Modeling

The schema is defined as a template using native Python data types:

# we will reuse this structure in examples below

spec = {
    'title': 'Untitled',
    'comments': [
        {
            'author': str,
            'date': datetime.datetime.utcnow,
            'text': str
        }
    ],
}

You are free to design as complex a document as you need. The manipulation and validation functions (described below) support arbitrary nested structures.

When this “natural” pythonic approach is not sufficient, you can mix it with a more verbose notation, e.g.:

title_spec = IsA(str, default='Untitled') | Equals(None)

There are also neat shortcuts:

spec = {
    'url': nullable(str),
    'status': one_of(['new', 'in progress', 'closed']),
    'comments': [str],
    'blob': None,
}

This could be written a bit more verbosely:

spec = {
    'url': IsA(str) | Equals(None),
    'status': Equals('new') | Equals('in progress') | Equals('closed'),
    'comments': ListOf(IsA(str)),
    'blob': Anything(),
}

It is even possible to define schemata for dictionary keys:

CATEGORIES = ['books', 'films', 'toys']
spec = {
    'title': str,
    optional('price'): float,    # key is optional; value is mandatory
    'similar_items': {
        one_of(CATEGORIES): [    # suggestions grouped by category
            {'url': str, 'title': str}
        ],
    }
}

# (what if the categories should be populated dynamically?
#  well, the schema is plain Python data, just copy/update on the fly.)

And, yes, you can mix notations. See FAQ.

This very short intro shows that Monk requires almost zero learning to start and then provides very powerful tools when you need them; you won’t have to rewrite the “intuitive” code, only augment complexity exactly in places where it’s inevitable.

Manipulation

The schema can be used to create full documents from incomplete data:

from monk import merge_defaults

# default values are set for missing keys

>>> merge_defaults(spec, {})
{
    'title': 'Untitled',
    'comments': [],
}

# it's easy to override the defaults

>>> merge_defaults(spec, {'title': 'Hello'})
{
    'title': 'Hello',
    'comments': [],
}

# nested lists of dictionaries can be auto-filled, too.
# by the way, note the date.

>>> merge_defaults(spec, {'comments': [{'author': 'john'}]})
{
    'title': 'Untitled',
    'comments': [
        {
            'author': 'john',
            'date': datetime.datetime(2013, 3, 3, 1, 8, 4, 152113),
            'text': None,
        }
    ]
}

Validation

The same schema can be used to ensure that the document has correct structure and the values are of correct types:

from monk.validation import validate

# correct data: staying silent

>>> validate(spec, data)

# a key is missing

>>> validate(spec, {'title': 'Hello'})
Traceback (most recent call last):
   ...
MissingKey: 'comments'

# a key is missing in a dictionary in a nested list

>>> validate(spec, {'comments': [{'author': 'john'}]}
Traceback (most recent call last):
   ...
MissingKey: 'comments': #0: 'date'

# type check; also works with functions and methods (by return value)

>>> validate(spec, {'title': 123, 'comments': []})
Traceback (most recent call last):
    ...
ValidationError: 'title': must be str

Custom validators can be used. Behaviour can be fine-tuned.

The validate() function translates the “natural” notation to a validator object under the hood. To improve performance you can “compile” the validator once (using translate() function or by creating a validator instance in place) and use it multiple times to validate different values:

>>> from monk import *
>>> translate(str) == IsA(str)
True
>>> validator = IsA(str) | IsA(int)
>>> validator('hello')
>>> validator(123)
>>> validator(5.5)
Traceback (most recent call last):
    ...
AllFailed: 5.5 (must be str; must be int)

The library can be also viewed as a framework for building ODMs (object-document mappers). See the MongoDB extension and note how it reuses mixins provided by DB-agnostic modules.

Here’s an example of the MongoDB ODM bundled with Monk:

from monk.mongo import Document

class Item(Document):
    structure = dict(text=unicode, slug=unicode)
    indexes = dict(text=None, slug=dict(unique=True))

# this involves manipulation (inserting missing fields)
item = Item(text=u'foo', slug=u'bar')

# this involves validation
item.save(db)