Numpy 数组作为 Pandas 中的一个元素 DataFrame

Question

How do I put a numpy array into an element (single cell) of a Pandas DataFrame? For instance,

Driver  Make  Model  Coordinates
Bob     Ford  Focus  [[1, 0, 1],[1, 2, 3], [2, 0, 2]]
Sally   Ford  Echo   [[0, 0, 1],[0, 2, 0]]

I've tried to store the array on each row, but the documentation doesn't seem to support this.

Context:

I am hoping to use df.to_json() to export the data to a json file, from which the data can later be read into a DataFrame where each row is one of the individuals. Should I be thinking about doing this differently?

Answer 1

Yes, you can. Use .at[] or .iat[] to avoid broadcasting behavior when attempting to put an iterable into a single cell. This also applies to list and set .

The bad thing: It may be quite challenging to do such assignment in an optimized way that does not involve iteration through rows. That said, this is still doable for reasonably-sized arrays. And if you really have to store millions of such arrays, a fundamental redesign may be required. Eg restructure your code, use MongoDB or other storage instruments instead, etc.

import pandas as pd
import numpy as np

# preallocate the output dataframe
df = pd.DataFrame(
    data=np.zeros((2,4), dtype=object),
    columns=["Driver", "Make", "Model", "Coordinates"]
)

# element-wise assignment
df.at[0, "Coordinates"] = np.array([[1, 0, 1],[1, 2, 3], [2, 0, 2]])
df.at[1, "Coordinates"] = np.array([[0, 0, 1],[0, 2, 0]])
# other elements were omitted

Result

print(df)
  Driver Make Model                        Coordinates
0      0    0     0  [[1, 0, 1], [1, 2, 3], [2, 0, 2]]
1      0    0     0             [[0, 0, 1], [0, 2, 0]]

print(df.at[0, "Coordinates"])
[[1 0 1]
 [1 2 3]
 [2 0 2]]

print(type(df.at[0, "Coordinates"]))
<class 'numpy.ndarray'>