Make migration Command doesn't allow changing the autodetector

[Ahmed Ibrahim]

I had a situation where I want to subclass the Autodetector to add more operations that the ones in Django, an operation that does not create a row, but rather, a table (different model)

This line is copied from django/django/core/management/commands/makemigrations.py

# Detect changes
changes = autodetector.changes(
     graph=loader.graph,
     trim_to_apps=app_labels or None,
     convert_apps=app_labels or None,
     migration_name=self.migration_name,
)

The suggested approach is to save the autodetector on the class level, and subclasses can override it to use their custom version.

P.S: I'm not sure if this is considered a feature or an optimization

[Claude Paroz]

I also do think having a way to specify a custom autodetector class would be an improvement. However, the autodetector class should be the same in makemigrations as the one in migrate command, so it might not be trivial to define it so it doesn't require the user to customize the class separately in both commands.

[Ahmed Ibrahim]

Replying to Claude Paroz:

I also do think having a way to specify a custom autodetector class would be an improvement. However, the autodetector class should be the same in makemigrations as the one in migrate command, so it might not be trivial to define it so it doesn't require the user to customize the class separately in both commands.

Thank you my friend for accepting the ticket and also stating the difficulty of the ticket, yes it bumped up a bit, I will do my best and mention the pr so that you may have a look if you don't mind

[Ahmed Ibrahim]

​https://github.com/django/django/pull/18445

[Sarah Boyce]

Marking as a new feature as we should highlight this change in the docs/release notes for people to use it. It also should have tests