pyobjtojson 0.10.1

A Python library that simplifies serializing any Python object to JSON-friendly structures, gracefully handling circular references.

pyobjtojson

A lightweight Python library that simplifies the process of serializing any Python object into a JSON-friendly structure without getting tripped up by circular references. With built-in support for dataclasses, Pydantic (v1 & v2), and standard Python collections, pyobjtojson helps you convert your objects into a cycle-free, JSON-ready format for logging, storage, or data transfer.

Features

• Automatic Circular Reference Detection Detects and replaces cyclical structures with "<circular reference>" to prevent infinite loops.
• Broad Compatibility Works seamlessly with dictionaries, lists, custom classes, dataclasses, and Pydantic models (including both model_dump() from v2 and dict() from v1).
• Extended Standard Types Support Native support for datetime, date, time, UUID, Decimal, bytes, Enum, Path, set, and frozenset.
• JSON-Safe Dictionary Keys Non-string keys (UUID, datetime, Enum, tuples, and other objects) are converted so the returned structure always survives json.dumps.
• Full Type Hints Support Complete type annotations for better IDE autocomplete, type checking with mypy, and improved code documentation.
• Non-Intrusive Serialization No special inheritance or overrides needed. Uses reflection and standard Python methods (__dict__, asdict(), to_dict(), etc.) where available.
• Easy to Integrate Just call obj_to_json() on your data structure. No additional configuration required.

Installation with uv:

uv add pyobjtojson

Quickstart

1. Basic Usage

from pyobjtojson import obj_to_json

# A simple dictionary with lists
data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "nested": {"inner_key": "inner_value"}
}

# obj_to_json returns a JSON-serializable structure (nested dicts, lists and
# primitives), not a JSON string. Pass it to json.dumps() when you need text:
json_obj = obj_to_json(data)

import json
json_text = json.dumps(json_obj)

Output (example):

{
  "key1": "value1",
  "key2": [
    1,
    2,
    3
  ],
  "nested": {
    "inner_key": "inner_value"
  }
}

2. Handling Circular References

from pyobjtojson import obj_to_json

a = {"name": "A"}
b = {"circular": a}
a["b"] = b  # Creates a circular reference

obj_to_json(a, check_circular=True)  # check_circular is True by default.

Output:

{
  "name": "A",
  "b": {
    "circular": "<circular reference>"
  }
}

The marker replaces the first object that is already being serialized on the current path: a is an ancestor of itself, so the reference back to it inside b is cut right there.

3. Working with Dataclasses and Pydantic

from dataclasses import dataclass
from pydantic import BaseModel
from pyobjtojson import obj_to_json

@dataclass
class MyDataClass:
    title: str
    value: int

class MyModel(BaseModel):
    name: str
    age: int

dataclass_instance = MyDataClass(title="Test", value=123)
pydantic_instance = MyModel(name="Alice", age=30)

obj = {
    "dataclass": dataclass_instance,
    "pydantic": pydantic_instance
}

obj_to_json(obj)

Output:

{
  "dataclass": {
    "title": "Test",
    "value": 123
  },
  "pydantic": {
    "name": "Alice",
    "age": 30
  }
}

4. Standard Python Types

pyobjtojson now supports many standard Python types out of the box:

from datetime import datetime, date, time
from decimal import Decimal
from enum import Enum
from pathlib import Path
from uuid import UUID
from pyobjtojson import obj_to_json

class Status(Enum):
    ACTIVE = "active"
    INACTIVE = "inactive"

data = {
    "timestamp": datetime(2024, 1, 15, 10, 30, 45),
    "date": date(2024, 1, 15),
    "time": time(14, 30, 0),
    "id": UUID("12345678-1234-5678-1234-567812345678"),
    "price": Decimal("99.99"),
    "binary": b"Hello",
    "status": Status.ACTIVE,
    "path": Path("/home/user/file.txt"),
    "tags": {"python", "json", "api"}
}

obj_to_json(data)

Output:

{
  "timestamp": "2024-01-15T10:30:45",
  "date": "2024-01-15",
  "time": "14:30:00",
  "id": "12345678-1234-5678-1234-567812345678",
  "price": 99.99,
  "binary": "SGVsbG8=",
  "status": "active",
  "path": "/home/user/file.txt",
  "tags": ["api", "json", "python"]
}

Supported Standard Types:

• datetime, date, time → ISO format strings
• UUID → string representation
• Decimal → float (default) or string (with decimal_as_float=False)
• bytes, bytearray → base64 encoded strings
• Enum → underlying value
• Path → string representation
• set, frozenset → sorted lists

5. Dictionary Keys

JSON object keys must be strings, so pyobjtojson normalizes non-string keys to keep the result compatible with json.dumps:

• str, int, float, bool, and None keys are kept as-is (json.dumps already coerces the non-string primitives to strings itself). The exception is a non-finite float key (inf, -inf, nan), which follows the non_finite policy like any other non-finite float.
• Typed keys such as UUID, datetime, Enum, Decimal, and Path are converted to their natural scalar form (e.g. UUID → string, datetime → ISO string), respecting decimal_as_float.
• Any remaining composite key (a tuple, frozenset, or custom object) is stringified as a last resort.

from uuid import UUID
from pyobjtojson import obj_to_json

data = {
    UUID("12345678-1234-5678-1234-567812345678"): "by uuid",
    (1, 2): "by tuple",
    42: "by int",
}

obj_to_json(data)

Output:

{
  "12345678-1234-5678-1234-567812345678": "by uuid",
  "[1, 2]": "by tuple",
  "42": "by int"
}

Note: If two distinct keys normalize to the same string, the last one wins. Primitive keys are the exception: they pass through unconverted, so a dict mixing 1 and "1" keeps both entries and json.dumps then emits a duplicate "1" key, exactly as it does when given that dict directly. Strict parsers may reject such a document, and json.loads keeps only the last entry.

API Reference

obj_to_json(obj, check_circular=True, decimal_as_float=True, non_finite="null")

Returns a cycle-free structure (nested dictionaries/lists) that is JSON-serializable.

Parameters:

• obj (Any): The object to serialize to JSON-like structures.

• check_circular (bool, optional): If True (default), detect and mark circular references as "<circular reference>".

• decimal_as_float (bool, optional): If True (default), convert Decimal to float. If False, convert to string for high precision.

• non_finite (str, optional): How to represent non-finite floats (inf, -inf, nan), which have no JSON literal. One of:

  • "null" (default): convert to None, matching JavaScript's JSON.stringify.
  • "string": convert to "Infinity", "-Infinity", or "NaN".
  • "keep": leave the float as-is. Note this is not valid JSON and will raise with json.dumps(..., allow_nan=False).

  An unknown value raises ValueError.

Returns:

• dict | list | Any: A JSON-serializable structure.

Non-finite floats

inf, -inf, and nan are valid Python floats but have no representation in the JSON spec. Left untouched they break json.dumps(..., allow_nan=False) and produce the non-standard Infinity/NaN tokens that strict parsers reject. By default pyobjtojson converts them to null so the output is always valid, portable JSON:

from pyobjtojson import obj_to_json

data = {"ratio": float("inf"), "value": float("nan"), "ok": 1.5}

obj_to_json(data)                      # {"ratio": None, "value": None, "ok": 1.5}
obj_to_json(data, non_finite="string") # {"ratio": "Infinity", "value": "NaN", "ok": 1.5}

Type Hints

pyobjtojson is fully typed and passes strict mypy checking. This provides:

• Better IDE Support: Autocomplete and inline documentation
• Type Safety: Catch errors before runtime with mypy
• Clear API: Type annotations serve as documentation

from typing import Any
from pyobjtojson import obj_to_json

# Your IDE will provide autocomplete and type checking
def serialize_data(data: dict[str, Any]) -> Any:
    return obj_to_json(
        data,
        check_circular=True,    # bool
        decimal_as_float=False  # bool
    )

To check types in your project:

mypy your_code.py

Contributing

Contributions, bug reports, and feature requests are welcome! Feel free to open an issue or submit a pull request.

License

MIT License