A bug in Matlab comparison? - matlab

Question

I write my own version ismember(find_element) in Matlab to check whether a row array a in matrix b or not, by comparing a with each row in b. If each element in a 'equals' each element of a row in b(if the absolute error is less than Tol), then return logic value 1 and row number.
However, when I test my code, I find that Matlab will return t=[1 1] when compare a=1.0e-11*[0.9063 0.0838] with B=[-1 0](B is the second row of b). Actually it gives correct absolute error error=[1.0000 0.0000], and clearly error(1) is greater than Tol=1e-6. Did I find a bug in Matlab? Or is there any error in my code?
The following is my find_element code:
function [Lia,Locb] = find_element(a,b)
%decide whether a is in b or not(compare each row); if in, return row number
%INPUT:
%a: salar or row array
%b: column array or matrix
%OUTPUT:
%Lia: 1 if a is in b, 0 if not
%Locb: location of a in b, row number if b is a matrix
Tol = 1e-6; %set tolerance when compare elements
Lia = 0;%initialization
Locb = 0;
t = zeros(size(a));%compare result of each element,logical array
for i = 1:size(b,1) %loop through each row
for j = 1:size(b,2) %compare each element in each row
if abs(a(j)-b(i,j))<Tol
t(j) = 1;
end
end
if t %all 1's
Lia = 1;%find a in b
Locb = i;%return row number
%%%%%%%%%%%%%%%%%%%%%%%%
%test outputs
a
B=b(i,:)
error = abs(a-b(i,:))
t
%%%%%%%%%%%%%%%%%%%%%%%%
break
end
end
The test code is :
a = [9.06319429228031e-12 8.37879678833309e-13];
b = [0 1;-1 0;1 0;0 1];
[Lia Locb] = find_element(a,b)
the output is:
a =
1.0e-11 *
0.9063 0.0838
B =
-1 0
error =
1.0000 0.0000
t =
1 1
Lia =
1
Locb =
2

Answer 1

After converting # to %, I find that your code works in Matlab for the following examples. Please provide a clear reproducible answer or edit your question if your problem occurs in Octave.
>> a=1.0e-11*[0.9063 0.0838];
>> b=[-1 0];
>> [Lia, Locb] = find_element(a,b)
Lia =
0
Locb =
0
>> [Lia, Locb] = find_element([-1.000001 0],b)
a =
-1 0
B =
-1 0
error =
1e-06 0
t =
1 1
Lia =
1
Locb =
1

Answer 2

I find the problem in my code now: t need to be initialized before checking each row. By putting t = zeros(size(a)) inside the i loop, it works now.
...
for i = 1:size(b,1) %loop through each row
t = zeros(size(a));%compare result of each element,logical array
for j = 1:size(b,2) %compare each element in each row
....
Test and output:
>> a = [9.06319429228031e-12 8.37879678833309e-13];
>> b = [0 1;-1 0;1 0;0 1];
>> [Lia Locb] = find_element(a,b)
Lia =
0
Locb =
0