[Python Interface] Regression: can't properly convert a vector to 2D/3D numpy array

[dpiparo]

Check duplicate issues.

• Checked for duplicates

Description

Issue opened fom a Forum post

Reproducer

import ROOT
import numpy as np

a = ROOT.vector("vector<vector<float>>")() # 3D
a1 = ROOT.vector("vector<float>")() # 2D
a2 = ROOT.vector("float")()
a2.push_back(1.)
a2.push_back(2.)
a2.push_back(3.)
a1.push_back(a2)
a1.push_back(a2)
a1.push_back(a2)
a.push_back(a1)
a.push_back(a1)
a.push_back(a1)

print(a)

print(np.array(a).shape) # 3D
print(np.array(a1).shape) # 2D

The output should be, as given by 6.30.06:

{ { { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f } }, { { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f } }, { { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f } } }
(3, 3, 3)
(3, 3)

However, for 6.34.04 it is:

{ { { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f } }, { { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f } }, { { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f }, { 1.00000f, 2.00000f, 3.00000f } } }
(3,)
(3,)

ROOT version

6.34.04

Installation method

all

Operating system

all

Additional context

No response

[lwpiotr]

Hi. Is anything happening regarding this regression? The inability to convert vectors to numpy arrays probably makes most PyROOT code that reads vectors unusable with recent ROOT versions.

[aaronj0]

Hi,

Thanks for this report! Adding a more minimal reproducer:

import cppyy
import numpy as np
from cppyy.gbl.std import vector

arr = np.random.randint(0, 10, size=(2,3,4,5), dtype=np.int32)
v1 = vector[vector[vector[vector[int]]]](arr)
print(v1) # Prints all the data

print(v1.__array__() ) # Only gives innermost 1D list (higher dimensions are dropped) 
print(v1.__array__().shape) # (5,)
print(len(v1.__array__() )) # 5

I can reproduce this with upstream cppyy (3.5.0), and it stems from the __array__ utility that was added to std::vector's python proxy. As this was new behaviour introduced with the cppyy upgrade in 6.32, this can safely be disabled to fix this bug ( I have opened a PR here: #18208), until a more robust solution can be developed upstream.