I'm working on an issue with initialization of affine and projective
morphisms over subschemes (Trac #20018).
In looking at this I've uncovered a sticky issue that I'm not sure what the
right fix is. Essentially, over a subscheme the coordinate ring is a
quotient ring. Thus, when initializing the morphism, the two choices are to
reduce the defining polynomials by the quotient or not to.
Currently, in elliptic curve/jacobian and toric/variety there are one or
two examples that use that reduction.
You get a different factorization without reducing
{{{
sage: R.<a,b,c> = QQ[]
sage: f = a^3+b^3+60*c^3
sage: Jacobian(f)
sage: Jacobian(f.subs(c=1))
sage: h = Jacobian(f, curve=Curve(f)); h
sage: h([1,-1,0])
Plugging in the polynomials defining `h` allows us to verify that
it is indeed a rational morphism to the elliptic curve::
sage: E = h.codomain()
sage: E.defining_polynomial()(h.defining_polynomials()).factor()
(-10077696000000000) * c^3 * b^3 * (a^3 + b^3 + 60*c^3) * (b^6 +
60*b^3*c^3 + 3600*c^6)^3
}}}
The form is different if you don't reduce.
{{{
sage: P1xP1 = toric_varieties.P1xP1()
sage: P1 = toric_varieties.P1()
sage: hom_set = P1xP1.Hom(P1); hom_set
sage: P1xP1.inject_variables()
sage: P1 = P1xP1.subscheme(s-t)
sage: hom_set = P1xP1.Hom(P1)
sage: hom_set([s,s,x,y])
Scheme morphism:
From: 2-d CPR-Fano toric variety covered by 4 affine patches
To: Closed subscheme of 2-d CPR-Fano toric variety covered
by 4 affine patches defined by:
s - t
Defn: Defined on coordinates by sending [s : t : x : y] to
[s : s : x : y]
sage: hom_set = P1.Hom(P1)
sage: hom_set([s,s,x,y])
Scheme endomorphism of Closed subscheme of 2-d CPR-Fano toric
variety covered by 4 affine patches defined by:
s - t
Defn: Defined on coordinates by sending [s : t : x : y] to
[t : t : x : y]
}}}
However, there are situations where reduction leads to issues. The original
bug is actually a case where the reduction cannot be done since the base
ring is not a field
{{{
sage: R.<t>=PolynomialRing(QQ)
sage: P.<x,y,z> = ProjectiveSpace(R,2)
sage: X = P.subscheme(x^2-y^2)
sage: H = Hom(X,X)
sage: f = H([x^2,y^2,x*z])
}}}
Another case is essentially any rational map over an affine subscheme since
typically Sage won't compute the quotient of a reduction.
{{{
sage: A.<x,y,z>=AffineSpace(QQ,3)
sage: X=A.subscheme([x^2-y^2])
sage: H=Hom(X,X)
sage: u,v,w = X.coordinate_ring().gens()
sage: H([u,v,(u+1)/v])
}}}
As an example of weird behavior: the map (x^2,y^2,z^2) on the subscheme
(x-y), a fixed point (1,1,1) gives back the multplier (0,2//0,2), when it
should be (2,0//0,2). This is because the map is reduced and initialized as
(y^2,y^2,z^2). Also, in this form it does not extend to a morphism of P^2,
where as mathematically it most certainly does.
{{{
sage: P.<x,y,z> = ProjectiveSpace(QQ,2)
sage: X=P.subscheme([x^2-y^2])
sage: H = End(X)
sage: f = H([x^2,y^2,z^2])
sage: f.multiplier(X([1,1]),1)
[0 2]
[0 2]
}}}
Perhaps compounding all this is that the actual representation of the map
is lifted to the covering ring (after reducing in the quotient). I
understand the reasoning for reducing and it is certainly not
mathematically incorrect to do so. However, neither is it mathematically
incorrect not to reduce as it is just a different representation. I find
myself uncertain what the "correct" fix is.
So, I'm looking for some feedback on what to do with initialization of
morphisms over subschemes. To reduce the defining polynomials or not to.
The third option is not to reduce by default and to provide an additional
function/option that provides a reduced representation.
Thanks,
Ben
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.
