Is there a possibility to use varargin()?

Is there any other way to perform such operation?

c = {rand,rand(4)};
z = @(...) mean(...);
z(c{:})

The same can be achieved using cellfun(), but is it possible only through anonymous function?

[Walter Roberson]

What part would you like to see done in an anonymous function?

Is it the case that you have something that you would like to be able to write something like

z(SomeFunction(SomeArguments){:})

except for the fact that you cannot use {:} on the result of a function call?

[Walter Roberson]

Some day it might be possible to define

C2S = @(C) struct('args', C);
z(C2S(SomeFunctionThatReturnsACell(SomeArguments)).args)

[madhan ravi]

Exactly (https://in.mathworks.com/matlabcentral/answers/477479-is-there-a-possibility-to-use-varargin#comment_738726 ) sir Walter.

For instance blkdiag() can accept comma separated list as an argument likewise it would be lot cooler if any functions could do the same:

z = @(varargin) mean(varargin{:}); % this is just for illustraton purposes
c = {rand,rand(2)};
z(c{:}) % this would return the mean for each c content as a cell for example 

[Bruno Luong]

What's wrong with using CELLFUN? Why you want to replace it with something else that is not designed for?

[madhan ravi]

There's nothing wrong with it, just curious to perform such an operation.

[Bruno Luong]

Well the answer is use CELLFUN

z = @(varargin) cellfun(@mean, varargin, 'unif', 0)

There is no valid reason asking anonymous function to replace cellfun.

I vote this question with -1

[madhan ravi]

The point of the question is not to replace cellfun()! but to just imply varargin in anonymous function.

[Bruno Luong]

Sorry I don't see the point.

[madhan ravi]

Thanks Bruno, I was exactly looking for a way , your comment (https://in.mathworks.com/matlabcentral/answers/477479-is-there-a-possibility-to-use-varargin#comment_738763 ) fullfilled it. How would you apply the same for the second dimension ?

[madhan ravi]

https://in.mathworks.com/matlabcentral/answers/477479-is-there-a-possibility-to-use-varargin#comment_738728 - seems interesting sir Walter.

[Walter Roberson]

"For instance blkdiag() can accept comma separated list as an argument"

Yes, and the calculation depends upon all of the inputs together.

"z(c{:}) % this would return the mean for each c content as a cell for example "

Which is a very different situation than blkdiag(), as the results would be independent of each other. For independent calculations, there is already cellfun if the input is a cell array.

In the case of comma separated lists like for blkdiag() you would need to think about the meaning of calls such as mean(1,2,3) . Is the result to be {mean(1), mean(2), mean(3)}? Or is it to be {mean(1) along the third dimension, mean(2) along the third dimension} ?

What would be a reasonable enhancement for a lot of routines including mean() would be to (for selected functions!) treat FUNCTION(CellArray, ...) as cellfun(@(ENTRY) FUNCTION(ENTRY,varargin{:}), CellArray, 'uniform', 0) . This should not be done for all functions, but it would make sense for a number of them.

[madhan ravi]

Thank you for the reply sir Walter. Perhaps I should have simply asked by reprasing it as, is there any upsides of using varargin in an anonymous function? I have seen some answerers use varargin in an anonymous function which gave an advantage but cannot find any specific examples at the moment.

[Walter Roberson]

Using varargin in an anonymous function is okay.

[madhan ravi]

Thank you, sir Walter :)