man3/fpclassify.3

   1 .\" Copyright 2002 Walter Harms (walter.harms@informatik.uni-oldenburg.de)
   2 .\"
   3 .\" %%%LICENSE_START(GPL_NOVERSION_ONELINE)
   4 .\" Distributed under GPL
   5 .\" %%%LICENSE_END
   6 .\"
   7 .\" This was done with the help of the glibc manual.
   8 .\"
   9 .\" 2004-10-31, aeb, corrected
  10 .TH FPCLASSIFY 3  2017-09-15 "" "Linux Programmer's Manual"
  11 .SH NAME
  12 fpclassify, isfinite, isnormal, isnan, isinf \- floating-point
  13 classification macros
  14 .SH SYNOPSIS
  15 .nf
  16 .B #include <math.h>
  17 .PP
  18 .BI "int fpclassify(" x );
  19 .PP
  20 .BI "int isfinite(" x );
  21 .PP
  22 .BI "int isnormal(" x );
  23 .PP
  24 .BI "int isnan(" x );
  25 .PP
  26 .BI "int isinf(" x );
  27 .fi
  28 .PP
  29 Link with \fI\-lm\fP.
  30 .PP
  31 .in -4n
  32 Feature Test Macro Requirements for glibc (see
  33 .BR feature_test_macros (7)):
  34 .in
  35 .PP
  36 .\" I haven't fully grokked the source to determine the FTM requirements;
  37 .\" in part, the following has been tested by experiment.
  38 .ad l
  39 .BR fpclassify (),
  40 .BR isfinite (),
  41 .BR isnormal ():
  42 .RS 4
  43 _ISOC99_SOURCE || _POSIX_C_SOURCE\ >=\ 200112L
  44 .RE
  45 .BR isnan ():
  46 .RS 4
  47 _ISOC99_SOURCE || _POSIX_C_SOURCE\ >=\ 200112L
  48     || _XOPEN_SOURCE
  49     || /* Since glibc 2.19: */ _DEFAULT_SOURCE
  50     || /* Glibc versions <= 2.19: */ _BSD_SOURCE || _SVID_SOURCE
  51 .RE
  52 .BR isinf ():
  53 .RS 4
  54 _ISOC99_SOURCE || _POSIX_C_SOURCE\ >=\ 200112L
  55     || /* Since glibc 2.19: */ _DEFAULT_SOURCE
  56     || /* Glibc versions <= 2.19: */ _BSD_SOURCE || _SVID_SOURCE
  57 .RE
  58 .ad
  59 .SH DESCRIPTION
  60 Floating point numbers can have special values, such as
  61 infinite or NaN.
  62 With the macro
  63 .BI fpclassify( x )
  64 you can find out what type
  65 .I x
  66 is.
  67 The macro takes any floating-point expression as argument.
  68 The result is one of the following values:
  69 .TP 14
  70 .B FP_NAN
  71 .I x
  72 is "Not a Number".
  73 .TP
  74 .B FP_INFINITE
  75 .I x
  76 is either positive infinity or negative infinity.
  77 .TP
  78 .B FP_ZERO
  79 .I x
  80 is zero.
  81 .TP
  82 .B FP_SUBNORMAL
  83 .I x
  84 is too small to be represented in normalized format.
  85 .TP
  86 .B FP_NORMAL
  87 if nothing of the above is correct then it must be a
  88 normal floating-point number.
  89 .PP
  90 The other macros provide a short answer to some standard questions.
  91 .TP 14
  92 .BI isfinite( x )
  93 returns a nonzero value if
  94 .br
  95 (fpclassify(x) != FP_NAN && fpclassify(x) != FP_INFINITE)
  96 .TP
  97 .BI isnormal( x )
  98 returns a nonzero value if
  99 (fpclassify(x) == FP_NORMAL)
 100 .TP
 101 .BI isnan( x )
 102 returns a nonzero value if
 103 (fpclassify(x) == FP_NAN)
 104 .TP
 105 .BI isinf( x )
 106 returns 1 if
 107 .I x
 108 is positive infinity, and \-1 if
 109 .I x
 110 is negative infinity.
 111 .SH ATTRIBUTES
 112 For an explanation of the terms used in this section, see
 113 .BR attributes (7).
 114 .ad l
 115 .TS
 116 allbox;
 117 lbw28 lb lb
 118 l l l.
 119 Interface       Attribute       Value
 120 T{
 121 .BR fpclassify (),
 122 .BR isfinite (),
 123 .BR isnormal (),
 124 .BR isnan (),
 125 .BR isinf ()
 126 T}      Thread safety   MT-Safe
 127 .TE
 128 .ad
 129 .SH CONFORMING TO
 130 POSIX.1-2001, POSIX.1-2008, C99.
 131 .PP
 132 For
 133 .BR isinf (),
 134 the standards merely say that the return value is nonzero
 135 if and only if the argument has an infinite value.
 136 .SH NOTES
 137 In glibc 2.01 and earlier,
 138 .BR isinf ()
 139 returns a nonzero value (actually: 1) if
 140 .I x
 141 is positive infinity or negative infinity.
 142 (This is all that C99 requires.)
 143 .SH SEE ALSO
 144 .BR finite (3),
 145 .BR INFINITY (3),
 146 .BR isgreater (3),
 147 .BR signbit (3)