"Expected a scalar" error and odd if-behavior

Hi,

I'm really confused with "Expected a scalar" error in simulink because I'm (almost) 100% sure that the values are indeed [1x1] scalars. Following matlab function is a small part of my simulink model with several inputs and outputs. I have extracted the function from the model and changed inputs to constant-blocks, but the error still occurs.

inputs check, P, hi and lo are constants with values as seen in figure above, arrays a, array1 and array have sizes [30801x7], [2531x5] and [12655x1], respectively.

Heres the function code:

function [res1,res2,res3,res4,state]= fcn(lo,hi,a,P,array1,array,check)
%#codegen
[r c] = find(array1==P);

if check == 1
    B = [a(lo+1:P+1,1),a(lo+1:P+1,3),a(lo+1:P+1,5),a(lo+1:P+1,7)];
    C = [a(P+2:hi+1,1),a(P+2:hi+1,2),a(P+2:hi+1,4),a(P+2:hi+1,6)];
    A = [B;C];
elseif check == 2
    i = array1(1,c);
    B = [a(lo+1:i,1),a(lo+1:i,3),a(lo+1:i,5),a(lo+1:i,7)];
    C = [a(i+1:hi+1,1),a(i+1:hi+1,2),a(i+1:hi+1,4),a(i+1:hi+1,6)];
    A = [B;C];
elseif check == 4
    max_val = max(array1);
    i = max_val(c);
    B = [a(lo+1:22770,1),a(lo+1:22770,3),a(lo+1:22770,5),a(lo+1:22770,7)];
    C = [a(22770+1:hi+1,1),a(22770+1:hi+1,2),a(22770+1:hi+1,4),a(22770+1:hi+1,6)];
    A = [B;C];
else
    B = [a(lo+1:P,1),a(lo+1:P,3),a(lo+1:P,5),a(lo+1:P,7)];
    C = [a(P+1:hi+1,1),a(P+1:hi+1,2),a(P+1:hi+1,4),a(P+1:hi+1,6)];
    A = [B;C];
end

[res1, i] = min(A(:,3));
res3 = A(i,2);
res4 = A(i,4);
res2 = A(i,1);

if (res2 < P) && (any(res2==array) == 0)
    state = 1;
elseif (res2 < P) && (any(res2==array) == 1)
    state = 2;
elseif (res2 >= P) && (any(res2==array) == 0)
    state = 3;
else
    state = 4;
end

and few example error codes:

Expected a scalar value.  This expression has size [1 x :?].

Function 'MATLAB Function1' (#35.326.327), line 11, column 17:
"i"
Launch diagnostic report.

 ---------------------------------------------

Expected a scalar. Non-scalars are not supported in IF or WHILE statements, or with logical operators.
Instead, use ALL to convert matrix logicals to their scalar equivalents.

Function 'MATLAB Function1' (#35.915.923), line 31, column 5:
"res2 < P"
Launch diagnostic report.

Why do I get these scalar-errors? If I run this in matlab line by line and try these variables with isscalar, I get boolean value 1 (meaning they are in fact scalars)

Also, why do I get the first error about i not being a scalar after elseif check == 2? check has value 1 and therefore should even go inside that elseif operator.

Answer 1

[Walter Roberson]

The code has no reason to know that

[r c] = find(array1==P);

is going to find exactly 1 value, not 0 values and not 2 or more values. Therefore r and c are being created as variable dimension. Then you have

i = array1(1,c);

with c being variable dimension, i is going to have to be variable dimension. So when you have

res2 = A(i,1);

then res2 is going to have to be variable dimension. When you then compare res2 < P that is a variable dimension, which is not permitted in an "if".

You would probably point out that, "But

[res1, i] = min(A(:,3));

and min() always returns a single index, so i is going to be a scalar, not a vector!"

The problem with that is that it ignores that for code generation, the size and data type of a variable is always determined by the first assignment to the variable. You assigned a variable dimension to it first, so it is going to remain variable dimension.

Hint: if you are sure that exactly one value is going to be returned, then

[r_ c_] = find(array1==P);
r = r_(1); c = c_(1);

and now the code knows for sure that r and c are scalars.

[mettala]

Thank you Walter Roberson for your detailed answer!

c = c_(1) did the trick