Re: Re: another Bug in Reduce ?

*To*: mathgroup at smc.vnet.net*Subject*: [mg60556] Re: [mg60533] Re: another Bug in Reduce ?*From*: Andrzej Kozlowski <andrzej at akikoz.net>*Date*: Tue, 20 Sep 2005 05:19:19 -0400 (EDT)*References*: <dgit7h$2ag$1@smc.vnet.net> <200509190845.EAA23572@smc.vnet.net> <ED1C56B6-A552-4CA6-A365-562B710B4271@mimuw.edu.pl>*Sender*: owner-wri-mathgroup at wolfram.com

On 19 Sep 2005, at 21:16, Andrzej Kozlowski wrote: > CylindricalDecomposition[Root[#1^3 + p*#1 - p - 1 & , 1] - Root > [#1^3 + p*#1 - p - 1 & , 2] == q, {p, q}] > > > (p <= -3 && q == (-(1/2))*Sqrt[-4*p - 3] - 3/2) || (-3 < p < -(3/4) > && q == -Sqrt[-4*p - 3]) || (p == -(3/4) && q == -(3/2)) > > > Now the case p== -3/4 does appear but look at the value of q. It is > -3/2 which is the difference -1/2 -1. But this is not the > difference between the first and the third root! So it looks like > we are seeing the cause of the bug: it comes form the fact that > CylindricalDecomposition forgets about double roots as in the above > example. I should have written "not the difference between the first and second root". As this is the key point I thought I should explain this again. Looking again at CylindricalDecomposition[y^3 + p*y - p - 1 == 0, {p, y}] (p < -3 && (y == (-(1/2))*Sqrt[-4*p - 3] - 1/2 || y == 1 || y == (1/2)*Sqrt[-4*p - 3] - 1/2)) || (p == -3 && (y == -2 || y == 1)) || (-3 < p < -(3/4) && (y == (-(1/2))*Sqrt[-4*p - 3] - 1/2 || y == (1/2)*Sqrt[-4*p - 3] - 1/2 || y == 1)) || (p == -(3/4) && (y == -(1/2) || y == 1)) || (p > -(3/4) && y == 1) we see that for the case p=-3/4 CylindricalDecomposition gives the point (y == -(1/2) || y == 1)) while the three roots (in order) are {-1/2,-1/2,1}. So in this case the difference between the first and secodn roots ought to be zero (as they are equal) but CylindricalDecomposition subtracts the two roots it has listed and gets -1/2-1 = -3/2. This appears to be what lies behind the whole problem. Andrzej Kozlowski

**References**:**Re: another Bug in Reduce ?***From:*Maxim <ab_def@prontomail.com>