Why 'str2double' is preferred over 'str2num?

Many times while writting matlab code, matlab forces to use 'str2double' function instead of 'str2num'.

I searched a lot, but could not get a convincing answer. Anyobody here has any idea?

Answer 1

[Walter Roberson]

MATLAB does not force one to use str2double(), but mlint may recommend it.

>> str2num('sqrt(pi)')

ans =

          1.77245385090552

Did you really want people to be able to enter arbitrary text and have it executed by MATLAB ?

[Andrei Bobrov]

>> cs = {'125' '.258' '12.36' '-2.369'}
cs = 
    '125'    '.258'    '12.36'    '-2.369'
>> str2num(cs)
Error using str2num (line 33)
Requires string or character array input.

>> str2double(cs)
ans =
          125        0.258        12.36       -2.369
>>

[Sachin Ganjare]

Thanks Walter & Andrei.

Answer 2

[Daniel Shub]

In this case I think it is best to actually look at the code of the two functions

type str2num
type str2double

One problem with str2num is that it uses EVAL. A malicious user could therefore cause major problems if str2num is used. That said, the use of eval within str2num is one of those cases where it is well encapsulated and unlikely to cause problems (but as I will point out below, can still lead to unexpected outcomes). It will in general be slow, but often that is not a big deal.

The real advantage is that str2double is a lot more powerful. It can handle all sorts of representations of numbers. For example, str2double can handle 1,000, while str2num cannot. In fact, str2num does not give you an error, but rather [1, 0], which might be unexpected.

[Sachin Ganjare]

Thanks Daniel.

Answer 3

[Jan]

STR2MUM evaluates the string using EVAL. Therefore there might be unexpected side effects:

str2num('system(''format C:''); fprintf(''9\n'')');

See: http://www.mathworks.com/matlabcentral/answers/51946-systematic-do-not-use-global-don-t-use-eval

[Sachin Ganjare]

Thank You Jan.