Modeling¶

Models¶

To create a Model, simply inherit from the Model class and then specify the field types and eventually their descriptors.

Collection¶

Each Model will be linked to its own collection. By default, the collection name will be created from the chosen class name and converted to snake_case. For example a model class named CapitalCity will be stored in the collection named capital_city.

If the class name ends with Model, ODMantic will remove it to create the collection name. For example, a model class named PersonModel will belong in the person collection.

It's possible to customize the collection name of a model by specifying the collection option in the Config class.

from odmantic import Model

class CapitalCity(Model):
    name: str
    population: int

    class Config:
        collection = "city"

Custom model validators¶

Exactly as done with pydantic, it's possible to define custom model validators as described in the pydantic: Root Validators documentation (this apply as well to Embedded Models).

In the following example, we will define a rectangle class and add two validators: The first one will check that the height is greater than the width. The second one will ensure that the area of the rectangle is less or equal to 9.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62

from typing import ClassVar

from pydantic import ValidationError, root_validator

from odmantic import Model


class SmallRectangle(Model):
    MAX_AREA: ClassVar[float] = 9

    length: float
    width: float

    @root_validator
    def check_width_length(cls, values):
        length = values.get("length", 0)
        width = values.get("width", 0)
        if width > length:
            raise ValueError("width can't be greater than length")
        return values

    @root_validator
    def check_area(cls, values):
        length = values.get("length", 0)
        width = values.get("width", 0)
        if length * width > cls.MAX_AREA:
            raise ValueError(f"area is greater than {cls.MAX_AREA}")
        return values


print(SmallRectangle(length=2, width=1))
#> id=ObjectId('5f81e3c073103f509f97e374'), length=2.0, width=1.0

try:
    SmallRectangle(length=2)
except ValidationError as e:
    print(e)
    """
    1 validation error for SmallRectangle
    width
      field required (type=value_error.missing)
    """

try:
    SmallRectangle(length=2, width=3)
except ValidationError as e:
    print(e)
    """
    1 validation error for SmallRectangle
    __root__
      width can't be greater than length (type=value_error)
    """

try:
    SmallRectangle(length=4, width=3)
except ValidationError as e:
    print(e)
    """
    1 validation error for SmallRectangle
    __root__
      area is greater than 9 (type=value_error)
    """

Advanced Configuration¶

The model configuration is done in the same way as with Pydantic models: using a Config class defined in the model body.

Available options:

collection: str

Customize the collection name associated to the model. see this section for more details about default collection naming.

parse_doc_with_default_factories: bool

Wether to allow populating field values with default factories while parsing documents from the database. See this section for more details.

Default: False

title: str (inherited from Pydantic)

Title inferred in the JSON schema.

Default: name of the model class

anystr_strip_whitespace: bool (inherited from Pydantic)

Whether to strip leading and trailing whitespaces for str & byte types.

Default: False

json_encoders: dict (inherited from Pydantic)

Customize the way types used in the model are encoded to JSON.

json_encoders example

For example, in order to serialize datetime fields as timestamp values:

class Event(Model):
    date: datetime

    class Config:
        json_encoders = {
            datetime: lambda v: v.timestamp()
        }

json_loads (inherited from Pydantic)

Function used to decode JSON data

Default: json.loads

json_dumps (inherited from Pydantic)

Function used to encode JSON data

Default: json.dumps

For more details and examples about the options inherited from Pydantic, you can have a look to Pydantic: Model Config

Embedded Models¶

Using an embedded model will store it directly in the root model it's integrated in. On the MongoDB side, the collection will contain the root documents and in inside each of them, the embedded models will be directly stored.

Embedded models are especially useful while building one-to-one or one-to-many relationships.

The creation of an Embedded model is done by inheriting the EmbeddedModel class. You can then define fields exactly as for the regular Models.

One to One¶

In this example, we will model the relation between a country and its capital city. Since one capital city can belong to one and only one country, we can model this relation as a One-to-One relationship. We will use an Embedded Model in this case.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

from odmantic import AIOEngine, EmbeddedModel, Model


class CapitalCity(EmbeddedModel):
    name: str
    population: int


class Country(Model):
    name: str
    currency: str
    capital_city: CapitalCity


countries = [
    Country(
        name="Switzerland",
        currency="Swiss franc",
        capital_city=CapitalCity(name="Bern", population=1035000),
    ),
    Country(
        name="Sweden",
        currency="Swedish krona",
        capital_city=CapitalCity(name="Stockholm", population=975904),
    ),
]

engine = AIOEngine()
await engine.save_all(countries)

Defining this relation is done in the same way as defining a new field. Here, the CapitalCity class will be considered as a field type during the model definition.

The Field descriptor can be used as well for Embedded Models in order to bring more flexibility (default values, Mongo key name, ...).

{
  "_id": ObjectId("5f79d7e8b305f24ca43593e2"),
  "name": "Sweden",
  "currency": "Swedish krona",
  "capital_city": {
    "name": "Stockholm",
    "population": 975904
  }
}
{
  "_id": ObjectId("5f79d7e8b305f24ca43593e1"),
  "name": "Switzerland",
  "currency": "Swiss franc",
  "capital_city": {
    "name": "Bern",
    "population": 1035000
  }
}

await engine.find_one(
    Country, Country.capital_city.name == "Stockholm"
)
#> Country(
#>     id=ObjectId("5f79d7e8b305f24ca43593e2"),
#>     name="Sweden",
#>     currency="Swedish krona",
#>     capital_city=CapitalCity(name="Stockholm", population=975904),
#> )

One to Many¶

Here, we will model the relation between a customer of an online shop and his shipping addresses. A single customer can have multiple addresses but these addresses belong only to the customer's account. He should be allowed to modify them without modifying others addresses (for example if two family members use the same address, their addresses should not be linked together).

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37

from typing import List

from odmantic import AIOEngine, EmbeddedModel, Model


class Address(EmbeddedModel):
    street: str
    city: str
    state: str
    zipcode: str


class Customer(Model):
    name: str
    addresses: List[Address]


customer = Customer(
    name="John Doe",
    addresses=[
        Address(
            street="1757  Birch Street",
            city="Greenwood",
            state="Indiana",
            zipcode="46142",
        ),
        Address(
            street="262  Barnes Avenue",
            city="Cincinnati",
            state="Ohio",
            zipcode="45216",
        ),
    ],
)

engine = AIOEngine()
await engine.save(customer)

As done previously for the One to One relation, defining a One to Many relationship with Embedded Models is done exactly as defining a field with its type being a sequence of Address objects.

{
  "_id": ObjectId("5f79eb116371e09b16e4fae4"),
  "name":"John Doe",
  "addresses":[
    {
      "street":"1757  Birch Street",
      "city":"Greenwood",
      "state":"Indiana",
      "zipcode":"46142"
    },
    {
      "street":"262  Barnes Avenue",
      "city":"Cincinnati",
      "state":"Ohio",
      "zipcode":"45216"
    }
  ]
}

Customization¶

Since the Embedded Models are considered as types by ODMantic, most of the complex type constructs that could be imagined should be supported.

Some ideas which could be useful:

• Combine two different embedded models in a single field using typing.Tuple.

• Allow multiple Embedded model types using a typing.Union type.

• Make an Embedded model not required using typing.Optional.

• Embed the documents in a dictionary (using the typing.Dict type) to provide an additional key-value mapping to the embedded documents.

• Nest embedded documents

Referenced models¶

Embedded models are really simple to use but sometimes it is needed as well to have many-to-one (i.e. multiple entities referring to another single one) or many-to-many relationships. This is not really possible to model those using embedded documents and in this case, references will come handy. Another use case where references are useful is for one-to-one/one-to-many relations but when the referenced model has to exist in its own collection, in order to be accessed on its own without any parent model specified.

Many to One (Mapped)¶

In this part, we will model the relation between books and publishers. Let's consider that each book has a single publisher. In this case, multiple books could be published by the same publisher. We can thus model this relation as a many-to-one relationship.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26

from odmantic import AIOEngine, Model, Reference


class Publisher(Model):
    name: str
    founded: int
    location: str


class Book(Model):
    title: str
    pages: int
    publisher: Publisher = Reference()


hachette = Publisher(name="Hachette Livre", founded=1826, location="FR")
harper = Publisher(name="HarperCollins", founded=1989, location="US")

books = [
    Book(title="They Didn't See Us Coming", pages=304, publisher=hachette),
    Book(title="This Isn't Happening", pages=256, publisher=hachette),
    Book(title="Prodigal Summer", pages=464, publisher=harper),
]

engine = AIOEngine()
await engine.save_all(books)

The definition of a reference field requires the presence of the Reference() descriptor. Once the models are defined, linking two instances is done simply by assigning the reference field of referencing instance to the referenced instance.

{
  "_id": ObjectId("5f7a0dc48a73b20f16e2a364"),
  "founded": 1826,
  "location": "FR",
  "name": "Hachette Livre"
}
{
  "_id": ObjectId("5f7a0dc48a73b20f16e2a365"),
  "founded": 1989,
  "location": "US",
  "name": "HarperCollins"
}

{
  "_id": ObjectId("5f7a0dc48a73b20f16e2a366"),
  "pages": 304,
  "publisher": ObjectId("5f7a0dc48a73b20f16e2a364"),
  "title": "They Didn't See Us Coming"
}
{
  "_id": ObjectId("5f7a0dc48a73b20f16e2a367"),
  "pages": 256,
  "publisher": ObjectId("5f7a0dc48a73b20f16e2a364"),
  "title": "This Isn't Happening"
}
{
  "_id": ObjectId("5f7a0dc48a73b20f16e2a368"),
  "pages": 464,
  "publisher": ObjectId("5f7a0dc48a73b20f16e2a365"),
  "title": "Prodigal Summer"
}

Many to Many (Manual)¶

Here, we will model the relation between books and their authors. Since a book can have multiple authors and an author can be authoring multiple books, we will model this relation as a many-to-many relationship.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

from typing import List

from bson import ObjectId

from odmantic import AIOEngine, Model


class Author(Model):
    name: str


class Book(Model):
    title: str
    pages: int
    author_ids: List[ObjectId]


david = Author(name="David Beazley")
brian = Author(name="Brian K. Jones")

python_cookbook = Book(
    title="Python Cookbook", pages=706, author_ids=[david.id, brian.id]
)
python_essentials = Book(
    title="Python Essential Reference", pages=717, author_ids=[brian.id]
)

engine = AIOEngine()
await engine.save_all((david, brian))
await engine.save_all((python_cookbook, python_essentials))

We defined an author_ids field which holds the list of unique ids of the authors (This id field in the Author model is generated implicitly by default).

Since this multi-reference is not mapped by the ODM, we have to persist the authors manually.

{
  "_id": ObjectId("5f7a37dc7311be1362e1da4e"),
  "name": "David Beazley"
}
{
  "_id": ObjectId("5f7a37dc7311be1362e1da4f"),
  "name": "Brian K. Jones"
}

{
  "_id": ObjectId("5f7a37dc7311be1362e1da50"),
  "title":"Python Cookbook"
  "pages":706,
  "author_ids":[
    ObjectId("5f7a37dc7311be1362e1da4e"),
    ObjectId("5f7a37dc7311be1362e1da4f")
  ],
}
{
  "_id": ObjectId("5f7a37dc7311be1362e1da51"),
  "title":"Python Essential Reference"
  "pages":717,
  "author_ids":[
    ObjectId("5f7a37dc7311be1362e1da4f")
  ],
}

1
2
3
4
5
6
7

book = await engine.find_one(Book, Book.title == "Python Cookbook")
authors = await engine.find(Author, Author.id.in_(book.author_ids))
print(authors)
#> [
#>   Author(id=ObjectId("5f7a37dc7311be1362e1da4e"), name="David Beazley"),
#>   Author(id=ObjectId("5f7a37dc7311be1362e1da4f"), name="Brian K. Jones"),
#> ]