A collection of reusable Django utility modules.
This is mostly for my own uses as I build more projects in Django. Whenever I create a bit of functionality that I'll probably need to re-use later, I'll add it here.
The microdev.models.ChangeLoggerMixin provides per-field change logging for any Django model.
The database can only keep track of a model instance's latest state, but there are plenty of use cases where it's important or useful to keep track of previous states. There are basic auditing uses but, more interestingly, tracking previous states opens up new avenues for data analysis.
For example, if you were running a dating site, you could average all users' popularity before and after changing their smoker status to "I quit".
Here's a simple Django model:
class MyModel(models.Model): favorite_color = models.CharField(max_length=128) favorite_number = models.IntegerField()
Next we define a trivial ChangeLog implementation class and add ChangeLoggerMixin to our model:
from microdev.models import ChangeLog from microdev.models import ChangeLoggerMixin class MyModelChangeLog(ChangeLog): pass class MyModel(models.Model, ChangeLoggerMixin): favorite_color = models.CharField(max_length=128) favorite_number = models.IntegerField() # Overrides from ChangeLoggerMixin _change_logger_mixin__change_log_class = MyModelChangeLog
The _change_logger_mixin__change_log_class variable needs to point to a ChangeLog implementation class so that the mix-in will know where to write the log entries. You don't have to have a dedicated ChangeLog implementation class for each model that you're logging. You could have all models write to the same ChangeLog implementation class, but it seems cleaner to me for each model to have its own separate set of logs. The mix-in supports either approach.
Remember to run:
This will add the MyModelChangeLog table to your database.
Now in your client code you can activate ChangeLoggerMixin logging with two simple calls: track_changes() and log_changes():
def do_something(request): # Create an instance just to have some initial data... my_model = MyModel() my_model.favorite_color = 'blue' my_model.favorite_number = 3 my_model.save() # Now activate change tracking my_model.track_changes() # Make changes my_model.favorite_color = 'purple' my_model.favorite_number = 9 my_model.save() # Log the changes my_model.log_changes(request.user)
The design assumes a specific Django User is driving these changes and expects the User to be passed in.
You'll see two new MyModelChangeLog entries:
2012-10-16 22:20:19+00:00 | User 1001 | MyModel 5 | favorite_color: blue -> purple 2012-10-16 22:20:19+00:00 | User 1001 | MyModel 5 | favorite_number: 3 -> 7
That's the __unicode__ output. The actual ChangeLog keeps all the data separated out to make it easy to run queries against it later. Here's the ChangeLog structure:
class ChangeLog(models.Model): VALUE_MAX_LENGTH = 1024 date_created = CreationDateTimeField() user = models.ForeignKey(User) model = models.CharField(max_length=128) obj_id = models.IntegerField() field_name = models.CharField(max_length=128) original_value = models.CharField(max_length=VALUE_MAX_LENGTH) updated_value = models.CharField(max_length=VALUE_MAX_LENGTH)
And obviously the ChangeLog will only log the fields that have changed. Unchanged fields are ignored and won't appear in the log.
Let's say your model has a field that always changes but you don't want to track it. No prob, just override the ignore_list:
class MyModel(models.Model, ChangeLoggerMixin): favorite_color = models.CharField(max_length=128) favorite_number = models.IntegerField() date_updated = models.DateTimeField() # Overrides from ChangeLoggerMixin _change_logger_mixin__ignore_list = ['date_updated',] _change_logger_mixin__change_log_class = MyModelChangeLog
Now any changes to the date_updated field will not be logged.
The ChangeLoggerMixin was adapted from Armin's original suggestion at: http://stackoverflow.com/a/111364/1639020