Find the closest index with "True" and calculating the distance (Pandas)-CodePudding

I have a DataFrame like this:

idx	Var1	Var2	Var3
0	True	False	False
1	False	True	False
2	True	False	True
3	False	False	False
4	True	False	True

I'd like to create three new columns with the distance (from each row) of the closest True, and if that row has a True show 0, so I would get this:

idx	Var1	Var2	Var3	distV1	distV2	distV3
0	True	False	False	0	1	2
1	False	True	False	1	0	1
2	True	False	True	0	1	0
3	False	False	False	1	2	1
4	True	False	True	0	3	0

I have read all other discussions related to this topic but haven't been able to find an answer for something like this.

CodePudding user response：

Here is one approach with numpy ops:

for c in df:
    r = np.where(df[c])[0]
    d = abs(df.index.values[:, None] - r)
    df[f'{c}_dist'] = abs(df.index - r[d.argmin(1)])

print(df)

    Var1   Var2   Var3  Var1_dist  Var2_dist  Var3_dist
0   True  False  False          0          1          2
1  False   True  False          1          0          1
2   True  False   True          0          1          0
3  False  False  False          1          2          1
4   True  False   True          0          3          0

CodePudding user response：

Code

from scipy.spatial import KDTree

array = df.to_numpy()
bmp = array.astype(np.uint8)
all_points = np.argwhere(bmp!=2)
true_points = np.argwhere(bmp==1)
distance = tree.query(points, k=1, p=1)[0]
distance.resize(array.shape)
df[[c "_dist" for c in df.columns]] = distance.astype(int)

Output

      Var1   Var2   Var3  Var1_dist  Var2_dist  Var3_dist
idx                                                      
0     True  False  False          0          1          2
1    False   True  False          1          0          1
2     True  False   True          0          1          0
3    False  False  False          1          2          1
4     True  False   True          0          1          0

Explain

Using np.array to make 0,1 data

array([[1, 0, 0],
       [0, 1, 0],
       [1, 0, 1],
       [0, 0, 0],
       [1, 0, 1]], dtype=uint8)

argwhere will return the position coordinate for eligible points.
KDTree is a classical algorithm to find the nearest point.
1. arg k means the top n nearest points
2. arg p=1 means "Manhattan" distance
Which Minkowski p-norm to use.

1 is the sum-of-absolute-values distance ("Manhattan" distance).

2 is the usual Euclidean distance.

Reference

scipy.KDTree