About isinteger command(confusion)

Hi, by the command of isinteger I can check if it is an integer, however, when defined at first, Matlab assume it is double precision right? So even a=3, isinteger(a) returns 0.

How to solve this problem?

Answer 1

[per isakson]

Matlab represents integers in differnt ways:

• int8, int16, etc. see Christians answer
• with a special use of double, which is called flint. See Floating Points

I use this function to test for flint

    function    isf = isflint( m )
    %   floating double only
        try
            bitand( abs( m ), 1 );
            isf = true;
        catch me
            isf = false;
        end
    end

I picked up the idea from a contribution by the "Pedestrian" in the FEX.

(I stripped off comments and error handling.)

[C Zeng]

Sort of lost of your code. While I am still confused, I use if floor(z)==z or ceil(z)==z to judge if it is indeed an integer.

[per isakson]

I cannot find the reference now, however see: isodd: a pedestrian parity checker

Most important is that Matlab use flint, which is described in Floating Points

Yes, there are many ways to decide if a double represents an integer: floor, round, ceil, rem, mod and bitand. The Pedestrian, alias us, argue that bitand behaves better with overflow - I've forgotten. I trust the Pedestrian.

[C Zeng]

Thanks, Isakson, will take a look.

[Jan]

The problem is to decide, if 1e17 is an integer or not: Because a double can store 16 digits only, this number does not contain any information about the fractional part for reasons of the precision. bitand() detect this and throw an error, while 1e17==round(1e17) replies true, because cropping the fractional part does not influence value.

[C Zeng]

Thanks, I understand it. It is good to know.

Answer 2

[Jan]

Joining the rounding with the checks for overflows:

function isf = isflint(m)
isf = (abs(m) <= bitmax && m == floor(m));

[per isakson]

I'll replace my cryptic isflint.

Answer 3

[ChristianW]

doc isinteger

isinteger(int8(3))

[C Zeng]

Thanks, however, isinteger(int8(2.5)) also returns 1, which I want to avoid.

[ChristianW]

isint = @(x) x==round(x)

[C Zeng]

yes, I used x==round(x) or floor(x). what do you mean by "isint = @(x)"?

[Jan]

@C Zeng: @(x) x==round(x) is an anonymous function. You find an exhaustive description ion the documentation.

[C Zeng]

OK