Python/Pydantic - using a list with json objects

Question

I have a working model to receive a json data set using pydantic. The model data set looks like this:

data = {'thing_number': 123, 
        'thing_description': 'duck',
        'thing_amount': 4.56}

What I would like to do is have a list of json files as the data set and be able to validate them. Ultimately the list will be converted to records in pandas for further processing. My goal is to validate an arbitrarily long list of json entries that looks something like this:

bigger_data = [{'thing_number': 123, 
                'thing_description': 'duck',
                'thing_amount': 4.56}, 
               {'thing_number': 456, 
                'thing_description': 'cow',
                'thing_amount': 7.89}]

The basic setup I have now is as follows. Note that adding the class ItemList is part of the attempt to get the arbitrary length to work.

from typing import List
from pydantic import BaseModel
from pydantic.schema import schema
import json

class Item(BaseModel):
    thing_number: int
    thing_description: str
    thing_amount: float

class ItemList(BaseModel):
    each_item: List[Item]

The basic code will then produce what I think I'm looking for in an array object that will take Item objects.

item_schema = schema([ItemList])
print(json.dumps(item_schema, indent=2)) 

    {
      "definitions": {
        "Item": {
          "title": "Item",
          "type": "object",
          "properties": {
            "thing_number": {
              "title": "Thing_Number",
              "type": "integer"
            },
            "thing_description": {
              "title": "Thing_Description",
              "type": "string"
            },
            "thing_amount": {
              "title": "Thing_Amount",
              "type": "number"
            }
          },
          "required": [
            "thing_number",
            "thing_description",
            "thing_amount"
          ]
        },
        "ItemList": {
          "title": "ItemList",
          "type": "object",
          "properties": {
            "each_item": {
              "title": "Each_Item",
              "type": "array",
              "items": {
                "$ref": "#/definitions/Item"
              }
            }
          },
          "required": [
            "each_item"
          ]
        }
      }
    }

The setup works on a singe json item being passed:

item = Item(**data)                                                      

print(item)

Item thing_number=123 thing_description='duck' thing_amount=4.56

But when I try and pass the single item into the ItemList model it returns an error:

item_list = ItemList(**data)

---------------------------------------------------------------------------
ValidationError                           Traceback (most recent call last)
<ipython-input-94-48efd56e7b6c> in <module>
----> 1 item_list = ItemList(**data)

/opt/conda/lib/python3.7/site-packages/pydantic/main.cpython-37m-x86_64-linux-gnu.so in pydantic.main.BaseModel.__init__()

/opt/conda/lib/python3.7/site-packages/pydantic/main.cpython-37m-x86_64-linux-gnu.so in pydantic.main.validate_model()

ValidationError: 1 validation error for ItemList
each_item
  field required (type=value_error.missing)

I've also tried passing bigger_data into the array thinking that it would need to start as a list. that also returns an error - - Although, I at least have a better understanding of the dictionary error I can't figure out how to resolve.

item_list2 = ItemList(**data_big)

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-100-8fe9a5414bd6> in <module>
----> 1 item_list2 = ItemList(**data_big)

TypeError: MetaModel object argument after ** must be a mapping, not list

Thanks.

Other Things I've Tried

I've tried passing the data into the specific key with a little more luck (maybe?).

item_list2 = ItemList(each_item=data_big)

---------------------------------------------------------------------------
ValidationError                           Traceback (most recent call last)
<ipython-input-111-07e5c12bf8b4> in <module>
----> 1 item_list2 = ItemList(each_item=data_big)

/opt/conda/lib/python3.7/site-packages/pydantic/main.cpython-37m-x86_64-linux-gnu.so in pydantic.main.BaseModel.__init__()

/opt/conda/lib/python3.7/site-packages/pydantic/main.cpython-37m-x86_64-linux-gnu.so in pydantic.main.validate_model()

ValidationError: 6 validation errors for ItemList
each_item -> 0 -> thing_number
  field required (type=value_error.missing)
each_item -> 0 -> thing_description
  field required (type=value_error.missing)
each_item -> 0 -> thing_amount
  field required (type=value_error.missing)
each_item -> 1 -> thing_number
  field required (type=value_error.missing)
each_item -> 1 -> thing_description
  field required (type=value_error.missing)
each_item -> 1 -> thing_amount
  field required (type=value_error.missing)

score 24 · Answer 1 · answered Oct 31 '19 at 03:55

24

To avoid having "each_item" in the ItemList, you can use the __root__ Pydantic keyword:

from typing import List
from pydantic import BaseModel

class Item(BaseModel):
    thing_number: int
    thing_description: str
    thing_amount: float

class ItemList(BaseModel):
    __root__: List[Item]    # ⯇-- __root__

To build the item_list:

just_data = [
    {"thing_number": 123, "thing_description": "duck", "thing_amount": 4.56},
    {"thing_number": 456, "thing_description": "cow", "thing_amount": 7.89},
]
item_list = ItemList(__root__=just_data)

a_json_duck = {"thing_number": 123, "thing_description": "duck", "thing_amount": 4.56}
item_list.__root__.append(a_json_duck)

The web-frameworks supporting Pydantic often jsonify such ItemList as a JSON array without intermediate __root__ keyword.

answered Oct 31 '19 at 03:55

oHo

51,447
27
165
200

For my own understanding, does the `__root__` effectively change the 'root' character of the `ItemList` to those item in `Item`? Whereas, using `each_item` effectively creates a thing inside `ItemList`? Thanks. – Bill Armstrong Nov 01 '19 at 04:11
2

The [docs](https://pydantic-docs.helpmanual.io/usage/models/#custom-root-types) lists this as a use case, so I prefer this, _although_ it feels slighly un-pythonic to ask users to use `__root__` keyword. Advantage of this method over the [other answer](https://stackoverflow.com/a/58081073/2670584) is that `ItemList.json()` returns the expected JSON structure. – Kristoffer Bakkejord Feb 24 '20 at 15:27
Unfortunately, when appending items to the __root__ (as done at the end of this answer), there is no validation of these items. They are simply appended. If (like in your case), the items in the list are pydantic models that might require validation, you need to trigger this yourself (e.g. using Item.validate(...)). – camo Nov 24 '20 at 19:01
For information, if you want to iterate over the `__root__` list or access items by an index -- you have to implement `__iter__` and `__getitem__` methods in the class. – Dmitry Apr 05 '21 at 09:06

ericbn · Answer 2 · 2021-10-12T14:54:44.333

20

The following also works, and does not require a root type.

To convert from a List[dict] to a List[Item]:

items = parse_obj_as(List[Item], bigger_data)

To convert from JSON str to a List[Item]:

items = parse_raw_as(List[Item], bigger_data_json)

To convert from a List[Item] to a JSON str:

bigger_data_json = json.dumps(items, default=pydantic_encoder)

or with a custom encoder:

def custom_encoder(**kwargs):
    def base_encoder(obj):
        if isinstance(obj, BaseModel):
            return obj.dict(**kwargs)
        else:
            return pydantic_encoder(obj)
    return base_encoder


bigger_data_json = json.dumps(items, default=custom_encoder(by_alias=True))

edited Oct 12 '21 at 14:54

answered Oct 08 '21 at 01:35

ericbn

10,163
3
47
55

10

I found this really useful. For others, the import for `pydantic_encoder` is: `from pydantic.json import pydantic_encoder` . – lukecampbell Oct 13 '21 at 17:05
1

This saved the day for me. It was exactly what I needed, without doing this, I received a non-serializable datamodel error. – RodriKing Nov 10 '21 at 13:20

score 13 · Accepted Answer · answered Sep 24 '19 at 13:09

from typing import List
from pydantic import BaseModel
import json


class Item(BaseModel):
    thing_number: int
    thing_description: str
    thing_amount: float


class ItemList(BaseModel):
    each_item: List[Item]

Base on your code with each_item as a List of Item

a_duck = Item(thing_number=123, thing_description="duck", thing_amount=4.56)
print(a_duck.json())

a_list = ItemList(each_item=[a_duck])

print(a_list.json())

Generate the following output:

{"thing_number": 123, "thing_description": "duck", "thing_amount": 4.56}
{"each_item": [{"thing_number": 123, "thing_description": "duck", "thing_amount": 4.56}]}

using these as "entry json":

a_json_duck = {"thing_number": 123, "thing_description": "duck", "thing_amount": 4.56}
a_json_list = {
    "each_item": [
        {"thing_number": 123, "thing_description": "duck", "thing_amount": 4.56}
    ]
}

print(Item(**a_json_duck))
print(ItemList(**a_json_list))

Work just fine and generates:

Item thing_number=123 thing_description='duck' thing_amount=4.56
ItemList each_item=[<Item thing_number=123 thing_description='duck' thing_amount=4.56>]

We are just left with the only datas:

just_datas = [
    {"thing_number": 123, "thing_description": "duck", "thing_amount": 4.56},
    {"thing_number": 456, "thing_description": "cow", "thing_amount": 7.89},
]
item_list = ItemList(each_item=just_datas)
print(item_list)
print(type(item_list.each_item[1]))
print(item_list.each_item[1])

Those works as expected:

ItemList each_item=[<Item thing_number=123 thing_description='duck'thing_amount=4.56>,<Item thin…
<class '__main__.Item'>
Item thing_number=456 thing_description='cow' thing_amount=7.89

So in case i'm missing something the pydantic librairy works as expected.

My pydantic version : 0.30 python 3.7.4

Reading from a lookalike file:

json_data_file = """[
{"thing_number": 123, "thing_description": "duck", "thing_amount": 4.56},
{"thing_number": 456, "thing_description": "cow", "thing_amount": 7.89}]"""

from io import StringIO
item_list2 = ItemList(each_item=json.load(StringIO(json_data_file)))

Work also fine.

I spent hours thinking the issue was the class/object structure - - not how i was loading the information. Works perfect. Thanks. — Bill Armstrong, Sep 25 '19 at 04:00

score 1 · Answer 4 · answered Jul 13 '22 at 13:56

1

what did the trick for me was fastapi.encoders.jsonable_encoder (take a look at https://fastapi.tiangolo.com/tutorial/encoder/)

So in your case I have appended the "single" items to a list result i.e. result.append(Item(thing_number=123, thing_description="duck", thing_amount=4.56))

and finally fastapi.JSONResponse(content=fastapi.encoders.jsonable_encoder(result))

answered Jul 13 '22 at 13:56

Kots

486
1
5
21

1

The question is asked in a Pydantic context. Not everyone that uses pydantic uses it in a FastAPI context... – acidjunk Sep 16 '22 at 21:28

Python/Pydantic - using a list with json objects

4 Answers4