weird behaviour of textscan

this is weird:

>> cell2mat(textscan('5','%d'))*0.5
ans =
        3

doesn't matlab supposed to know to automatically interpret integers as floats when necessary? For example:

>> sscanf('5','%d')*0.5

ans =

      2.5000

Any ideas?

Answer 1

[Friedrich]

Hi,

I would rather say that sscanf behaves weired. The textscan does it correct:

http://www.mathworks.com/help/techdoc/matlab_prog/f2-52790.html#f2-98095

(Integers or integer arrays and scalar double-precision floating-point numbers. This yields a result that has the same data type as the integer operands)

So textscan gives an int, cell2mat keeps it and than its an int multiplied by a double which gives an int like states in the link above

But sscanf returns a double, whereas I would hope to get an int. But this is not the case. This is also documented:

http://www.mathworks.com/help/techdoc/ref/sscanf.html

"Only numeric specifiers, A is numeric. If format includes only 64-bit signed integer specifiers, A is of class int64. Similarly, if format includes only 64-bit unsigned integer specifiers, A is of class uint64. Otherwise, A is of class double. If sizeA is inf or n, then A is a column vector. If the input contains fewer than sizeA elements, MATLAB pads A with zeros"

Since there is no ld or lu specified as format, you get a double back.

So overall everything behaves as it should.

[Yosef]

Thanks for the answer Friedrich! I guess you are right and it works as intended, but in my humble opinion this is a design flaw. This type of obvious inconsistency is a recipe for trouble. It costed me more then two hours of my time this morning.

[Friedrich]

I agree. I would wish that sscanf return an int instead of a double in such a situation.

So your question is answered? Please mark it as answered than.

Thanks,

Friedrich