How do you call an inline or anonymous function within cellfun?

I have three variables (simplifying this and changing variable names for the question since the actual variables are over 200,000 rows long)

A= [1; 2; 3; ...] % numbers that arent necessarily in order but they're unique and there arent gaps
B{1,1}= [1 3]
B{2,1}= [9 11 12]

and so on (the cell array B has varying sizes within each cell)

C= [25 50 75; 
    30 60 90; 
    10 20 30;
    ...] % same length as A since each row is basically an ID and coordinates for each value in A

I want to look at each row of B, find the locations of those B values in A and then extract the corresponding values in C into something that looks like this:

result{1,1}= [ 25 50 75;
               30 60 90];

result{2,1}= basically the values corresponding to those 3 numbers in B{2,1};

If I use a for loop it takes TOO LONG (more than an hour). I can do a cellfun if I define the function in a separate file or by defining it with "function" within the script but I would like to just do an inline function and call that using cellfun. So I saved A, B, and C into a 1xn cell array where each row is a 1x3 cell (A, B, C) so that cellfun would work (single input)

findVal = @(x)x{3}(ismember(x{1},x{2}),:);
coords=cellfun(@findVal, vals, 'UniformOutput',false);

But then I get this error:

Error: "findVal" was previously used as a variable, conflicting with its use here as the name of a function or command.

See "How MATLAB Recognizes Command Syntax" in the MATLAB documentation for details.

if I just call findVal(vals{1,1}) then it gives me what I want. And if I define the function in a different file it works too. Why doesn't this work for inline defined functions?

EDIT: I reverted back to the for loop because I realized cellfun was quite slow. But would still like to know how to use anonymous functions within cellfun

Answer 1

[Matt J]

But would still like to know how to use anonymous functions within cellfun

There's no difference. Since findVal is already a function handle, you need to omit the '@',

coords=cellfun(findVal, vals, 'UniformOutput',false);

But as you mentioned, cellfun is not expected to outperform a for-loop.

[Megna Hari]

oh wow I just assumed all functions being called in cellfun required the @. Good to know!

Answer 2

[Matt J]

I reverted back to the for loop

This might be faster than the for-loop,

allB=[B{:}];
idx=ismember(allB,A);
Blens=cellfun('length',B);
G=repelem(1:numel(Blens),Blens);

rlens=sparse(G,1:numel(G),1)*idx(:);

results=mat2cell( C(allB(idx),:)  ,rlens,size(C,2)  );

[Megna Hari]

That's an interesting way to do it. I'll try it out - thanks!

[Megna Hari]

OMG THANK YOU

This helped reduce my run time by 10 whole minutes.

Instead of 40 minutes it's 30 minutes which is definitely much better!