Math::BigFloat::fcmp() incorrect for +0E+0

[p5pRT]

Migrated from rt.perl.org#1880 (status was 'resolved')

Searchable as RT1880$

[p5pRT]

From @tamias

Math​::BigFloat​::fcmp() incorrectly determines that 0 is greater than any
positive non-integer. This occurs because in BigFloat's representation of
0 (+0E+0) has an exponent of 0, while its representation of a non-integer
such as 0.9 (+9E-1) has a negative exponent. fcmp() compares the exponents
without noting that one of the operands is equal to 0.

# compare 2 values returns one of undef, <0, =0, >0
# returns undef if either or both input value are not numbers
sub fcmp #(fnum_str, fnum_str) return cond_code
{
  local($x, $y) = (fnorm($_[$[]),fnorm($_[$[+1]));
  if ($x eq "NaN" || $y eq "NaN") {
  undef;
  } else {
  ord($y) <=> ord($x)
  ||
  ( local($xm,$xe,$ym,$ye) = split('E', $x."E$y"),
  (($xe <=> $ye) * (substr($x,$[,1).'1') # <=== BUG!
  || Math​::BigInt​::cmp($xm,$ym))
  );
  }
}

I note the following in Math​::BigFloat's documentation​:

  BUGS

  The current version of this module is a preliminary version of the
  real thing that is currently (as of perl5.002) under development.

Is the "real thing" still under active development?

Perl Info

Site configuration information for perl 5.00503:

Configured by sting at Mon Jul 26 10:22:37 EDT 1999.

Summary of my perl5 (5.0 patchlevel 5 subversion 3) configuration:
  Platform:
    osname=irix, osvers=6.5, archname=IP32-irix
    uname='irix linguist 6.5 01221644 ip32 '
    hint=recommended, useposix=true, d_sigaction=define
    usethreads=undef useperlio=undef d_sfio=undef
  Compiler:
    cc='cc -n32', optimize='-O3', gccversion=
    cppflags='-D_BSD_TYPES -D_BSD_TIME -OPT:Olimit=0:space=ON -I/usr/local/include -DLANGUAGE_C'
    ccflags ='-D_BSD_TYPES -D_BSD_TIME -woff 1009,1110,1184 -OPT:Olimit=0:space=ON -I/usr/local/include -DLANGUAGE_C'
    stdchar='unsigned char', d_stdstdio=define, usevfork=false
    intsize=4, longsize=4, ptrsize=4, doublesize=8
    d_longlong=define, longlongsize=8, d_longdbl=define, longdblsize=16
    alignbytes=8, usemymalloc=y, prototype=define
  Linker and Libraries:
    ld='cc -n32', ldflags ='-L/usr/local/lib -L/usr/linguist/lib -L/usr/local/lib32'
    libpth=/usr/local/lib /usr/lib32 /lib32 /lib /usr/lib /usr/linguist/lib
    libs=-lm
    libc=/usr/lib32/libc.so, so=so, useshrplib=false, libperl=libperl.a
  Dynamic Linking:
    dlsrc=dl_dlopen.xs, dlext=so, d_dlsymun=undef, ccdlflags=' '
    cccdlflags=' ', lddlflags='-n32 -shared -L/usr/local/lib -L/usr/linguist/lib -L/usr/local/lib32'

Locally applied patches:
    


@INC for perl 5.00503:
    /usr/linguist/lib/perl5/5.00503/IP32-irix
    /usr/linguist/lib/perl5/5.00503
    /usr/linguist/lib/perl5/site_perl/5.005/IP32-irix
    /usr/linguist/lib/perl5/site_perl/5.005
    .


Environment for perl 5.00503:
    HOME=/usr/people/rjk
    LANG=C
    LANGUAGE (unset)
    LD_LIBRARY_PATH (unset)
    LOGDIR (unset)
    PATH=/usr/bin:/usr/linguist/bin:/usr/sbin:/usr/bsd:/sbin:/usr/bin/X11:/usr/local/bin:/usr/afsws/bin:/usr/etc:/usr/people/rjk/scripts:/usr/people/rjk/bin:.
    PERL_BADLANG (unset)
    SHELL=/bin/tcsh

[p5pRT]

From @tamias

Here's a patch for Math​::BigFloat​::fcmp() to give correct results when one
of the operands is +0E+0. The patch to bigfltpm.t also fixes some minor
inconsistent formatting. Patched against 5.005_63.

I assume the answer to my question at the end is 'No'.

Ronald

*** lib/Math/BigFloat.pm.orig Tue Jul 20 17​:18​:00 1999
--- lib/Math/BigFloat.pm Thu Dec 16 00​:33​:59 1999
***************
*** 240,251 ****
  if ($x eq "NaN" || $y eq "NaN") {
  undef;
  } else {
  ord($y) <=> ord($x)
! ||
! ( local($xm,$xe,$ym,$ye) = split('E', $x."E$y"),
! (($xe <=> $ye) * (substr($x,$[,1).'1')
! || Math​::BigInt​::cmp($xm,$ym))
! );
  }
  }
 
--- 240,252 ----
  if ($x eq "NaN" || $y eq "NaN") {
  undef;
  } else {
+ local($xm,$xe,$ym,$ye) = split('E', $x."E$y");
+ if ($xm eq '+0' || $ym eq '+0') {
+ return $xm <=> $ym;
+ }
  ord($y) <=> ord($x)
! || ($xe <=> $ye) * (substr($x,$[,1).'1')
! || Math​::BigInt​::cmp($xm,$ym);
  }
  }
 
*** t/bigfltpm.t.orig Tue Sep 14 08​:55​:35 1999
--- t/bigfltpm.t Thu Dec 16 00​:39​:19 1999
***************
*** 9,15 ****
 
  $test = 0;
  $| = 1;
! print "1..358\n";
  while (<DATA>) {
  chop;
  if (s/^&//) {
--- 9,15 ----
 
  $test = 0;
  $| = 1;
! print "1..362\n";
  while (<DATA>) {
  chop;
  if (s/^&//) {
***************
*** 41,55 ****
  $try .= "0+\$x->fsqrt;";
  } else {
  $try .= "\$y = new Math​::BigFloat \"$args[1]\";";
! if ($f eq fcmp){
  $try .= "\$x <=> \$y;";
! }elsif ($f eq fadd){
  $try .= "\$x + \$y;";
! }elsif ($f eq fsub){
  $try .= "\$x - \$y;";
! }elsif ($f eq fmul){
  $try .= "\$x * \$y;";
! }elsif ($f eq fdiv){
  $try .= "\$x / \$y;";
  } else { warn "Unknown op"; }
  }
--- 41,55 ----
  $try .= "0+\$x->fsqrt;";
  } else {
  $try .= "\$y = new Math​::BigFloat \"$args[1]\";";
! if ($f eq "fcmp") {
  $try .= "\$x <=> \$y;";
! } elsif ($f eq "fadd") {
  $try .= "\$x + \$y;";
! } elsif ($f eq "fsub") {
  $try .= "\$x - \$y;";
! } elsif ($f eq "fmul") {
  $try .= "\$x * \$y;";
! } elsif ($f eq "fdiv") {
  $try .= "\$x / \$y;";
  } else { warn "Unknown op"; }
  }
***************
*** 271,276 ****
--- 271,280 ----
  +1​:-1​:1
  -1​:-1​:0
  +1​:+1​:0
+ -1.1​:0​:-1
+ +0​:-1.1​:1
+ +1.1​:+0​:1
+ +0​:+1.1​:-1
  +123​:+123​:0
  +123​:+12​:1
  +12​:+123​:-1

--

On Sun, Dec 12, 1999 at 02​:12​:25AM -0500, Ronald J. Kimball wrote​:

This is a bug report for perl from rjk@​linguist.dartmouth.edu,
generated with the help of perlbug 1.26 running under perl 5.00503.

-----------------------------------------------------------------
[Please enter your report here]

Math​::BigFloat​::fcmp() incorrectly determines that 0 is greater than any
positive non-integer. This occurs because in BigFloat's representation of
0 (+0E+0) has an exponent of 0, while its representation of a non-integer
such as 0.9 (+9E-1) has a negative exponent. fcmp() compares the exponents
without noting that one of the operands is equal to 0.

I note the following in Math​::BigFloat's documentation​:

BUGS

The current version of this module is a preliminary version of the
real thing that is currently (as of perl5.002) under development.

Is the "real thing" still under active development?