MATLAB one-liners

One of the joys of using MATLAB is that it has a stock of matrix functions like diff, sort, all, and so on, that can be combined in all sorts of interesting ways. For example, in a recent question, the challenge was to find a compact code to determine which columns of a matrix A have all elements equal. Matt Tearle came up with this nifty answer:

all(~diff(A))

What are your favorite one-line MATLAB expressions?

Answer 1

[Walter Roberson]

cell2mat(arrayfun(@(K) accumarray(C, F(:,K), [], @mean), 1:size(F,2), 'Uniform', 0))

In response to a cssm question:

 I have a feature matrix, F(m, n) and a cluster vector, C(m, 1).
 Now I want to get the mean of feature in F according to C. Make
 it simple as below:
 F = [2 5; 3 7; 8 4]
 C = [2; 1; 2]
 output should be [3 7; (2+8)/2 (5+4)/2]
 =[3 7; 5 4.5]

[Matt Fig]

ARRAYFUN, and ACUMARRAY? About all that is missing is to somehow work BSXFUN in there too!

[Andrew Newell]

Even though BSXFUN is missing, I decided to accept this one. Very clever, @Walter!

Answer 2

[Matt Tearle]

I'm a huge fan of logical indexing. Expressions like

mean(frogs(wombats > 42))

rock my world.

[Andrew Newell]

Or how about mean(people(suck))?

[Sean de Wolski]

I use the function 'keep' from the FEX which performs the inverse of clear.

I also use the variables 'in', 'out' a lot. So the other day I typed:

'keep out'

which kind of made my day.

[Andrew Newell]

I like it!

Answer 3

[Sean de Wolski]

Here's another from a thread today:

Given a connected components analysis (bwconncomp) and some criteria for objects to meet: remove objects that don't meet that criteria from your binary image:

 I(cell2mat(CC.PixelIdxList(~idx)')) = false;

[Sean de Wolski]

idx can usually be a one line expression from cellfun making this a super-awesome-long-one-liner.

Answer 4

[Matt Fig]

Here is a good one. After already writing a solution to this Question, I stumbled upon this:

groups = mat2cell(A,diff([0;find(diff(A) ~= 1);length(A)]),1);

Pretty Slick.

[Andrew Newell]

Very nice, but Matt would be happier if you used logical indexing.

[Matt Fig]

LOL, LI is tops, but not always necessary or useful.

[Walter Roberson]

Strange, I am sure I wrote code much like that in response to one of the numerous copies of that question. I cannot seem to find it now, though. I had the 1 first, and I used > instead of ~=

[Matt Fig]

I had it stored in my "One liners" file, which is made up of one-liners either I or somebody else on CSSM wrote in response to some question. The file is too big to be much use anymore...

[Matt Tearle]

Matt Fig: "LI is tops, but not always necessary or useful"

An unbeliever! Persecute! Kill the heretic!

[Walter Roberson]

Ah, I found my copy, and it was _not_ a 1 liner. I had used

b=diff(a); %find differences

idx = find([b 2]>1); %find their indexes

cel = mat2cell(a, 1, [idx(1) diff(idx)]); %break up the matrix

Answer 5

[Matt Tearle]

Inspired by something I'm working on right now & your comment to my previous answer...

If you have an n-by-1 structure array people with a field suck (which contains a scalar for each struct element), and you want to find the average:

mean([people.suck])

Extract multiple elements, concatenate, apply function. All in one line.

[Andrew Newell]

I like compact expressions with social commentary.

Answer 6

[Andrew Newell]

Turn a structure array S into a cell array with the names of the fields in the first column:

C = horzcat(fieldnames(S), squeeze(struct2cell(S)))

Answer 7

[Andrew Newell]

Here is a real beauty from a comment by Tim Davis on a guest blog for Loren Shure. Suppose you have three 2D vectors p1, p2 and p3, and you want to know if they are collinear. The shortest solution also happens to be numerically the most reliable:

rank ([p2-p1 ; p3-p1]) < 2

It is also easily generalized to more dimensions!

Answer 8

[Oleg Komarov]

eval('fliplr(['''' 33 33 33 33 33 76 105 118 69 32 109 39 73 ''''])')

MuAhahauHAh!!!

[Andrew Newell]

If you were really evil (or a quitter), you'd do

eval(char([113 117 105 116]))

[Oleg Komarov]

You're the evilest!

[the cyclist]

I used to be an admin on a chess server where "qu" could be used as a shorthand to quit out of the interface. A common prank was to tell newbies that "qu" could be used to display the "quote of the day".

Answer 9

[Andrew Newell]

Here is an interesting way of calculating n rows of Pascal's triangle:

round(expm(diag(1:n-1,-1)))

Now if we could just remove the zeros and center it on the same line!

[Walter Roberson]

sprintf() it and regexrep() on the result, substituting spaces for leading space-zero-space; another regexprep() call could substitute spaces for trailing space-zero-space.

[Doug Eastman]

It's not pretty but just for fun, here's one way to do it:

trimmedTriangle = cell2mat(cellfun(@(x) x(1:size(num2str(expm(diag(1:n-1,-1))),2)),cellfun(@(x,y)[x y],cellfun(@(x) repmat(' ',1,x),num2cell(round(linspace(size(num2str(expm(diag(1:n-1,-1))),2)/2,0,n))'),'UniformOutput',false),regexprep(mat2cell(num2str(expm(diag(1:n-1,-1))),ones(n,1)),' 0',' '),'UniformOutput',false),'UniformOutput',false))

[Andrew Newell]

Wow! Three nested cellfuns! This deserves a separate answer.

Answer 10

[Drew Weymouth]

Read in an image and convert it to a grayscale, double matrix of data range 0..1

im= rgb2gray(double(imread('filename.jpg'))/255);

[Walter Roberson]

or more generally, rgb2gray(im2double(imread('filename.jpg')))

Your code would fail for images that happened to be already double or happened to be uint16.