Definitive answer on the correct str2double behavior for the string '10 20'

Under Windows 7 with R2014a, str2double('10 20') evaluates to NaN. I believe this is the correct behavior, per the documentation.

A colleague working in a Linux environment with R2014a reports that str2double('10 20') evaluates to an array [10 20].

1. Is this possible (without overloading)? 2. Am I correct that the intended evaluation (without overloading) is NaN?

Thanks so much.

Answer 1

[per isakson]

"str2double('10 20') evaluates to an array [10 20]." . If that is the case it should be reported as a bug

1. It should not be possible without overloading
2. You are correct

I advise against unrestricted use of str2num. It uses eval and ... , e.g.

    >> str2num( 'sin([0:0.8:2*pi])')
    ans =
      0    0.7174    0.9996    0.6755   -0.0584   -0.7568   -0.9962   -0.6313

IMO: sscanf is a better alternative

    >> vec = sscanf( '10 20', '%f' )
    vec =
        10
        20

[Nicholas]

Thanks Per.

Answer 2

[Azzi Abdelmalek]

str2num('10 20')

will give an array, maybe that's what your colleague has used

[Nicholas]

Thanks Azzi. Yes, we ended up changing the code to use str2num, to give OS independent results. But I am still curious about: a) the definitive (Mathworks intended) behavior of str2double; and b) whether OS can in fact affect behavior (I don't have a Linux machine here to investigate further).