Yet Another Document Mapper

It’s small and simple ODM for use with MongoDB.


YAMD support MongoDB version 3.x only. MongoDB 2.x is not supported.

Minimal version of python is 3.5.

Quick start

Create the models

from datetime import datetime

import pymongo

from yadm import Database
from yadm import Document, EmbeddedDocument
from yadm import fields
from yadm.serialize import to_mongo

class User(Document):
    __collection__ = 'users'

    name = fields.StringField()
    email = fields.EmailField()

class PostLogItem(EmbeddedDocument):
    op = fields.StringField(choices=['created', 'comment_added'])
    at = fields.DatetimeField()
    data = fields.MongoMapField()

class Post(Document):
    __collection__ = 'posts'

    user = fields.ReferenceField(User)
    created_at = fields.DatetimeField(auto_now=True)
    title = fields.StringField()
    body = fields.StringField()
    log = fields.ListField(fields.EmbeddedDocumentField(PostLogItem))

class Comment(Document):
    __collection__ = 'comments'

    user = fields.ReferenceField(User)
    created_at = fields.DatetimeField(auto_now=True)
    post = fields.ReferenceField(Post)
    text = fields.StringField()

All documents creates from class Document. You can use multiple inheritance.

__collection__ magic attribute setups the collection name for documents of model.

Connect to database

client = pymongo.MongoClient('mongodb://localhost:27017')
db = Database(client, 'blog')

Database object is a wrapper about pymongo or motor Database.

Create documents


user = User(name='Bill', email='bill@galactic.hero')

Just insert document to database.


post = Post()
post.user = user
post.title = 'Small post'
post.body = 'Bla-bla-bla...'
post.log = [PostLogItem(op='created', at=datetime.utcnow())]

You can fill documents as above.

Comment the post

comment = Comment()
comment.user = user = post
comment.text = "RE: Bla-bla-bla..."
db.update_one(post, push={
    'log': to_mongo(PostLogItem(op='comment_added',

We add log item to post’s log. This is very usefull case.



qs = db(Post).find({'title': {'$regex': '^S'}})
assert qs.count() > 0
  1. db(Post) creates the QuerySet object;
  2. find method get the raw-query and return new QuerySet object with updated criteria;
  3. count method make the query to database and return value.
for post in qs:
    assert post.title.startswith('S')

__iter__ method make the find-query and returns the generator of documents.


post = db(Post).find_one({'user':})

find_one get the first finded document.

user = post.user

Get attribute with reference makes the query to referred collection. Warning: N+1 problem! We have a cache in QuerySet object and get one referred document only once for one queryset.


Yes! We have a joins! In userspace…

comments = db(Comment).find({'post':}).sort(('created_at', 1))
for comment in comments.join('user'):
    print(, comment.text)
  1. Create the queryset for comments;
  2. In join we get all documents for qs, get all users with one $in-query and bind to documents.


agg = (db.aggregate(Comment)
       .group(_id='post', count={'$sum': 1})

for item in agg:


1.5.0 (2017-12-31)

  • Experimental asyncio support;
  • Add ReferencesListField
    for lists of references.

1.4.15 (2017-12-27)

  • Add projection argument to get_document
    and reload;
  • Add Document.__default_projection__ attribute.

1.4.14 (2017-11-06)

  • Add EnumField
    for save enum.Enum;
  • Add EnumStateField
    for simple state machines based on enum.Enum.

1.4.13 (2017-10-31)

  • Add QuerySet.batch_size
    method for setup batch size for cursor;
  • Some minor fixes.

1.4.10 (2017-07-07)

  • ReferenceField.from_mongo
    try to get document from primary if not found by default.

1.4.9 (2017-07-06)

  • Add QuerySet.read_primary method
    for simple setup pymongo.read_preference.Primary.

1.4.4 (2017-05-17)

  • Add TimedeltaField for stores durations;
  • Add SimpleEmbeddedDocumentField
    for simply create embedded documents.
class Doc(Document):
    embedded = SimpleEmbeddedDocumentField({
        'i': IntegerField(),
        's': StringField(),

1.4.3 (2017-05-14)

  • Add StaticField for static data.

1.4.2 (2017-04-09)

  • Additional arguments (like write_concern) for write operations;
  • create_fake save the documents with write concern “majority” by default.

1.4.0 (2017-04-05)

  • Drop pymongo 2 support;
  • Additional options for databases and collections;
  • Add Database.get_document;
  • Add TypedEmbeddedDocumentField;
  • reload argument of Database.update_one <yadm.database.Database.update_one>
    must be keyword (may be backward incompotable).

1.3.1 (2017-02-21)

  • Change raw data for Money;

1.3.0 (2017-02-19)

  • Add currency support to Money:
    • Totaly rewrite Money type. Now it is not subclass of Decimal;
    • Add storage for currencies:;

1.2.1 (2017-01-19)

  • Add QuerySet.find_in for $in queries with specified order;

1.2.0 (2016-12-27)

  • Drop MongoDB 2.X support;
  • Objects for update and remove results;
  • Use Faker instead fake-factory;

1.1.4 (2016-08-20)

  • Add some features to :py:module:`Bulk <yadm.bulk>`:
    • Bulk.update_one(document, **kw): method for add update one document in bulk;
    • Bulk.find(query).update(**kw): update many documents by query;
    • Bulk.find(query).upsert().update(**kw): upsert document;
    • Bulk.find(query).remove(**kw): remove documents;

1.1.3 (2016-07-23)

  • Add QuerySet.ids
    method for get only documents id’s from queryset;
  • Add Money.total_cents
    method and Money.total_cents classmethod;

1.1 (2016-04-26)

  • Add cacheing on queryset level and use it for
  • Add mongo aggregation framework support;
  • Add exc argument to
    QuerySet.find_one for raise specified exception if not found;
  • Add multi argument to
  • Deprecate QuerySet.find_one
  • Refactoring.

1.0 (2015-11-14)

  • Change document structure. No more bad BaseDocument.__data__ attribute:
    • BaseDocument.__raw__: raw data from mongo;
    • BaseDocument.__cache__: cached objects, casted with fields;
    • BaseDocument.__changed__: changed objects.
  • Changes api for custom fields:
    • Not more need create field descriptors for every field;
    • prepare_value called only for setattr;
    • to_mongo called only for save objects to mongo;
    • from_mongo called only for load values from BaseDocument.__raw__;
    • Remove Field.default attribute. Use Field.get_default method;
    • Add get_if_not_loaded and get_if_attribute_not_set method;
    • By default raise NotLoadedError if field not loaded from projection;
  • Changes in ReferenceField:
    • Raise BrokenReference if link is bloken;
    • Raise NotBindingToDatabase if document not saved to database;
  • smart_null keyword for Field;
  • Fields in document must be instances (not classes!);
  • Remove ArrayContainer and ArrayContainerField;
  • Remove old MapIntKeysField and MapObjectIdKeysField. Use new MapCustomKeysField;
  • Add Database.update_one method for run simple update query with specified document;
  • Add QuerySet.distinct;
  • serialize.from_mongo now accept not_loaded sequence with filed names who must mark as not loaded, parent and name;
  • serialize.to_mongo do not call FieldDescriptor.__set__;
  • Fakers! Subsystem for generate test objects;
  • Tests now use pytest;
  • And more, and more…