bug in collect and/or term ordering in symbolics

[zimmerma]

This seems a bug (note the instances of -x^2 and x^2):

var('a b x y z')
sage: p = -a*x^3 - a*x*y^2 + 2*b*x^2*y + 2*y^3 + x^2*z + y^2*z + x^2 + y^2 + a*x
sage: p.collect(x)
-a*x^3 + (2*b*y + z + 1)*x^2 - x^2 - (a*y^2 - a)*x + x^2 + 2*y^3 + y^2*z + y^2

[burcin]

Here is the same session using GiNaC directly via ginsh:

> t= -a*x^3 - a*x*y^2 + 2*b*x^2*y + 2*y^3 + x^2*z + y^2*z + x^2 + y^2 + a*x;
x^2+2*y*b*x^2+y^2+y^2*z+a*x+2*y^3-a*x^3-y^2*a*x+z*x^2
> t;
x^2+2*y*b*x^2+y^2+y^2*z+a*x+2*y^3-a*x^3-y^2*a*x+z*x^2
> collect(t, x);
(1+2*y*b+z)*x^2+y^2+y^2*z+2*y^3-a*x^3-(y^2*a-a)*x
> u = (x^2+(y-x^2)*(y+x));
x^2-(y+x)*(x^2-y)
> collect(u, x);
x^3-(x^2-y)*x+y^2-(-1+y+x)*x^2

It seems that one needs to call expand() explicitly before calling collect(). I think this should just be documented in the docstring.

The problem with -x^2 + x^2 appearing in the output is probably a bug I introduced while playing with the ordering of the terms. I will take a look at it when I find a chance. It's likely to be later than a week though.

[kcrisman]

This is not critical.

[kcrisman]

It turns out that #11839 was opened and did the first part of this ticket, including documenting the expand() issue.

However, the bug remains, so I'll just change this ticket to be about it.

[burcin]

With the new description, this is probably a duplicate of #9880. I will check if the pynac changes for that ticket fix this one.

[zimmerma]

bug still present in Sage 5.0.

Paul

[burcin]

This is a duplicate of #9880. With  the Pynac patch queue and patches listed on #9880, I get:

sage: var('a b x y z')
(a, b, x, y, z)
sage: p = -a*x^3 - a*x*y^2 + 2*b*x^2*y + 2*y^3 + x^2*z + y^2*z + x^2 + y^2 + a>
sage: p.collect(x)
-a*x^3 + (2*b*y + z + 1)*x^2 + 2*y^3 + y^2*z - (a*y^2 - a)*x + y^2

We should close this after adding it as a doctest to #9880.

[burcin]

Doctest is in attachment:trac_9880-doctest_for_9046.patch:ticket:9880 . This can be closed now.

[zimmerma]

Burcin,

first the ticket number is wrong (13107 instead of 9046) then the input p was mangled (ends with a> instead of a*x).

Paul

[burcin]

I replaced the patch attached to #9880:

attachment:trac_9880-doctest_for_9046.patch:ticket:9880

I hope I got it right this time. Sorry for the noise.

[zimmerma]

the patch is ok now. But since #9880 is not yet fixed, the doctest will fail. Thus we should wait for #9880 to review this one...

Paul