The other day I had to make an update to our database schema because we started tracking the language developers use when they integrate their website with Clef.

Piece of cake right? We use Flask and SQLAlchemy so adding a column to our MySQL database is as easy as doing this:

languages = db.Column(db.String(64))

But we don’t want to allow just any string to be stored in this field. We want to restrict the data that’s persisted so we don’t end up in a situation where JavaScript can be represented as ‘javascript’, ‘JS’, and ‘JavaScript.’

So what’s the best way to do this?

Declaring a column of type Enum (similar to a Django ChoiceField) would allow us to limit this field to a set of fixed string values.

language_choices = ('javascript', 'python', 'ruby', 'java', 'go')
language = db.Column(db.Enum(*language_choices))

But it would also lock us in to a set of choices that could only be updated by running a migration. (There are also higher-level issues with using Enum — namely that it violates normalization by storing data in the column definition rather than the actual database fields.)

Whereas the Enum type is too rigid, a String type column is too permissive, so we needed to find a way to enforce some constraints. Luckily, Python’s namedtuple module makes this a breeze.

As its name suggests, a namedtuple assigns names, in addition to the numerical index, to each of its members. This allows us to have tuple-like objects whose contents can be accessed with dot-notation.

from collections import namedtuple

LANGUAGES = [ 'javascript', 'python', 'ruby', 'java', 'go']
LANGUAGE_CHOICES = namedtuple('Languages', LANGUAGES)(*LANGUAGES)

In the code above, we declare a namedtuple and specify its field names and then create an instance by unpacking the LANGUAGES array into positional arguments.

>> print LANGUAGE_CHOICES
>> Languages(javascript='javascript', python='python', ruby='ruby', java='java', go='go')

Now wherever we set the language attribute on the model, we can just use the constant we’ve created.

model.language = LANGUAGE_CHOICES.python

Or if we need to validate an input, we can easily do that with the _asdict() method that maps a namedtuple’s field names to corresponding values.

selection = request.values.get('language')

if not LANGUAGE_CHOICES._asdict().get(selection):
    raise InvalidLanguageException('Invalid language')

Now, if we want to update the list of languages we track, we only have to modify our constant instead of running a migration. This allows for a more sane database configuration while still maintaining data integrity.

Are you a fan of namedtuples? Show your love by sharing this!

A practical use for namedtuples was originally published in Two Factor Authenticity on Medium, where people are continuing the conversation by highlighting and responding to this story.