# numpy-json
[](https://opensource.org/licenses/MIT)
[](https://www.python.org/downloads/)
A robust JSON encoder for NumPy arrays and extended Python data types. No more writing custom serialization wrappers — just drop in `NumpyJSONEncoder` and go.
## Features
**numpy-json** seamlessly handles:
- **NumPy arrays** → JSON lists
- **NumPy scalars** (int, float, bool) → native JSON types
- **NumPy datetime64/timedelta64** → ISO 8601 strings
- **Python datetime/date/time** → ISO 8601 strings
- **UUID** → string representation
- **Decimal** → float (default) or string (configurable)
- **set/tuple** → JSON lists
- **bytes/bytearray** → base64 encoded strings (configurable)
- **pathlib.Path** → string paths
- **pandas types** (optional, no hard dependency) → appropriate JSON types
## Installation
```bash
pip install numpy-json
```
Or install from source:
```bash
git clone https://github.com/featrix/numpy-json.git
cd numpy-json
pip install -e .
```
## Quick Start
```python
import json
import numpy as np
from numpy_json import NumpyJSONEncoder
# Your data with NumPy arrays
data = {
"array": np.array([1, 2, 3, 4, 5]),
"matrix": np.array([[1, 2], [3, 4]]),
"scalar": np.float64(3.14159),
"date": np.datetime64('2025-01-15'),
}
# Encode to JSON - it just works!
json_str = json.dumps(data, cls=NumpyJSONEncoder)
print(json_str)
```
**Output:**
```json
{
"array": [1, 2, 3, 4, 5],
"matrix": [[1, 2], [3, 4]],
"scalar": 3.14159,
"date": "2025-01-15"
}
```
## Usage
### Basic Usage
Simply pass `NumpyJSONEncoder` as the `cls` parameter to `json.dumps()`:
```python
import json
from numpy_json import NumpyJSONEncoder
json_str = json.dumps(your_data, cls=NumpyJSONEncoder)
```
### Configuration Options
The encoder provides class-level configuration:
#### Decimal Handling
```python
from numpy_json import NumpyJSONEncoder
import decimal
# Default: convert Decimal to float (lossy for very precise decimals)
data = {"price": decimal.Decimal("19.99")}
json.dumps(data, cls=NumpyJSONEncoder) # → {"price": 19.99}
# Alternative: preserve as string
NumpyJSONEncoder.DECIMAL_AS_STR = True
json.dumps(data, cls=NumpyJSONEncoder) # → {"price": "19.99"}
```
#### Binary Data Handling
```python
from numpy_json import NumpyJSONEncoder
data = {"binary": b"hello world"}
# Default: base64 encoding
json.dumps(data, cls=NumpyJSONEncoder)
# → {"binary": "aGVsbG8gd29ybGQ="}
# Alternative: emit as integer list
NumpyJSONEncoder.BASE64_BYTES = False
json.dumps(data, cls=NumpyJSONEncoder)
# → {"binary": [104, 101, 108, 108, 111, 32, 119, 111, 114, 108, 100]}
```
### Handling NaN and Infinity
By default, Python's `json` module allows `NaN` and `Infinity` values (which are not valid in RFC 8259 JSON). If you need strict JSON compliance:
```python
import json
import numpy as np
from numpy_json import NumpyJSONEncoder, sanitize_nans
data = {
"valid": np.array([1.0, 2.0, 3.0]),
"invalid": np.array([1.0, np.nan, np.inf, -np.inf]),
}
# Clean the data first
clean_data = sanitize_nans(data)
# NaN/Inf values are replaced with None
# Then encode with allow_nan=False for strict JSON
json_str = json.dumps(clean_data, cls=NumpyJSONEncoder, allow_nan=False)
```
**Result:**
```json
{
"valid": [1.0, 2.0, 3.0],
"invalid": [1.0, null, null, null]
}
```
### Advanced Example
```python
import json
import numpy as np
from datetime import datetime
from uuid import uuid4
from pathlib import Path
from numpy_json import NumpyJSONEncoder
complex_data = {
"id": uuid4(),
"timestamp": datetime.now(),
"measurements": np.array([1.5, 2.3, 3.7, 4.1]),
"matrix": np.random.rand(3, 3),
"metadata": {
"path": Path("/tmp/data.csv"),
"tags": {"ml", "experiment", "2025"},
"coordinates": (40.7128, -74.0060),
},
"config": {
"enabled": np.bool_(True),
"threshold": np.float32(0.85),
"max_items": np.int64(1000),
}
}
json_str = json.dumps(complex_data, cls=NumpyJSONEncoder, indent=2)
print(json_str)
```
## API Reference
### `NumpyJSONEncoder`
**Class Attributes:**
- `DECIMAL_AS_STR` (bool): If `True`, encode `Decimal` as string. Default: `False`
- `BASE64_BYTES` (bool): If `True`, encode bytes as base64. Default: `True`
**Methods:**
- `default(obj)`: Override method that handles type conversion
### `sanitize_nans(obj)`
Recursively replace `NaN`, `Inf`, and `-Inf` values with `None`.
**Parameters:**
- `obj`: Any Python object (dict, list, NumPy array, etc.)
**Returns:**
- Object with all NaN/Inf values replaced with `None`
## Type Conversion Table
| Python/NumPy Type | JSON Output | Notes |
|-------------------|-------------|-------|
| `np.ndarray` | Array | Recursive conversion via `.tolist()` |
| `np.int*` | Number | All NumPy integer types |
| `np.float*` | Number | All NumPy float types |
| `np.bool_` | Boolean | NumPy boolean |
| `np.datetime64` | String | ISO 8601 format |
| `np.timedelta64` | String | Duration string |
| `datetime`/`date`/`time` | String | ISO 8601 format |
| `uuid.UUID` | String | Standard UUID string |
| `decimal.Decimal` | Number or String | Configurable via `DECIMAL_AS_STR` |
| `set`/`tuple` | Array | Converted to list |
| `bytes`/`bytearray` | String or Array | Configurable via `BASE64_BYTES` |
| `pathlib.Path` | String | String representation |
| pandas types | String or `null` | Optional support, no hard dependency |
## Requirements
- Python 3.8+
- NumPy
## Development
### Setup
```bash
git clone https://github.com/featrix/numpy-json.git
cd numpy-json
pip install -e ".[dev]"
```
### Running Tests
```bash
pytest tests/
```
### Code Style
This project uses:
- `black` for code formatting
- `flake8` for linting
- `mypy` for type checking
```bash
black numpy_json/
flake8 numpy_json/
mypy numpy_json/
```
## Contributing
Contributions are welcome! Please feel free to submit a Pull Request.
1. Fork the repository
2. Create your feature branch (`git checkout -b feature/amazing-feature`)
3. Commit your changes (`git commit -m 'Add some amazing feature'`)
4. Push to the branch (`git push origin feature/amazing-feature`)
5. Open a Pull Request
## License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
## Copyright
Copyright (c) 2025 Featrix, Inc.
## Acknowledgments
- Built to solve real-world data serialization challenges in scientific computing and machine learning workflows
- Inspired by the need for seamless NumPy integration with JSON APIs
## Support
For issues, questions, or contributions, please visit the [GitHub repository](https://github.com/featrix/numpy-json).
Raw data
{
"_id": null,
"home_page": "https://github.com/featrix/numpy-json",
"name": "numpy-json",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": "\"Featrix, Inc.\" <info@featrix.com>",
"keywords": "numpy, json, encoder, serialization, array, scientific computing, data science",
"author": "Featrix, Inc.",
"author_email": "\"Featrix, Inc.\" <info@featrix.com>",
"download_url": "https://files.pythonhosted.org/packages/ba/d7/7103a6e64dce6a37ef04c0439b7c10ed843c526bfb108b7be0c0d34868bd/numpy_json-0.1.0.tar.gz",
"platform": null,
"description": "# numpy-json\n\n[](https://opensource.org/licenses/MIT)\n[](https://www.python.org/downloads/)\n\nA robust JSON encoder for NumPy arrays and extended Python data types. No more writing custom serialization wrappers \u2014 just drop in `NumpyJSONEncoder` and go.\n\n## Features\n\n**numpy-json** seamlessly handles:\n\n- **NumPy arrays** \u2192 JSON lists\n- **NumPy scalars** (int, float, bool) \u2192 native JSON types\n- **NumPy datetime64/timedelta64** \u2192 ISO 8601 strings\n- **Python datetime/date/time** \u2192 ISO 8601 strings\n- **UUID** \u2192 string representation\n- **Decimal** \u2192 float (default) or string (configurable)\n- **set/tuple** \u2192 JSON lists\n- **bytes/bytearray** \u2192 base64 encoded strings (configurable)\n- **pathlib.Path** \u2192 string paths\n- **pandas types** (optional, no hard dependency) \u2192 appropriate JSON types\n\n## Installation\n\n```bash\npip install numpy-json\n```\n\nOr install from source:\n\n```bash\ngit clone https://github.com/featrix/numpy-json.git\ncd numpy-json\npip install -e .\n```\n\n## Quick Start\n\n```python\nimport json\nimport numpy as np\nfrom numpy_json import NumpyJSONEncoder\n\n# Your data with NumPy arrays\ndata = {\n \"array\": np.array([1, 2, 3, 4, 5]),\n \"matrix\": np.array([[1, 2], [3, 4]]),\n \"scalar\": np.float64(3.14159),\n \"date\": np.datetime64('2025-01-15'),\n}\n\n# Encode to JSON - it just works!\njson_str = json.dumps(data, cls=NumpyJSONEncoder)\nprint(json_str)\n```\n\n**Output:**\n```json\n{\n \"array\": [1, 2, 3, 4, 5],\n \"matrix\": [[1, 2], [3, 4]],\n \"scalar\": 3.14159,\n \"date\": \"2025-01-15\"\n}\n```\n\n## Usage\n\n### Basic Usage\n\nSimply pass `NumpyJSONEncoder` as the `cls` parameter to `json.dumps()`:\n\n```python\nimport json\nfrom numpy_json import NumpyJSONEncoder\n\njson_str = json.dumps(your_data, cls=NumpyJSONEncoder)\n```\n\n### Configuration Options\n\nThe encoder provides class-level configuration:\n\n#### Decimal Handling\n\n```python\nfrom numpy_json import NumpyJSONEncoder\nimport decimal\n\n# Default: convert Decimal to float (lossy for very precise decimals)\ndata = {\"price\": decimal.Decimal(\"19.99\")}\njson.dumps(data, cls=NumpyJSONEncoder) # \u2192 {\"price\": 19.99}\n\n# Alternative: preserve as string\nNumpyJSONEncoder.DECIMAL_AS_STR = True\njson.dumps(data, cls=NumpyJSONEncoder) # \u2192 {\"price\": \"19.99\"}\n```\n\n#### Binary Data Handling\n\n```python\nfrom numpy_json import NumpyJSONEncoder\n\ndata = {\"binary\": b\"hello world\"}\n\n# Default: base64 encoding\njson.dumps(data, cls=NumpyJSONEncoder) \n# \u2192 {\"binary\": \"aGVsbG8gd29ybGQ=\"}\n\n# Alternative: emit as integer list\nNumpyJSONEncoder.BASE64_BYTES = False\njson.dumps(data, cls=NumpyJSONEncoder)\n# \u2192 {\"binary\": [104, 101, 108, 108, 111, 32, 119, 111, 114, 108, 100]}\n```\n\n### Handling NaN and Infinity\n\nBy default, Python's `json` module allows `NaN` and `Infinity` values (which are not valid in RFC 8259 JSON). If you need strict JSON compliance:\n\n```python\nimport json\nimport numpy as np\nfrom numpy_json import NumpyJSONEncoder, sanitize_nans\n\ndata = {\n \"valid\": np.array([1.0, 2.0, 3.0]),\n \"invalid\": np.array([1.0, np.nan, np.inf, -np.inf]),\n}\n\n# Clean the data first\nclean_data = sanitize_nans(data)\n# NaN/Inf values are replaced with None\n\n# Then encode with allow_nan=False for strict JSON\njson_str = json.dumps(clean_data, cls=NumpyJSONEncoder, allow_nan=False)\n```\n\n**Result:**\n```json\n{\n \"valid\": [1.0, 2.0, 3.0],\n \"invalid\": [1.0, null, null, null]\n}\n```\n\n### Advanced Example\n\n```python\nimport json\nimport numpy as np\nfrom datetime import datetime\nfrom uuid import uuid4\nfrom pathlib import Path\nfrom numpy_json import NumpyJSONEncoder\n\ncomplex_data = {\n \"id\": uuid4(),\n \"timestamp\": datetime.now(),\n \"measurements\": np.array([1.5, 2.3, 3.7, 4.1]),\n \"matrix\": np.random.rand(3, 3),\n \"metadata\": {\n \"path\": Path(\"/tmp/data.csv\"),\n \"tags\": {\"ml\", \"experiment\", \"2025\"},\n \"coordinates\": (40.7128, -74.0060),\n },\n \"config\": {\n \"enabled\": np.bool_(True),\n \"threshold\": np.float32(0.85),\n \"max_items\": np.int64(1000),\n }\n}\n\njson_str = json.dumps(complex_data, cls=NumpyJSONEncoder, indent=2)\nprint(json_str)\n```\n\n## API Reference\n\n### `NumpyJSONEncoder`\n\n**Class Attributes:**\n- `DECIMAL_AS_STR` (bool): If `True`, encode `Decimal` as string. Default: `False`\n- `BASE64_BYTES` (bool): If `True`, encode bytes as base64. Default: `True`\n\n**Methods:**\n- `default(obj)`: Override method that handles type conversion\n\n### `sanitize_nans(obj)`\n\nRecursively replace `NaN`, `Inf`, and `-Inf` values with `None`.\n\n**Parameters:**\n- `obj`: Any Python object (dict, list, NumPy array, etc.)\n\n**Returns:**\n- Object with all NaN/Inf values replaced with `None`\n\n## Type Conversion Table\n\n| Python/NumPy Type | JSON Output | Notes |\n|-------------------|-------------|-------|\n| `np.ndarray` | Array | Recursive conversion via `.tolist()` |\n| `np.int*` | Number | All NumPy integer types |\n| `np.float*` | Number | All NumPy float types |\n| `np.bool_` | Boolean | NumPy boolean |\n| `np.datetime64` | String | ISO 8601 format |\n| `np.timedelta64` | String | Duration string |\n| `datetime`/`date`/`time` | String | ISO 8601 format |\n| `uuid.UUID` | String | Standard UUID string |\n| `decimal.Decimal` | Number or String | Configurable via `DECIMAL_AS_STR` |\n| `set`/`tuple` | Array | Converted to list |\n| `bytes`/`bytearray` | String or Array | Configurable via `BASE64_BYTES` |\n| `pathlib.Path` | String | String representation |\n| pandas types | String or `null` | Optional support, no hard dependency |\n\n## Requirements\n\n- Python 3.8+\n- NumPy\n\n## Development\n\n### Setup\n\n```bash\ngit clone https://github.com/featrix/numpy-json.git\ncd numpy-json\npip install -e \".[dev]\"\n```\n\n### Running Tests\n\n```bash\npytest tests/\n```\n\n### Code Style\n\nThis project uses:\n- `black` for code formatting\n- `flake8` for linting\n- `mypy` for type checking\n\n```bash\nblack numpy_json/\nflake8 numpy_json/\nmypy numpy_json/\n```\n\n## Contributing\n\nContributions are welcome! Please feel free to submit a Pull Request.\n\n1. Fork the repository\n2. Create your feature branch (`git checkout -b feature/amazing-feature`)\n3. Commit your changes (`git commit -m 'Add some amazing feature'`)\n4. Push to the branch (`git push origin feature/amazing-feature`)\n5. Open a Pull Request\n\n## License\n\nThis project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.\n\n## Copyright\n\nCopyright (c) 2025 Featrix, Inc.\n\n## Acknowledgments\n\n- Built to solve real-world data serialization challenges in scientific computing and machine learning workflows\n- Inspired by the need for seamless NumPy integration with JSON APIs\n\n## Support\n\nFor issues, questions, or contributions, please visit the [GitHub repository](https://github.com/featrix/numpy-json).\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "JSON encoder for NumPy arrays and extended Python data types",
"version": "0.1.0",
"project_urls": {
"Bug Tracker": "https://github.com/featrix/numpy-json/issues",
"Documentation": "https://github.com/featrix/numpy-json#readme",
"Homepage": "https://github.com/featrix/numpy-json",
"Repository": "https://github.com/featrix/numpy-json"
},
"split_keywords": [
"numpy",
" json",
" encoder",
" serialization",
" array",
" scientific computing",
" data science"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "1d50657ee74d759fd7b050542f4c4c7252304c6fbb07a732991108b5b0a928d8",
"md5": "f858b9e5d2998e7d63657f09845eb443",
"sha256": "598aba0886791eee94ffe125684a09b8ba3e3079e282959100a609089444a31b"
},
"downloads": -1,
"filename": "numpy_json-0.1.0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "f858b9e5d2998e7d63657f09845eb443",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.8",
"size": 6982,
"upload_time": "2025-10-08T02:15:14",
"upload_time_iso_8601": "2025-10-08T02:15:14.659697Z",
"url": "https://files.pythonhosted.org/packages/1d/50/657ee74d759fd7b050542f4c4c7252304c6fbb07a732991108b5b0a928d8/numpy_json-0.1.0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "bad77103a6e64dce6a37ef04c0439b7c10ed843c526bfb108b7be0c0d34868bd",
"md5": "b7601811277ebbf6f58df020d3a2be13",
"sha256": "59f1e5d9247d4796af12fcf7c8e18c2a7bdcc084567d402aa6f71674b0c4d86f"
},
"downloads": -1,
"filename": "numpy_json-0.1.0.tar.gz",
"has_sig": false,
"md5_digest": "b7601811277ebbf6f58df020d3a2be13",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 7898,
"upload_time": "2025-10-08T02:15:15",
"upload_time_iso_8601": "2025-10-08T02:15:15.954534Z",
"url": "https://files.pythonhosted.org/packages/ba/d7/7103a6e64dce6a37ef04c0439b7c10ed843c526bfb108b7be0c0d34868bd/numpy_json-0.1.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-10-08 02:15:15",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "featrix",
"github_project": "numpy-json",
"github_not_found": true,
"lcname": "numpy-json"
}
JSON
