Bug #6986: Inconsistent result of BigDecimal#power

• Status: Closed
• Assignee: mrkn (Kenta Murata)
• Target version: -
• ruby -v: ruby 2.0.0dev (2012-09-05 trunk 36913) [i386-mswin32_100]

[phasis68 (Heesob Park)]

I found BigDecimal#power returns incorrect result for a large value.

C:\tmp>ruby -rbigdecimal -e "p BigDecimal('1e-10').power(1e7)"
#BigDecimal:a2bdb8,'0.1E-99999999',9(100000026)

C:\tmp>ruby -rbigdecimal -e "p BigDecimal('1e-10').power(1e8)"
-e:1:in power': failed to allocate memory (NoMemoryError) from -e:1:in '

C:\tmp>ruby -rbigdecimal -e "p BigDecimal('1e-10').power(1e10)"
#BigDecimal:f7cb3c,'0.0',9(18)

C:\tmp>ruby -rbigdecimal -e "p BigDecimal('1e-10').power(1e19)"
#BigDecimal:f7cb3c,'Infinity',9(18)

[shyouhei (Shyouhei Urabe)]

• Category set to ext
• Status changed from Open to Assigned
• Assignee set to mrkn (Kenta Murata)

[mrkn (Kenta Murata)]

I cannot reproduce NoMemoryError on my environment.
However I think the following behaviors for the large parameters may be inconsistent, so I will attempt to fix these behavior.

Unfortunately I may not be able to fix NoMemoryError because I don't have the environment of mswin32.
Please submit your patch for fixing it.

$ ruby-head -v
ruby 2.0.0dev (2012-09-06 trunk 36915) [x86_64-darwin11.4.0]
$
$ ruby-head -rbigdecimal -e 'p BigDecimal("1e-10").power(1e7)'
#BigDecimal:7fbd51033f98,'0.1E-99999999',9(100000026)
$
$ ruby-head -rbigdecimal -e 'p BigDecimal("1e-10").power(1e8)'
#BigDecimal:7f8401033f90,'0.1E-999999999',9(1000000026)
$
$ ruby-head -rbigdecimal -e 'p BigDecimal("1e-10").power(1e10)'
-e:1:in power': integer 10000000000 too big to convert to int' (RangeError)
from -e:1:in `'
$
$ ruby-head -rbigdecimal -e 'p BigDecimal("1e-10").power(1e19)'
#BigDecimal:7fe595004998,'0.0',9(18)

[phasis68 (Heesob Park)]

Could you try "p BigDecimal('1e-10').power(1e50)" ?
On Windows, the parameter higher than 1e19 returns Infinity instead of 0.0.

Here is a patch of NoMemoryError on Windows.

diff --git a/bigdecimal.c b/bigdecimal.c.new
index e798781..976503b 100644
--- a/bigdecimal.c
+++ b/bigdecimal.c.new
@@ -2184,7 +2184,7 @@ retry:
if (exp != NULL) {
return rmpd_power_by_big_decimal(x, exp, n);
}

• else if (TYPE(vexp) == T_BIGNUM) {

• else if (TYPE(vexp) == T_BIGNUM || FIX2INT(vexp)>=10000000) {
  VALUE abs_value = BigDecimal_abs(self);
  if (is_one(abs_value)) {
  return ToValue(VpCreateRbObject(n, "1"));

[usa (Usaku NAKAMURA)]

Hello,

In message "[ruby-core:47446] [ruby-trunk - Bug #6986] Inconsistent result of BigDecimal#power"
on Sep.06,2012 14:27:38, [e-mail] wrote:

I cannot reproduce NoMemoryError on my environment.
However I think the following behaviors for the large parameters may be inconsistent, so I will attempt to fix these behavior.

Unfortunately I may not be able to fix NoMemoryError because I don't have the environment of mswin32.
Please submit your patch for fixing it.

Generally, you should fix platform independent problem first.
If some platform dependent problem remains after it, I'll fix it.

Regards,

U.Nakamura [e-mail]

[mrkn (Kenta Murata)]

• Status changed from Assigned to Closed

I think it is fixed in bigdecimal version 1.2.3.

[phasis68 (Heesob Park)]

There is still NoMemoryError on Windows with trunk.

C:\Users\phasis>ruby -v -rbigdecimal -e "p BigDecimal('1e-10').power(1e8)"
ruby 2.1.0dev (2013-11-28 trunk 43886) [i386-mingw32]
-e:1:in power': failed to allocate memory (NoMemoryError) from -e:1:in '