Enforcing a schema

Introduction to MongoDB in Python

Filip Schouwenaars

Machine Learning Researcher

From flexible to validated schema

MongoDB lets you store data without a fixed schema
Great for rapid prototyping and evolving your data model

mov.insert_one({
    "title": "knives out",
    "genre": ["comedy", "crime", "drama"],
    "year": 2019, # oops
    "rating": 7.9,
})


mov.find_one({
  "title": "knives out",
  "release_year": 2019
})

When schema is clear, important to configure validation

Enforce a schema with pydantic

Data validation library
Define expected fields and their types
Blueprint for every document

from pydantic import BaseModel
from typing import Optional

class Movie(BaseModel):
    title: str
    genre: list[str]
    release_year: int
    rating: float
    won_oscar: Optional[bool] = None

Inserting with typed data models

# Before
new_movie = {
  "title": "knives out",
  "genre": ["comedy", "crime", "drama"],
  "year": 2019, # oops
  "rating": 7.9,
}

# No output

No checks on the data format

# Now
new_movie = Movie(
  title = "knives out",
  genre = ["comedy", "crime", "drama"],
  year = 2019, # oops
  rating = 7.9,
)

pydantic.error_wrappers.ValidationError:
1 validation error for Movie
release_year: field required

Typos and missing fields are caught before they make it into the collection!

Fixing our mistake

from pydantic import BaseModel
from typing import Optional

class Movie(BaseModel):
    title: str
    genre: list[str]
    release_year: int
    rating: float
    won_oscar: Optional[bool] = None

# Correct set of fields and field values
new_movie = Movie(
  title = "knives out",
  genre = ["comedy", "crime", "drama"],
  release_year = 2019,
  rating = 7.9,
)


mov.insert_one(dict(new_movie))

InsertOneResult(...)

MongoDB's built-in schema validation

client.film.create_collection(
  "movies_v2",
  validator={
    "$jsonSchema": {
      "required": ["title", "genre", "release_year", "rating"],
      "properties": {
        "title": { "bsonType": "string" },
        "genre": { 
          "bsonType": "array",
          "items": { "bsonType": "string" }
        },
        "release_year": { "bsonType": "int" },
        "rating": { "bsonType": "double" },
        "won_oscar": { "bsonType": "bool" }
      }
    }
  }
)

Testing MongoDB's built-in schema validation

client.film.movies_v2.insert_one({
  "title": "knives out",
  "genre": ["comedy", "crime", "drama"],
  "year": 2019, # oops
  "rating": 7.9,
})

pymongo.errors.WriteError: Document failed validation, [...]
'missingProperties': ['release_year'], 'errmsg': 'Document failed validation'}

Schema validation at the database level
Works across all applications accessing MongoDB

Summary

Application-side validation: pydantic.BaseModel
Database-side validation: MongoDB's built-in schema validation
Prevent mistakes
Enforce structure

Let's practice!

Introduction to MongoDB in Python