How to check if a float is a true number in c++?

Question

How to check if a float number is a true number? That is: it is not infinity, negative infinity, NaN ...

float f;
???

You mean functions like [`std::isnan`](http://en.cppreference.com/w/cpp/numeric/math/isnan)? That's C++11 but does the job. — tadman, Jun 09 '16 at 15:33
Have you seen [std::isfinite](http://en.cppreference.com/w/cpp/numeric/math/isfinite) and the related functions? — James Adkison, Jun 09 '16 at 15:34
you mean a *real* number? Because a number is true if it's different from 0.0 — phuclv, Jun 09 '16 at 15:45
Please [edit](http://stackoverflow.com/posts/37730313/edit) your question to not use `...` and instead fully explain what aspects must be considered to meet your definition of _true_. — James Adkison, Jun 09 '16 at 15:46

score 11 · Accepted Answer · answered Jun 09 '16 at 15:40

11

Simpler than std::fpclassify() is to use std::isfinite()

Determines if the given floating point number arg has finite value i.e. it is normal, subnormal or zero, but not infinite or NaN.

answered Jun 09 '16 at 15:40

eerorika

232,697
12
197
326

score 5 · Answer 2 · edited May 23 '17 at 12:22

5

std::isnormal() does what you want but it also checks for 0.0. So you might check that case extra:

float f;
bool isNumber = (std::isnormal(f) || f == 0.0);

Edit: as pointed out by user2079303 isnormal also returns false for a subnormal number which the OP probably does not want.

However, maybe std::isfinite does the right thing.

float f;
bool isNumber = std::isfinite(f) ;

it returns false for NaN and Inf.

edited May 23 '17 at 12:22

Community

1
1

answered Jun 09 '16 at 15:34

463035818_is_not_an_ai

109,796
11
89
185

3

I don't think you want `&&` here. – n. m. could be an AI Jun 09 '16 at 15:35
3

Are you sure that subnormals are not "*true*" numbers for OP? – eerorika Jun 09 '16 at 15:36
1

@user2079303 to be honest I didnt know what subnormal is until 5 seconds ago and I just ignored it ;) Seems like my answer is wrong then – 463035818_is_not_an_ai Jun 09 '16 at 15:38

Bathsheba · Answer 3 · 2016-06-09T16:00:51.953

1

For C++11 and onwards, use !std::isnan(f) && !std::isinf(f) to test for a number being neither NaN nor infinity (positive or negative).

Pre-C++11 it's a bit more difficult, but you can still do it. If your platform uses IEEE754 for the float then you can use:

f == f will be false only for the NaN case.
Something of the form f == f / 2 will be true only for infinity or a number close to zero, and you can trivially eliminate the latter case.

Boost (www.boost.org) also provides methods for pre C++11. See http://www.boost.org/doc/libs/1_41_0/libs/math/doc/sf_and_dist/html/math_toolkit/utils/fpclass.html

edited Jun 09 '16 at 16:00

answered Jun 09 '16 at 15:34

Bathsheba

231,907
34
361
483

`isnan` returns false for `inf` – 463035818_is_not_an_ai Jun 09 '16 at 15:35
1

You should probably mention that the pre-c++11 methods are guaranteed by IEEE-754 but not by c++ standard. – eerorika Jun 09 '16 at 15:57

score 1 · Answer 4 · answered Jun 09 '16 at 15:34

std::fpclassify() looks like what you are looking for.

int fpclassify( float arg );
int fpclassify( double arg );
int fpclassify( long double arg );
int fpclassify( Integral arg );
Return value

one of FP_INFINITE, FP_NAN, FP_NORMAL, FP_SUBNORMAL, FP_ZERO or implementation-defined type, specifying the category of arg.

How to check if a float is a true number in c++?

4 Answers4