What does [:, :] mean on NumPy arrays

Question

Sorry for the stupid question.

I'm programming in PHP but found some nice code in Python and want to "recreate" it in PHP.

But I'm quite frustrated about the line:

self.h = -0.1    
self.activity = numpy.zeros((512, 512)) + self.h
self.activity[:, :] = self.h

I don't understand what [:, :] means.

I couldn't find an answer by googling it.

Full code

import math
import numpy
import pygame
from scipy.misc import imsave
from scipy.ndimage.filters import gaussian_filter


class AmariModel(object):

    def __init__(self, size):
        self.h = -0.1
        self.k = 0.05
        self.K = 0.125
        self.m = 0.025
        self.M = 0.065

        self.stimulus = -self.h * numpy.random.random(size)
        self.activity = numpy.zeros(size) + self.h
        self.excitement = numpy.zeros(size)
        self.inhibition = numpy.zeros(size)

    def stimulate(self):
        self.activity[:, :] = self.activity > 0

        sigma = 1 / math.sqrt(2 * self.k)
        gaussian_filter(self.activity, sigma, 0, self.excitement, "wrap")
        self.excitement *= self.K * math.pi / self.k

        sigma = 1 / math.sqrt(2 * self.m)
        gaussian_filter(self.activity, sigma, 0, self.inhibition, "wrap")
        self.inhibition *= self.M * math.pi / self.m

        self.activity[:, :] = self.h
        self.activity[:, :] += self.excitement
        self.activity[:, :] -= self.inhibition
        self.activity[:, :] += self.stimulus


class AmariMazeGenerator(object):

    def __init__(self, size):
        self.model = AmariModel(size)

        pygame.init()
        self.display = pygame.display.set_mode(size, 0)
        pygame.display.set_caption("Amari Maze Generator")

    def run(self):
        pixels = pygame.surfarray.pixels3d(self.display)

        index = 0
        running = True
        while running:
            self.model.stimulate()

            pixels[:, :, :] = (255 * (self.model.activity > 0))[:, :, None]
            pygame.display.flip()

            for event in pygame.event.get():
                if event.type == pygame.QUIT:
                    running = False
                elif event.type == pygame.KEYDOWN:
                    if event.key == pygame.K_ESCAPE:
                        running = False
                    elif event.key == pygame.K_s:
                        imsave("{0:04d}.png".format(index), pixels[:, :, 0])
                        index = index + 1
                elif event.type == pygame.MOUSEBUTTONDOWN:
                    position = pygame.mouse.get_pos()
                    self.model.activity[position] = 1

        pygame.quit()


def main():
    generator = AmariMazeGenerator((512, 512))
    generator.run()


if __name__ == "__main__":
    main()

Answer 1

numpy uses tuples as indexes. In this case, this is a detailed slice assignment.

[0]     #means line 0 of your matrix
[(0,0)] #means cell at 0,0 of your matrix
[0:1]   #means lines 0 to 1 excluded of your matrix
[:1]    #excluding the first value means all lines until line 1 excluded
[1:]    #excluding the last param mean all lines starting form line 1 
         included
[:]     #excluding both means all lines
[::2]   #the addition of a second ':' is the sampling. (1 item every 2)
[::]    #exluding it means a sampling of 1
[:,:]   #simply uses a tuple (a single , represents an empty tuple) instead 
         of an index.

It is equivalent to the simpler

self.activity[:] = self.h

(which also works for regular lists as well)

Answer 2

The [:, :] stands for everything from the beginning to the end just like for lists. The difference is that the first : stands for first and the second : for the second dimension.

a = numpy.zeros((3, 3))

In [132]: a
Out[132]: 
array([[ 0.,  0.,  0.],
       [ 0.,  0.,  0.],
       [ 0.,  0.,  0.]])

Assigning to second row:

In [133]: a[1, :] = 3

In [134]: a
Out[134]: 
array([[ 0.,  0.,  0.],
       [ 3.,  3.,  3.],
       [ 0.,  0.,  0.]])

Assigning to second column:

In [135]: a[:, 1] = 4

In [136]: a
Out[136]: 
array([[ 0.,  4.,  0.],
       [ 3.,  4.,  3.],
       [ 0.,  4.,  0.]])

Assigning to all:

In [137]: a[:] = 10

In [138]: a
Out[138]: 
array([[ 10.,  10.,  10.],
       [ 10.,  10.,  10.],
       [ 10.,  10.,  10.]])

Answer 3

This is slice assignment. Technically, it calls¹

self.activity.__setitem__((slice(None,None,None),slice(None,None,None)),self.h)

which sets all of the elements in self.activity to whatever value self.h is storing. The code you have there really seems redundant. As far as I can tell, you could remove the addition on the previous line, or simply use slice assignment:

self.activity = numpy.zeros((512,512)) + self.h

or

self.activity = numpy.zeros((512,512))
self.activity[:,:] = self.h

Perhaps the fastest way to do this is to allocate an empty array and .fill it with the expected value:

self.activity = numpy.empty((512,512))
self.activity.fill(self.h)

^{1Actually, __setslice__ is attempted before calling __setitem__, but __setslice__ is deprecated, and shouldn't be used in modern code unless you have a really good reason for it.}