[Django]-How to squash recent Django migrations?

184👍

python manage.py squashmigrations <appname> <squashfrom> <squashto>

python manage.py help squashmigrations

https://docs.djangoproject.com/en/dev/topics/migrations/#migration-squashing

This will give you more granular control over which migrations to squash, and let you keep a cleaner commit history. Deleting + recreating all migrations may cause other issues such as circular dependencies depending on how models are constructed.

👤A—

33👍

You can just delete the migration files and run makemigrations again. If you have a dev deployment that uses these, you should migrate back to the one before the first one you delete.

Also, it’s probably a good idea to commit your code first, in case something goes wrong.

Also:

The slight complication with this is that if there’s custom RunPython code, it won’t be included in the new migration created by makemigrations

👤Dan

6👍

Squash migrations command was introduced in Django 1.9

If you are using Django 1.8 you need to

👤pymen

5👍

I created django-squash https://pypi.org/project/django-squash/ as a way to not have to deal with migrations on a per-app level or worse a per-app-specific-migration level, and handle it on a per-project level. The idea is to hopefully integrate it inside core Django at some point.

Basic idea:

  • You have a product, nothing open source that other people enhance, but yours, your teams, you deal with it.
  • After each release you want to squash all migrations you did in the past release and start a new because your product has evolved from what it was last release and your data model as well.
  • You squash, it looks to see if you’ve squashed previously, if you have, it will deleted any VERY old migrations that have no business in your codebase anymore. Finally, create a new snapshot of your migrations, and keep what migrations you had around.
  • You will do this every release/when ever you feel your tests are taking too long running all your migrations.

Example:

/app1/migrations/__init__.py
/app1/migrations/0001_initial.py
/app1/migrations/0002_created_user_model.py
/app1/migrations/0003_added_username.py
/app1/migrations/0004_added_password.py
/app1/migrations/0005_last_name.py

You’ve applied them all.

But every time you run your tests, every single one of those steps need to run, taking valuable time. So we squash. The new directory will look like this:

/app1/migrations/__init__.py
/app1/migrations/0001_initial.py
/app1/migrations/0002_created_user_model.py
/app1/migrations/0003_added_username.py
/app1/migrations/0004_added_password.py
/app1/migrations/0005_last_name.py
/app1/migrations/0006_squash.py

inside 0006_squash.py you will find a replaces = [..] with the names of migrations 1-5. You will also find a Migration.operations = [..] with everything you would expect if you deleted all your migrations and did a ./manage.py makemigrations + any RunSQL/RunPython with elidable=False. If you deploy to an environment that is missing any of migrations 1-5 it will apply it from source and not use 0006 AT ALL. (this is standard Django migrations)

Some time passes, now your migrations look like this:

/app1/migrations/__init__.py
/app1/migrations/0001_initial.py
/app1/migrations/0002_created_user_model.py
/app1/migrations/0003_added_username.py
/app1/migrations/0004_added_password.py
/app1/migrations/0005_last_name.py
/app1/migrations/0006_squash.py
/app1/migrations/0007_change_username_to_100_char.py
/app1/migrations/0008_added_dob.py

You squash again. This time the following will happen. Anything inside the replaces = [..] will be deleted. 0006_squash.py will be modified to have replaces be an empty list. Lastly the squash will be recreated with the new changes. All told, will look like this:

/app1/migrations/0006_squash.py
/app1/migrations/0007_change_username_to_100_char.py
/app1/migrations/0008_added_dob.py
/app1/migrations/0009_squash.py

Starting the cycle once again.

Leave a comment