I have one piece of code in MATLAB, and I try to translate that code to Python. In MATLAB, I can write this:
x = [1,2,3;4,5,6;7,8,9];
which is just a 3x3 matrix. Then if I use x(1:5), MATLAB will first transfer the matrix x into a 1x9 vector and then return me a 1x5 vector as follows: ans=[1,4,7,2,5];
So could you please tell me that what piece of simple code in Python can have that same outcome?
You can convert your matrix to a numpy array and then use unravel_index to convert your linear indices into subscripts which you can then use to index into your original matrix. Note that all commands below use the 'F' input to use column-major ordering (the default for MATLAB) rather than row-major ordering (the default for numpy)
import numpy as np
a = np.array([[1,2,3],[4,5,6],[7,8,9]])
inds = np.arange(5);
result = a[np.unravel_index(inds, a.shape, 'F')]
# array([1, 4, 7, 2, 5])
Also, if you want to flatten a matrix like MATLAB you can do that as well:
a.flatten('F')
# array([1, 4, 7, 2, 5, 8, 3, 6, 9])
If you are converting a bunch of MATLAB code to python, it's strong recommended to use numpy and look at the documentation on notable differences
An alternative way to directly access the 2D array without making a transformed copy is to use the integer division and the modulo operators.
import numpy as np
# example array
rect_arr = np.array([[1, 2, 3, 10], [4, 5, 6, 11], [7, 8, 9, 12]])
rows, cols = rect_arr.shape
print("Array is:\n", rect_arr)
print(f"rows = {rows}, cols = {cols}")
# Access by Linear Indexing
# Reference:
# https://upload.wikimedia.org/wikipedia/commons/4/4d/Row_and_column_major_order.svg
total_elems = rect_arr.size
# Row major order
print("\nRow Major Sequence:")
for linear_index in range(total_elems):
# do something with rect_arr[linear_index // cols][linear_index % cols]
# Sequence will be 1, 2, 3, 10, 4, 5, 6, 11, 7, 8, 9, 12
print(rect_arr[linear_index // cols][linear_index % cols])
# Columnn major order
print("\nColumn Major Sequence:")
for linear_index in range(total_elems):
# do something with rect_arr[linear_index % rows][linear_index // rows]
# Sequence will be 1, 4, 7, 2, 5, 8, 3, 6, 9, 10, 11, 12
print(rect_arr[linear_index % rows][linear_index // rows])
# With unravel_index
# Row major order
row_indices = range(total_elems)
row_transformed_arr = rect_arr[np.unravel_index(row_indices, rect_arr.shape, "C")]
print(row_transformed_arr)
# Columnn major order
col_indices = range(total_elems)
col_transformed_arr = rect_arr[np.unravel_index(row_indices, rect_arr.shape, "F")]
print(col_transformed_arr)
Useful for plotting in subplots:
# <df> is a date-indexed dataframe with 8 columns containing time-series data
fig, axs = plt.subplots(nrows=4, ncols=2)
rows, cols = axs.shape
# Order plots in row-major
for i, colname in enumerate(df):
df[colname].plot(ax=axs[i // cols][i % cols], title=colname)
plt.show()
# Order plots in column-major
for i, colname in enumerate(df):
df[colname].plot(ax=axs[i % rows][i // rows], title=colname)
plt.show()
I'm not sure what MATLAB's x(1:5) syntax is supposed to do but, according to your desired output, it would seem to be transposing the matrix, flattening it, and then returning a slice. This is how to do that in Python:
>>> from itertools import chain
>>>
>>> x = [[1,2,3],
... [4,5,6],
... [7,8,9]]
>>>
>>> list(chain(*zip(*x)))[0:5]
[1, 4, 7, 2, 5]
A=[1,2;3,4] A(:) would return [1;2;3;4]`. No transposing involved, just plain old column major ordering. What you call "flattening" is apparently what Python does, which is the same procedure but in row-major order, so yes, then you'd have to transpose a matrix to get the same in column-major.
x(1:5)simply takes the first five elements in column-major order, which you can think of like how you described, but MATLAB doesn't make the in-between step.