delete=models.CASCADE is set.

ForeignKey Parameters

to : target model name.

to_field : target field name.

on_delete : behavior when the referenced row is removed ( CASCADE , DO_NOTHING , PROTECT , SET_NULL , SET_DEFAULT , SET(value) ).

related_name : custom name for reverse relation (e.g., cc ).

related_query_name : name used in reverse lookups (e.g., aa_set if set to aa ).

limit_choices_to : filter options shown in Django admin or ModelForm.

Forward and Reverse Lookups

Forward lookup example:

v1 = models.User.objects.all().values("user", "ut__name")

Reverse lookup example:

v3 = models.UserType.objects.all().values("name", "user__user")

Both can also be performed with explicit loops.

Many‑to‑Many Relationships

Three ways to create many‑to‑many relations:

Let Django create an implicit through table using ManyToManyField :

Manually create an explicit through table:

Manually create a through table and attach it via through and through_fields for finer control (recommended).

ORM Basic CRUD Operations

Examples of creating, reading, updating, and deleting records:

models.Tb1.objects.create(c1='xx', c2='oo')
obj = models.Tb1(c1='xx', c2='oo')
obj.save()
models.Tb1.objects.get(id=123)
models.Tb1.objects.filter(name='seven')
models.Tb1.objects.exclude(name='seven')
models.Tb1.objects.filter(name='seven').update(gender='0')
models.Tb1.objects.filter(name='seven').delete()

Additional QuerySet methods covered include count , range filters ( id__gt , id__lt ), in , isnull , contains , ordering, grouping with annotate , pagination ( [10:20] ), and raw SQL execution via raw() .

Performance Optimization

Using select_related('ut') performs a SQL join and fetches related UserType data in a single query, while prefetch_related('ut') runs two separate queries and caches the related objects, reducing the number of database hits in loops.

Model Validation

Calling full_clean() on a model instance triggers field‑level validation and the clean() hook. An example custom validation ensures the name field is unique:

def clean(self):
    if UserInfo.objects.filter(name=self.name).count():
        raise ValidationError("用户名已经存在", code="cc1")

The validation order is: field regex validation → clean() hook.

Overall, the article serves as a practical reference for Django developers needing detailed guidance on model design, indexing strategies, relationship handling, query optimization, and data validation.