micromongo

micromongo is a tiny layer around pymongo that allows you to create simple ORM-style classes that can perform validation, allow dot access to documents, auto-wrap queryset results, and give you pre/post save hooks.

It's designed with microframeworks in mind, but is application and framework agnostic. It is meant to simplify usage of pymongo and provide tools for common idioms, not to obscure pymongo or mongodb from your data structures.

You are welcome to open issues or send pull requests on micromongo's github

micromongo makes a few design decisions in the name of simplification that might not work for you:

• micromongo maintains a single global connection, so you cannot have models that connect to multiple mongodb servers
• there are a handfull of model names and document attribute names that will not work with micromongo models; these will be covered in the full docs
• you can only have one model per collection

getting started

To start off with micromongo, just import it:

>>> from micromongo import connect, Model
>>> c = connect()

connect takes the same arguments as pymongo's Connection object, and behaves almost identically, except that it attempts to automatically return query results wrapped in the appropriate Model classes. The connection object that you create via this call will be cached and used by the various ORM-style facilities, like Model.save(), Model.proxy, etc. If you want a clean, standard Connection object, you can get one easily:

>>> from micromongo import clean_connection
>>> clean = clean_connection()

Note that clean_connection does not take arguments and will always return a clean Connection class with the same settings as the current micromongo connection.

With these connection objects, you can create databases or do whatever you would with normal pymongo objects:

>>> db = c.test_db
>>> collection = db.test_collection
>>> collection.save({"docid": 1, "fail": False})
>>> collection.find_one()
{u'_id': ObjectId('...'), u'fail': False, u'docid': 1}

You can also declare your own Model for a particular collection in declarative style:

>>> class TestModel(Model):
        collection = 'test_db.test_collection'

>>> collection.find_one()
<TestModel: {u'_id': ObjectId('...'), u'fail': False, u'docid': 1}>

These classes have a number of additional features over a dictionary that can make them much more convenient to use. The document keys are all accessible as attributes:

>>> t = collection.find_one()
>>> t.fail
False
>>> t.docid
1

The documents are also easily persisted to the database:

>>> t.docid = 17
>>> t.save()
>>> clean.test_db.test_collection.find_one()
{u'_id': ObjectId('...'), u'fail': False, u'docid': 17}

defining models

Above, the collection attribute was assigned to our Foo model. This was a shortcut, though; if database and collection are assigned separately, the Model can figure out the full collection name. If the collection and database are not present, micromongo attempts to figure it out based on the class and module name of your Model. For instance, blog.Post will become blog.post, or stream.StreamEntry will become stream.stream_entry. Explicit is better than implicit, and it's encouraged that you set the collection manually.

Besides packing and unpacking results from the database, models can also define a spec document which can define defaults and perform validation before saving the model. Take a trivial blog post model:

>>> from micromongo.spec import *
>>> class Post(Model):
        collection = 'test_db.blog_posts'
        spec = dict(
            author=Field(required=True, default='jmoiron', type=basestring),
            title=Field(required=False, default='', type=basestring),
            published=Field(required=True, default=False, type=[True, False]),
            body=Field(type=unicode),
            timestamp=Field(),
        )

>>> p = Post.new()
>>> p
<Post: {'title': u'', 'author': u'jmoiron', 'published': False}>

A few things are going on here. Fields that have a default are initialized to that default whether they are required or not. If a required field does not have a default, it's initialized to None.

Fields can take a type argument, which can either be a callable that takes a value and returns True or False, one or more base types, or one or more values. If one or more types are provided, isinstance is used to test that values are the right type. If one or more values are provided, the Field acts as an enum type, checking that values are in its set of values. If no type is given, validation always passes on a field unless it is required and absent.

If a field in p is given an invalid type, then a ValueError is raised:

>>> p.title = 10
>>> p.save()
Traceback (most recent call last):
  ...
ValueError: Keys did not match spec: ['title']
>>> del p.author
>>> p.save()
Traceback (most recent call last):
  ...
ValueError: Missing fields: ['author'], Invalid fields: ['title']
>>> p.title = 'My first blogpost'
>>> p.author = 'jmoiron'
>>> p.published = True
>>> p.body = u"This is my first blog post..  I'm so excited!"
>>> p.save()

Model.find

For convenience and DRY, Model.find is a classmethod that will use micromongo's cursor to issue a find against the right collection. This method behaves exactly the same as pymongo's Collection.find.

micromongo's slightly modified Cursor class also makes a django-inspired order_by method available to all cursors (find and anything you chain off if it returns a cursor). You can pass one or more field names, with an optional leading '-', to sort things by ascending or descending order.

These changes allow you to use most of the power of pymongo without having to import it, and lets you avoid needless repetition of the location of your data.

field subclassing

You are encouraged to create your own Fields that do what you want. Field subclasses have a hook function pre_validate which take an incoming value and can transform it however they want. Note that this will only work if the fields are actually present; so to get something like an auto_now_add on a DateTimeField, you will want to make it required and have its pre_validate turn None into datetime.datetime.now().