Strange behaviour using double

I'm currently using Matlab 2020b

The doumentation for the double function states that 'Converting a string that does not represent a single numeric value to double will produce a NaN result'. Based on this, I am using isnan(double(string('chars'))) to check whether a user inputted character vector represents a number or not. One particular case I wish to guard against is users entering a ',' instaed of '.'. This works very well, except if the vector in question has a comma followed by exactly 3 numerals, in which case double tells me it is a number. For example, the vector '1.23' will produce NaN, but '1,234' gives the number 1234. Any other number of characters after the comma gives NaN (as far as I can tell).

Does anybody understand why this is and is there a way to get it to do what I actually want?

Thanks

[Torsten]

Can you include the exact code that produces these results ?

[Matt J]

For example, the vector '1.23' will produce NaN

No, it won't. You probably meant something else.

double(string('1.23'))

ans = 1.2300

isnan(ans)

ans = logical
   0

[Stephen23]

"For example, the vector '1.23' will produce NaN"

From the context we can assume that the example uses a comma:

double("1,23")

ans = NaN

[dpb]

How do you know the user didn't intend to enter "1,234" as the value 1234?

Is there instruction that valid MATLAB numeric optional thousands separators are not allowed in this application?

Where is the user input coming from? If interactive, you could set the uieditfield to be 'numeric' instead of 'text' and MATLAB will inform the user when tries to enter a nonallowable character.

Answer 1

[Stephen23]

"Does anybody understand why this is"

The comma is an optional thousands delimiter in english: https://en.wikipedia.org/wiki/Decimal_separator#Digit_grouping

" and is there a way to get it to do what I actually want?"

Nope, accepting optional thousands delimiters is what STR2DOUBLE and DOUBLE (likely for compatibility reasons) both do. There is nothing stopping you from modifying its input data, e.g. replacing the comma with '@'.

Answer 2

[Matt J]

You could replace the commas with some other NaN-triggering character, e.g.,

c={'1.23', '1,234'}

c = 1×2 cell array
    {'1.23'}    {'1,234'}

double(replace(string(c),',','?') )

ans = 1×2

1.2300 NaN

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[Matt J]

Or,

c={'1.23', '1,234'};

isnan(double(string(c))) | contains(c,',')

ans = 1×2 logical array
   0   1