Using end as argument to function

So today i noticed something i think is weird.

Say i define an anonyomous function.

f = @(x) ... Now if i input f(end) it will evaluate the function at f(1).

Is this a bug or is there a reason for this behaviour?

Regards Michael

Answer 1

[James Tursa]

Weird, but it does match the doc for end:

"... end = (size(x,k)) when used as part of the kth index ..."

E.g.,

>> f = @(x)x+5
f = 
    @(x)x+5
>> size(f,1)
ans =
     1
>> f(1)
ans =
     6
>> f(end)
ans =
     6

f(end) appears as an indexing expression to MATLAB. Since the size of the f variable is 1x1, it uses 1 for end.

[Stephen23]

It is weird: why does it get interpreted as an indexing operation in the first place?

If a function handle has size 1x1, then why can't we create non-scalar function handles?

[James Tursa]

Seems the parser processes the 'end' first and effectively does a replacement, without regard to whether the variable in question is a function handle or not. It does have a weird effect, however, in that the index itself gets magically turned into an argument for the function handle. E.g., using the same example:

>> f = @(x)x+5
f = 
    @(x)x+5
>> f(end-5)
ans =
     1
>> f(end+10)
ans =
    16

So IMO this behavior should be disallowed and should throw an error. But that may mean the parser would have to be smarter and examine the variable class before doing the replacement. Maybe a bug report or enhancement request to TMW is in order.

[Michael]

Sent a message to TMW. Thanks for the explanations of the phenomenon.