How to Resolve Python "TypeError: Object of type ndarray is not JSON serializable"

When working with data analysis libraries like NumPy and Pandas in Python, you frequently need to serialize your data, often into JSON format for APIs, storage, or web applications. However, attempting to directly serialize NumPy arrays (ndarray) or Pandas DataFrames using Python's standard json.dumps() function leads to errors like TypeError: Object of type ndarray is not JSON serializable or TypeError: Object of type DataFrame is not JSON serializable. This occurs because the default JSON encoder doesn't natively understand these complex data structures.

This guide explains why these errors happen and provides standard solutions using built-in methods and custom encoders.

Understanding the Error: JSON Serializable Types

Python's built-in json library can directly serialize standard Python types into their JSON equivalents:

Specialized objects from external libraries like NumPy's ndarray or Pandas' DataFrame are not on this list. The default json.dumps() function doesn't know how to represent their internal structure using standard JSON types, hence the TypeError.

Error 1: TypeError: Object of type ndarray is not JSON serializable

Cause: json.dumps() Can not Handle ndarray

You are trying to pass a NumPy array directly into json.dumps().

import json
import numpy as np

my_array = np.array([10, 20, 30, 40])
data = {'id': 'A123', 'values': my_array}

print(f"Type of data['values']: {type(data['values'])}")

try:
  json_string = json.dumps(data)
  print(json_string)
except TypeError as e:
  print(e)

Solution 1: Convert ndarray to list using .tolist() (Recommended)

The simplest and most common solution is to convert the NumPy array into a standard Python list before serialization using the array's .tolist() method. Lists are JSON serializable.

import json
import numpy as np

my_array = np.array([10, 20, 30, 40])
data = {'id': 'A123', 'values': my_array.tolist()} # ✅ Convert here

print(f"Type of data['values'] now: {type(data['values'])}")

json_string = json.dumps(data)
print(f"Serialized JSON: {json_string}")

if json_string:
    loaded_data = json.loads(json_string)
    recreated_array = np.array(loaded_data['values'])
    print(f"Recreated NumPy array: {recreated_array}")
    print(f"Type: {type(recreated_array)}")

This is generally the preferred method due to its simplicity and directness.

Solution 2: Use a Custom JSONEncoder

For more complex scenarios or to handle various NumPy types centrally, create a custom encoder.

import json
import numpy as np

class NumpyEncoder(json.JSONEncoder):
    """ Custom encoder for numpy data types """
    def default(self, obj):
        if isinstance(obj, np.integer):
            return int(obj) # Convert NumPy int to Python int
        elif isinstance(obj, np.floating):
            return float(obj) # Convert NumPy float to Python float
        elif isinstance(obj, np.ndarray):
            return obj.tolist() # Convert ndarray to Python list
        else:
            return super(NumpyEncoder, self).default(obj) # Default handling

my_array = np.array([1.1, 2.2, 3.3])
data = {'id': 'B456', 'count': np.int64(5), 'measurements': my_array}

json_string = json.dumps(data, cls=NumpyEncoder)
print(f"Serialized JSON (Custom Encoder): {json_string}")

This encoder handles basic NumPy integer/float types and arrays by converting them to their Python equivalents.

Solution 3: Use the default Argument in json.dumps()

Provide a function to the default argument that handles the conversion for specific types.

import json
import numpy as np

def numpy_serializer(obj):
    """Function to serialize NumPy types for json.dumps default."""
    if isinstance(obj, np.ndarray):
        return obj.tolist()
    return obj # Or let default handle/error out

my_array = np.array([5, 10, 15])
data = {'run': 1, 'samples': my_array}

json_string = json.dumps(data, default=numpy_serializer)
print(f"Serialized JSON (default arg): {json_string}")

This is similar to the custom encoder but might be simpler for one-off conversions.

Solution 4: Using Pandas Series.to_json() (If using Pandas)

If your array is part of a Pandas workflow, you can convert it to a Series and use its to_json() method.

import numpy as np
import pandas as pd # Requires pandas installation

my_array = np.array([100, 200, 300])

json_string = pd.Series(my_array).to_json(orient='values')

print(f"Serialized JSON (Pandas Series): {json_string}")

This is specific to workflows already involving Pandas.

Error 2: TypeError: Object of type DataFrame is not JSON serializable

Cause: json.dumps() Can not Handle DataFrame

Similar to NumPy arrays, Pandas DataFrames are complex objects that the standard json.dumps doesn't know how to serialize.

import json
import pandas as pd # Requires pandas installation

df = pd.DataFrame(
    {
        "ID": ["X1", "X2", "X3"],
        "Value": [10.5, 12.1, 13.8],
        "Status": ["OK", "OK", "Error"],
    }
)
print("Original DataFrame:")
print(df)
print(f"Type of df: {type(df)}")

try:
  json_string = json.dumps({'data': df})
except TypeError as e:
  print(f"Error: {e}")

Solution 1: Use Pandas .to_json() Method (Recommended)

Pandas DataFrames have their own built-in .to_json() method designed specifically for this purpose. It offers various formatting options via the orient parameter.

import pandas as pd

df = pd.DataFrame(
    {
        "ID": ["X1", "X2", "X3"],
        "Value": [10.5, 12.1, 13.8],
        "Status": ["OK", "OK", "Error"],
    }
)

json_string_cols = df.to_json()
print(f"Serialized JSON (orient='columns'): {json_string_cols}")

json_string_records = df.to_json(orient='records')
print(f"Serialized JSON (orient='records'): {json_string_records}")

json_string_split = df.to_json(orient='split')
print(f"Serialized JSON (orient='split'): {json_string_split}")

print(f"Type of result: {type(json_string_cols)}") # Output: <class 'str'>

Choose the orient value that produces the JSON structure most suitable for your needs. This is the standard way to serialize DataFrames.

Solution 2: Convert DataFrame to dict using .to_dict()

You can convert the DataFrame to a Python dictionary using .to_dict() (which also has an orient parameter) and then serialize that dictionary using json.dumps().

import json
import pandas as pd

df = pd.DataFrame(
    {
        "ID": ["X1", "X2", "X3"],
        "Value": [10.5, 12.1, 13.8],
    }
)

data_dict = df.to_dict()
print(f"DataFrame converted to dict: {data_dict}")

json_string_from_dict = json.dumps(data_dict)
print(f"Serialized JSON (from dict): {json_string_from_dict}")

This works because the standard json library can serialize Python dictionaries. However, using the DataFrame's native .to_json() method (Solution 1) is usually more direct.

Conclusion

The TypeError indicating that NumPy ndarray or Pandas DataFrame objects are not JSON serializable arises because Python's standard json.dumps only handles built-in types.

• For NumPy arrays (ndarray): The best practice is to convert them to Python lists using .tolist() before calling json.dumps(). Custom encoders or the default argument are alternatives.
• For Pandas DataFrames: The recommended approach is to use the DataFrame's own .to_json() method, selecting the orient option that fits your desired JSON output structure. Converting to a dictionary first using .to_dict() is also possible.

By applying these appropriate conversion methods, you can successfully serialize your NumPy and Pandas data structures into JSON format.