Skip to main content

Adding marshmallow schemas

Something that we're lacking in our API at the moment is validation. We've done a tiny bit of it with this kind of code:

if (
    "price" not in item_data
    or "store_id" not in item_data
    or "name" not in item_data
):
    abort(
        400,
        message="Bad request. Ensure 'price', 'store_id', and 'name' are included in the JSON payload.",
    )

But there's so much more we can do. For starters, some data points may be optional in some endpoints. We also want to check the data type is correct (i.e. price shouldn't be a string, for example).

To do this kind of checking we can construct a massive if statement, or we can use a library that is made specifically for it.

The marshmallow[^1] library is used to define what data fields we want, and then we can pass incoming data through the validator. We can also go the other way round, and give it a Python object which marshmallow then turns into a dictionary.

Writing the ItemSchema​

Here's the definition of an Item using marshmallow (this is called a schema):

schemas.py
from marshmallow import Schema, fields


class ItemSchema(Schema):
    id = fields.Str(dump_only=True)
    name = fields.Str(required=True)
    price = fields.Float(required=True)
    store_id = fields.Str(required=True)

A couple of weird things maybe!

The id field is a string, but it has the dump_only=True argument. This means that when we use marshmallow to validate incoming data, the id field won't be used or expected. However, when we use marshmallow to serialize data to be returned to a client, the id field will be included in the output.

The other fields will be used for both validation and serialization, and since they have the required=True argument, that means that when we do validation if the fields are not present, an error will be raised.

marshmallow will also check the data type with fields.Float and fields.Int.

Writing the ItemUpdateSchema​

Something that even to do this day sits a bit weird with me is having multiple different schemas for different applications.

When we want to update an Item, we have different requirements than when we want to create an item.

The main difference is that the incoming data to our API when we update an item is different than when we create one. Fields are optional, such that not all item fields should be required. Also, you may not want to allow certain fields at all.

This is the ItemUpdateSchema:

schemas.py
class ItemUpdateSchema(Schema):
    name = fields.Str()
    price = fields.Float()

As you can see, these are not required=True. I've also taken off the id and store_id fields, because:

  • This schema will only be used for incoming data, and we will never receive an id.
  • We don't want clients to be able to change the store_id of an item. If you wanted to allow this, you can add the store_id field here as well.

Writing the StoreSchema​

schemas.py
class StoreSchema(Schema):
    id = fields.Str(dump_only=True)
    name = fields.Str(required=True)

There's not much to explain here! Similar to the ItemSchema, we have id and name since those are the only fields we need for a store.