/django-migrate-sql

Django Migrations for raw SQL

Primary LanguagePythonISC LicenseISC

django-migrate-sql-deux

Note

This package is a fork of the django-migrate-sql package, originally published by Bogdan Klichuk. This package appears unmaintained, so we decided to start a fork as we depended on it. Most of the code is from him.

Build Status codecov.io

Django Migrations support for raw SQL.

About

This tool implements mechanism for managing changes to custom SQL entities (functions, types, indices, triggers) using built-in migration mechanism. Technically creates a sophistication layer on top of the RunSQL Django operation.

What it does

  • Makes maintaining your SQL functions, custom composite types, indices and triggers easier.
  • Structures SQL into configuration of SQL items, that are identified by names and divided among apps, just like models.
  • Automatically gathers and persists changes of your custom SQL into migrations using makemigrations.
  • Properly executes backwards/forwards keeping integrity of database.
  • Create -> Drop -> Recreate approach for changes to items that do not support altering and require dropping and recreating.
  • Dependencies system for SQL items, which solves the problem of updating items, that rely on others (for example custom types/functions that use other custom types), and require dropping all dependency tree previously with further recreation.

What it does not

  • Does not parse SQL nor validate queries during makemigrations or migrate because is database-agnostic. For this same reason setting up proper dependencies is user's responsibility.
  • Does not create ALTER queries for items that support this, for example ALTER TYPE in Postgre SQL, because is database-agnostic. In case your tools allow rolling all the changes through ALTER queries, you can consider not using this app or restructure migrations manually after creation by nesting generated operations into `state_operations of RunSQL <https://docs.djangoproject.com/en/1.8/ref/migration-operations/#runsql>`__ that does ALTER.
  • (TODO)During migrate does not restore full state of items for analysis, thus does not notify about existing changes to schema that are not migrated nor does not recognize circular dependencies during migration execution.

Installation

Install from PyPi:

$ pip install django-migrate-sql-deux

Add migrate_sql to INSTALLED_APPS:

INSTALLED_APPS = [
    # ...
    'migrate_sql',
]

App defines a custom makemigrations command, that inherits from Django's core one, so in order migrate_sql app to kick in put it after any other apps that redefine makemigrations command too.

Usage

  1. Create sql_config.py module to root of a target app you want to manage custom SQL for.
  2. Define SQL items in it (sql_items), for example:
# PostgreSQL example.
# Let's define a simple function and let `migrate_sql` manage it's changes.

from migrate_sql.config import SQLItem

sql_items = [
    SQLItem(
        'make_sum',   # name of the item
        'create or replace function make_sum(a int, b int) returns int as $$ '
        'begin return a + b; end; '
        '$$ language plpgsql;',  # forward sql
        reverse_sql='drop function make_sum(int, int);',  # sql for removal
    ),
]
  1. Create migration ./manage.py makemigrations:

    Migrations for 'app_name':
      0002_auto_xxxx.py:
    - Create SQL "make_sum"
    

You can take a look at content this generated:

# -*- coding: utf-8 -*-
from __future__ import unicode_literals
from django.db import migrations, models
import migrate_sql.operations


class Migration(migrations.Migration):
    dependencies = [
        ('app_name', '0001_initial'),
    ]
    operations = [
        migrate_sql.operations.CreateSQL(
            name='make_sum',
            sql='create or replace function make_sum(a int, b int) returns int as $$ begin return a + b; end; $$ language plpgsql;',
            reverse_sql='drop function make_sum(int, int);',
        ),
    ]
  1. Execute migration ./manage.py migrate:

    Operations to perform:
      Apply all migrations: app_name
    Running migrations:
      Rendering model states... DONE
      Applying app_name.0002_xxxx... OK
    

Check result in ./manage.py dbshell:

db_name=# select make_sum(12, 15);
 make_sum
----------
       27
(1 row)

Now, say, you want to change the function implementation so that it takes a custom type as argument:

  1. Edit your sql_config.py:
# PostgreSQL example #2.
# Function and custom type.

from migrate_sql.config import SQLItem

sql_items = [
    SQLItem(
        'make_sum',  # name of the item
        'create or replace function make_sum(a mynum, b mynum) returns mynum as $$ '
        'begin return (a.num + b.num, 'result')::mynum; end; '
        '$$ language plpgsql;',  # forward sql
        reverse_sql='drop function make_sum(mynum, mynum);',  # sql for removal
        # depends on `mynum` since takes it as argument. we won't be able to drop function
        # without dropping `mynum` first.
        dependencies=[('app_name', 'mynum')],
    ),
    SQLItem(
        'mynum'   # name of the item
        'create type mynum as (num int, name varchar(20));',  # forward sql
        reverse_sql='drop type mynum;',  # sql for removal
    ),
]
  1. Generate migration ./manage.py makemigrations:
Migrations for 'app_name':
  0003_xxxx:
    - Reverse alter SQL "make_sum"
    - Create SQL "mynum"
    - Alter SQL "make_sum"
    - Alter SQL state "make_sum"

You can take a look at the content this generated:

# -*- coding: utf-8 -*-
from __future__ import unicode_literals
from django.db import migrations, models
import migrate_sql.operations


class Migration(migrations.Migration):
    dependencies = [
        ('app_name', '0002_xxxx'),
    ]
    operations = [
        migrate_sql.operations.ReverseAlterSQL(
            name='make_sum',
            sql='drop function make_sum(int, int);',
            reverse_sql='create or replace function make_sum(a int, b int) returns int as $$ begin return a + b; end; $$ language plpgsql;',
        ),
        migrate_sql.operations.CreateSQL(
            name='mynum',
            sql='create type mynum as (num int, name varchar(20));',
            reverse_sql='drop type mynum;',
        ),
        migrate_sql.operations.AlterSQL(
            name='make_sum',
            sql='create or replace function make_sum(a mynum, b mynum) returns mynum as $$ begin return (a.num + b.num, \'result\')::mynum; end; $$ language plpgsql;',
            reverse_sql='drop function make_sum(mynum, mynum);',
        ),
        migrate_sql.operations.AlterSQLState(
            name='make_sum',
            add_dependencies=(('app_name', 'mynum'),),
        ),
    ]

*NOTE: Previous function is completely dropped before creation because definition of it changed. CREATE OR REPLACE would create another version of it, so DROP makes it clean.*

*If you put ``replace=True`` as kwarg to an ``SQLItem`` definition, it will NOT drop + create it, but just rerun forward SQL, which is ``CREATE OR REPLACE`` in this example.*

  1. Execute migration ./manage.py migrate:
Operations to perform:
  Apply all migrations: app_name
Running migrations:
  Rendering model states... DONE
  Applying brands.0003_xxxx... OK

Check results:

db_name=# select make_sum((5, 'a')::mynum, (3, 'b')::mynum);
  make_sum
------------
 (8,result)
(1 row)

db_name=# select make_sum(12, 15);
ERROR:  function make_sum(integer, integer) does not exist
LINE 1: select make_sum(12, 15);
               ^
HINT:  No function matches the given name and argument types. You might need to add explicit type casts.

For more examples see tests.

Feel free to open new issues.