MATLAB reshape to Python: the order='F' column-major gotcha
Convert MATLAB reshape to NumPy. MATLAB is column-major and NumPy is row-major, so reshape(A,m,n) needs order='F' to match. The same applies to A(:) flattening.
reshape looks identical — but silently reorders your data
reshape exists in both MATLAB and NumPy with the same name and arguments, which makes it one of the most dangerous functions to port. The signatures match; the element order does not.
MATLAB is column-major (it fills down columns first). NumPy is row-major (it fills across rows first). So the naive translation produces a *different matrix* with no error:
`matlab
% MATLAB — fills column by column
A = [1 2 3 4 5 6];
B = reshape(A, 2, 3)
% B =
% 1 3 5
% 2 4 6
`
`python
# Python — A.reshape(2,3) fills ROW by row → WRONG vs MATLAB
B = A.reshape(2, 3)
# array([[1, 2, 3],
# [4, 5, 6]]) ← not the same matrix
`
Both run. Both give a 2×3 array. Only one matches MATLAB. This is the single most common *silent* bug in a MATLAB-to-Python migration.
The fix: order='F'
Pass order='F' (Fortran / column-major) to make NumPy match MATLAB exactly:
`python
import numpy as np
A = np.array([1, 2, 3, 4, 5, 6])
B = A.reshape(2, 3, order='F')
# array([[1, 3, 5],
# [2, 4, 6]]) ← matches MATLAB
`
Use order='F' whenever you're reshaping data that originated in MATLAB and the *layout* matters — building a matrix from a vector, unrolling an image, packing/unpacking state vectors. If you only ever reshape, compute, and reshape back with the same order, the round-trip is self-consistent and the choice cancels out.
The same trap hides in A(:)
MATLAB's colon flatten A(:) is also column-major. The faithful NumPy equivalent is A.flatten(order='F') (or A.ravel('F')), not the default A.flatten():
`matlab
% MATLAB
A = [1 2; 3 4];
v = A(:) % [1; 3; 2; 4] (down columns)
`
`python
# Python
A = np.array([[1, 2], [3, 4]])
v = A.flatten(order='F') # [1 3 2 4] ✓ matches MATLAB
# A.flatten() → [1 2 3 4] ✗ row-major
`
Anywhere you see A(:) feeding sum, reshape, or a linear index, carry the order='F' through.
When you can ignore order
Column vs row order is irrelevant when:
- The array is 1-D — there's only one axis to traverse. - You created the data in Python and never compared it element-for-element against MATLAB output. - You reshape and reshape back with the same order within a self-contained calculation.
It matters when a specific element must land in a specific position — which is most of the time when you're reproducing a MATLAB result. When in doubt, add order='F' and verify a couple of elements by hand.
Convert your MATLAB code automatically
The [MATLAB-to-Python converter](/convert) translates reshape, indexing, and the rest of your script in one pass. For column-major-sensitive reshapes, double-check the order='F' on the output — it's the one place where "looks right" and "is right" can differ.
Paste your MATLAB into the [converter](/convert) to get a runnable NumPy starting point in seconds.
Free for 50 lines. No account required.
More like this, once a week
New articles on MATLAB-to-Python migration. Short, practical, no fluff — the same tone as the one you just read.