help links subscribe changes refresh edit

	IssueTracker
#342 axiom from CVS not building on Fedora core 6 ←	↑	→ #344 Tube Plots

#343 Bug in FreeAbelianGroup

Edit detail for #343 Bug in FreeAbelianGroup revision 3 of 5

	1 2 3 4 5
Editor: japp Time: 2008/10/16 11:09:05 GMT-7
Note:

added:

From japp Thu Oct 16 11:09:05 -0700 2008
From: japp
Date: Thu, 16 Oct 2008 11:09:05 -0700
Subject: 
Message-ID: <20081016110905-0700@axiom-wiki.newsynthesis.org>

Status: fix proposed => fixed somewhere

The order in FreeAbelianGroup is broken and makes the domain unusable (e.g., comparison of Set FreeAbelianGroup S does not work properly) For example here is a cycle:

  ----------------------------------------
  Z2:=FreeAbelianGroup Symbol         

  a:= a::FreeAbelianGroup Symbol         
  b:= b::FreeAbelianGroup Symbol
  z:= 0::FreeAbelianGroup Symbol

  a < -b 
  -b < z
  z < a
  ----------------------------------------

thus

axiom

a:= a::FreeAbelianGroup Symbol

(1)

Type: FreeAbelianGroup? Symbol

axiom

b:= b::FreeAbelianGroup Symbol

(2)

Type: FreeAbelianGroup? Symbol

axiom

z:= 0::FreeAbelianGroup Symbol

(3)

Type: FreeAbelianGroup? Symbol

axiom

(a < -b)::Boolean

(4)

Type: Boolean

axiom

(-b < z)::Boolean

(5)

Type: Boolean

axiom

(z < a)::Boolean

(6)

Type: Boolean

proposed patch in the last few lines of free.spad:

  545,546c545,546
  <         ta.gen < tb.gen => true
  <         ta.gen > tb.gen => false
  ---
  >         ta.gen < tb.gen => tb.exp > 0
  >         ta.gen > tb.gen => ta.exp < 0

apparently revlex order is intended but is not achieved in the current implementation, as the sign of the trailing exponent is not taken into account.

Fixed in wh-sandbox Revision: 496 --kratt6, Mon, 09 Apr 2007 13:50:56 -0500 reply