Open kschwede opened 4 months ago
I wonder if this is related to https://github.com/Macaulay2/M2/issues/3165#issue-2206886314, since kernel
also calls the pushForward
from Core.
Is there a homogeneous version of this issue you've observed, or does it happen only in the inhomogeneous case?
cc: @Devlin-Mallory
I had a conversation with Claudiu Raicu about especially the third example.
He provided the following minimal code showing the kernel giving an incorrect answer that didn't need pushFwd
.
i2 : R = QQ[x]/ideal(x^3-2);
i3 : S = R[y]/ideal(y^2+y+1);
i4 : U = (flattenRing S)#0;
i5 : A = QQ;
i6 : f = map(U, A)
o6 = map (U, QQ, {})
o6 : RingMap U <-- QQ
i7 : g = map(U^1,A^6,f,matrix{{1,y,y*x,y*x^2,x,x^2}})
o7 = | 1 y yx yx2 x x2 |
1 6
o7 : Matrix U <-- QQ
i8 : ker g
o8 = image | 0 0 0 0 |
| 0 0 0 0 |
| 1 0 0 0 |
| 0 1 0 0 |
| 0 0 1 0 |
| 0 0 0 1 |
6
o8 : QQ-module, submodule of QQ
could it be related to #2004 ?
I've discovered this in the
PushForward
package, but I believe the issue is withkernel
. Notably, sometimespushFwd
gives the wrong answer and other times it has an error. This is an issue on both 1.24.05 and 1.22.For instance, using a multi-graded setup,
In a singly graded setup, I get the right answer.
What's worse though, in the multi-graded setup, sometimes I get no error but instead the wrong answer if the source is QQ (instead of QQ[]). In the singly graded setup I don't have this issue.
You can see in the above I am getting the wrong module out of
pushFwd