jsonschema-fill-default

Fill a JSON instance in Python with the missing defaults from its JSON Schema Draft 2020-12-valid schema.

from jsonschema_fill_default import fill_default

schema = {
    "properties": {
        "text": {"default": "Hello"},
        "font": {"default": 12},
    }
}

instance = {"text": "Goodbye"}

fill_default(instance, schema)  # Mutates instance!

>>> instance
    {
        "text": "Goodbye",
        "font": 12
    }

[!CAUTION] Filled instances are not automatically validated.

See Load, validate, deference, fill for how you can validate instances and schemas.

Install

jsonschema-fill-default is available on PyPI. You can install using pip:

pip install jsonschema-fill-default

Features

• Fills all missing defaults, including nested ones.

• Uses the first applicable default if multiple defaults exist for a single property.

• Works with the following keywords and any combination thereof (see examples for details):

  • "properties"
  • "allOf"
  • "anyOf"
  • "oneOf"
  • "dependentSchemas"
  • "if-then(-else)"
  • "prefixItems"
  • "items"

[!IMPORTANT]

• The instance must already be valid to its schema.
• The schema itself must be a valid Draft 2020-12 JSON Schema.
• The filled instance is not automatically validated.

Examples

Load, validate, dereference, fill

See unabridged script at examples/load_validate_dereference_fill.py.

import json

from jsonschema import validate, protocols
from jsonref import replace_refs
from jsonschema_fill_default import fill_default


schema_filename = "bicycle.schema.json"
instance = {
    "style": "road",
    "color": "purple",
    "tire": {
        "width": 28
    }
}

with open(schema_filename, 'r') as file:
    schema = json.load(file)

protocols.Validator.check_schema(schema)  # Validate schema
validate(instance, schema)  # Validate instance against schema

schema = replace_refs(schema) # De-reference schema "$refs"

fill_default(instance, schema)  # Fill instance (mutates)

validate(instance, schema)  # Validate filled instance

print(f"\nFilled:\n{json.dumps(instance, indent=4)}")

Nested defaults

from jsonschema_fill_default import fill_default

schema = {
    "properties": {
        "someString": {"default": "The default string"},
        "someObject": {
            "properties": {
                "someNumber": {"default": 3.14},
                "someBoolean": {"default": True}}}}}

instance = {
    "someObject": {
        "someNumber": -1
    }
}

fill_default(instance, schema)

original
    {
        "someObject": {
            "someNumber": -1
        }
    }

filled
    {
        "someString": "The default string",
        "someObject": {
            "someNumber": -1,
            "someBoolean": True
        }
    }

Conditional properties with defaults with "dependentSchemas"

from jsonschema_fill_default import fill_default

schema = {
    "properties": {"some_number": {"default": 100}},
    "dependentSchemas": {
        "some_bool": {
            "properties": {
              "some_string": {"default": "some_bool given"}}}}}

without_bool = {}
with_bool = {"some_bool": False}

fill_default(without_bool, schema)`
fill_default(with_bool, schema)

original  {}
filled    {"some_number": 100}

original
    {
        "some_bool": False
    }
filled
    {
        "some_number": 100,
        "some_bool": False,
        "some_string": "some_bool given"
    }

Conditional defaults with "if-then-else"

from jsonschema_fill_default import fill_default

schema = {
    "if": {
        "required": ["someInteger"]
    },
    "then": {
        "if": {
            "properties": {
                "someInteger": {"multipleOf": 2}
            }
        },
        "then": {"properties": {
            "conditionalString": {"default": "Even integer"}
        }},
        "else": {"properties": {
            "conditionalString": {"default": "Odd integer"}
        }}
    },
    "else": {"properties": {
        "conditionalString": {"default": "someInteger not given"}
    }}
}

none = {}
odd = {"someInteger": 3}
even = {"someInteger": 4}

fill_default(none, schema)
fill_default(odd, schema)
fill_default(even, schema)

original  {}
filled    {"conditionalString": "someInteger not given"}

original  {"someInteger": 3}
filled    {"someInteger": 3, "conditionalString": "Odd integer"}

original  {"someInteger": 4}
filled    {"someInteger": 4, "conditionalString": "Even integer"}

Different properties and defaults with "oneOf"

from jsonschema_fill_default import fill_default

schema = {
    "unevaluatedProperties": False,
    "oneOf": [
        {
            "additionalProperties": False,
            "properties": {
                "food": {"enum": ["cake", "taco"]},
                "price": {"default": 9.95}
            },
            "required": ["food"]
        },
        {
            "additionalProperties": False,
            "properties": {
                "activity": {
                    "enum": ["walk", "talk", "eat"]
                },
                "duration": {
                    "default": 30
                }
            },
            "required": ["activity"]
        }
    ],
}

A = {"food": "cake"}
B = {"activity": "eat"}

fill_default(A, schema)
fill_default(B, schema)

original  {"food": "cake"}
filled    {"food": "cake", "price": 9.95}


original  {"activity": "eat"}
filled    {"activity": "eat", "duration": 30}

Fill array defaults with "prefixItems" and "items"

from jsonschema_fill_default import fill_default

schema = {
    "type": "array",
    "prefixItems": [
        {"type": "number"},
        {"type": "string"},
        {"enum": ["Street", "Avenue", "Drive"], "default": "Drive"}
    ],
    "items": {
        "type": "object",
        "properties": {
            "name": {"type": "string"},
            "age": {"type": "integer", "default": 11}
        },
        "required": ["name"]
    }
}

a = [4, "Privet"]
fill_default(a, schema)

# Missing prefixItems are only filled if there are only default-resolving prefixItem schemas remaining.

original  [4]
filled    [4]

original  [4, "Privet"]
filled    [4, "Privet", "Drive"]


# Existing prefixItems and items are filled

original  [4, "Privet", "Drive",
           {"name": "Harry"},
           {"name": "Dudley"}]
filled    [4, "Privet", "Drive",
           {"name": "Harry", "age": 11},
           {"name": "Dudley", "age": 11}]


original  [1428, "Elm", "Street"]
filled    [1428, "Elm", "Street"]

Developers

Development environment with conda and poetry

I use conda to create a virtual environment with Python, pip, and poetry.

I then add the dependencies using poetry install, which automatically adds them to that conda environment.

Here's how:

1. Clone the repo

2. Create and activate a virtual environment using conda

For example, create and activate a virtual environment env in the root of the project repo using requirements.dev.txt as reference:

cd /root/of/this/repo
conda env create --prefix ./env python=3.9
conda activate ./env
pip install poetry==1.8.5

I don't use an environment.yml to solve and install the conda environment because it's typically slower than just running the above "manual" install.

3. Install poetry dependencies

poetry install

4. Use

Once set up, you can use the development environment in the future by simply activating the conda environment.

If you used the example above, that would be:

cd /root/of/this/repo
conda activate ./env

Paradigms

Use the top-level __init__.py to declare a 'public' API for the module

From this post by reostra:

For example, having

stuff/
  __init__.py
  bigstuff.py
    Stuffinator()
    Stuffinatrix()
  privateStuff.py

where init.py contains

from .bigstuff import Stuffinator, Stuffinatrix

and thereby users can import those with

from stuff import Stuffinator, Stuffinatrix

which essentially says that stuff.Stuffinator and stuff.Stuffinatrix are the only parts of the module intended for public use.

While there's nothing stopping people from doing an 'import stuff.bigstuff.Stuffometer' or 'import stuff.privateStuff.HiddenStuff', they'll at least know they're peeking behind the curtain at that point.

Rather than being implicit, I find it's rather explicit.