Updating Field Values in PostgreSQL Migration

When converting a BooleanField to a CharField in PostgreSQL, you're on the right track with using the update() method for efficiency. After the migration, PostgreSQL typically represents the boolean values as strings 'true' and 'false' (or sometimes 't' and 'f').

Efficient Update Approach

The most efficient way to perform this update is to use Django's update() method with a case expression directly in the migration. This avoids loading all objects into memory and performs the update in a single SQL operation:

PYTHONCopied
from django.db import migrations, models
from django.db.models import Case, When, Value

class Migration(migrations.Migration):

    dependencies = [
        # Your dependencies here
    ]

    operations = [
        # First operation: Change the field type
        migrations.AlterField(
            model_name='mymodel',
            name='field',
            field=models.CharField(max_length=2),
        ),
        
        # Second operation: Update the values
        migrations.RunSQL(
            sql="UPDATE myapp_mymodel SET field = CASE WHEN field::text = 'true' THEN '01' WHEN field::text = 'false' THEN '02' END;",
            reverse_sql="",  # No reverse operation needed
        ),
    ]

Alternatively, you can use Django's ORM approach in a data migration:

PYTHONCopied
def update_field_values(apps, schema_editor):
    MyModel = apps.get_model('myapp', 'MyModel')
    # Update True values to "01"
    MyModel.objects.filter(field='true').update(field='01')
    # Update False values to "02"
    MyModel.objects.filter(field='false').update(field='02')

class Migration(migrations.Migration):

    dependencies = [
        # Your dependencies here
    ]

    operations = [
        migrations.AlterField(
            model_name='mymodel',
            name='field',
            field=models.CharField(max_length=2),
        ),
        migrations.RunPython(update_field_values),
    ]

This approach is efficient because:

1. It uses Django's update() method which translates directly to SQL
2. It doesn't load objects into memory
3. It performs bulk updates rather than individual saves
4. It doesn't trigger model save methods or signals

For 250,000 records, this approach should be quite performant as it leverages PostgreSQL's ability to perform bulk updates efficiently.