Hey there! I’ve dealt with this exact scenario before when working with Python.NET and C# interop—totally get how clunky those System.Object[,] matrices can feel when you want to work with them in Pandas. Let me walk you through a few solid approaches depending on your use case:

1. Basic Conversion: Nested List to DataFrame #

This is the most straightforward method, great for small to medium-sized matrices. We’ll first extract values from the C# matrix into a Python nested list, then feed that into Pandas.

import pandas as pd
import clr

# Assume this is the matrix you get from your C# function
csharp_matrix = your_csharp_function_call()

# Get dimensions of the C# 2D array
row_count = csharp_matrix.GetLength(0)
col_count = csharp_matrix.GetLength(1)

# Convert to a Python nested list
python_matrix = [[csharp_matrix[i, j] for j in range(col_count)] for i in range(row_count)]

# Convert to Pandas DataFrame
df = pd.DataFrame(python_matrix)

Notes for this method: #

• Python.NET automatically converts most basic C# types (int, double, string, etc.) to their Python equivalents, so you won’t have to handle those manually.
• If your matrix contains null values from C#, they’ll show up as None in Python, which Pandas will convert to NaN (for numeric columns) or keep as None (for object columns).

2. Faster Conversion for Numeric Matrices: Use NumPy First #

If your matrix is large and contains only numeric values (ints, floats), using NumPy first will be more efficient than building a nested list.

import pandas as pd
import numpy as np
import clr

csharp_matrix = your_csharp_function_call()
row_count = csharp_matrix.GetLength(0)
col_count = csharp_matrix.GetLength(1)

# Initialize a NumPy array (start with object dtype to handle any initial conversions)
np_array = np.empty((row_count, col_count), dtype=object)

# Populate the NumPy array
for i in range(row_count):
    for j in range(col_count):
        np_array[i, j] = csharp_matrix[i, j]

# Convert to a numeric dtype (adjust based on your data: int32, float64, etc.)
np_array = np_array.astype(np.float64)

# Convert to DataFrame
df = pd.DataFrame(np_array)

3. Handling Custom C# Objects in the Matrix #

If your System.Object[,] contains custom C# objects (not basic types), you’ll need to extract the specific properties you want before creating the DataFrame. For example, if each element is a Person class with Name and Age properties:

import pandas as pd
import clr

csharp_matrix = your_csharp_function_call()
row_count = csharp_matrix.GetLength(0)
col_count = csharp_matrix.GetLength(1)

# Option 1: Create a DataFrame with nested dictionaries (each cell is a dict of properties)
python_matrix = [[{"Name": csharp_matrix[i,j].Name, "Age": csharp_matrix[i,j].Age} 
                  for j in range(col_count)] 
                 for i in range(row_count)]
df = pd.DataFrame(python_matrix)

# Option 2: Flatten into a long-format DataFrame (useful if you want row/col indices + properties)
flat_data = []
for i in range(row_count):
    for j in range(col_count):
        person = csharp_matrix[i,j]
        flat_data.append({
            "Row_Index": i,
            "Col_Index": j,
            "Name": person.Name,
            "Age": person.Age
        })
df = pd.DataFrame(flat_data)

Quick Troubleshooting Tips #

• Performance with large matrices: Unfortunately, Python.NET doesn’t have a direct way to cast System.Object[,] to a NumPy array without looping (since it’s an object-type array). The NumPy method is still faster than list comprehensions for big data, though.
• Type mismatches: If you get errors converting to a numeric dtype, check for non-numeric values in your matrix (like strings or nulls). You might need to clean the data first or stick with object dtype in Pandas.

内容的提问来源于stack exchange，提问作者cabo