I definitely agree about tests! There should be a generic "test suite" that should systematically be run with these all different implementations. It is a bit ridiculous that your example was not caught.
On Fri, 29 Sept 2023 at 17:53, John H Palmieri <jhpalmier...@gmail.com> wrote: > > First, an error in the original message: it should have been > "identity_matrix(F, 1)". That doesn't make any difference in the behavior. A > little exploration suggests that the problem goes away if you make the > identity matrix sparse: the problem may be with multiplying a dense matrix > with a vector. I'm not sure about this, though. > > This is with the most recent beta, built from scratch on several different OS > X machines. It also happens on Sagecell. I have not yet opened an issue, > because there are deeper existential questions: if linear algebra over this > field is broken, and if there are no doctests to detect that it's broken, how > viable is this implementation? How long has it been broken? Why has no one > noticed? What is it used for? Should we keep it? > > I think that if someone wanted to add a new implementation of a field to Sage > today, we would insist on doctests verifying that it worked as a drop-in > replacement for the existing implementations everywhere in Sage where fields > get used. GF(2, impl='ntl') does not meet this standard. Someone can > presumably fix this one bug, but is someone willing to add all of the > doctests that should really be added? > > On Friday, September 29, 2023 at 12:32:14 AM UTC-7 Vincent Delecroix wrote: >> >> What is your sage version? How did you install it? Did you open an >> issue on github? >> >> Best >> Vincent >> >> On Tue, 26 Sept 2023 at 07:00, John H Palmieri <jhpalm...@gmail.com> wrote: >> > >> > Setup: >> > >> > sage: F = GF(2, impl='ntl') >> > sage: m_ntl = identity_matrix(1, F) >> > sage: v_ntl = vector(F, (1,)) >> > >> > Now consider >> > >> > sage: m_ntl * v_ntl >> > sage: v_ntl * m_ntl >> > >> > I'm trying to multiply a 1x1 matrix by a 1-dimensional vector, in one >> > order or the other. Here's what happens: the first line fails with a >> > SignalError, and the second actually crashes Sage. If we are defining a >> > field that can't do linear algebra, shouldn't there be big warnings posted >> > somewhere? If we are defining a field like this, are there any >> > expectations that it should work broadly with Sage types and >> > constructions? I just discovered that cup products in mod 2 cohomology >> > don't work when "mod 2" means coefficients in GF(2, impl='ntl'), and I >> > don't know if I should bother trying to fix this. >> > >> > For what it's worth, I get the same with `GF(2, impl='givaro')`. Also for >> > what it's worth, explicitly converting `v_ntl` to a matrix allows the >> > matrix multiplication to work. >> > >> > I see this on two different OS X machines, one Intel and one Apple Silicon. >> > >> > -- >> > John >> > >> > -- >> > 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+...@googlegroups.com. >> > To view this discussion on the web visit >> > https://groups.google.com/d/msgid/sage-devel/faeffb72-3b16-4f37-90c4-f969e4d5c017n%40googlegroups.com. > > -- > 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 view this discussion on the web visit > https://groups.google.com/d/msgid/sage-devel/378f06a5-2008-42fb-8642-32019a3a4976n%40googlegroups.com. -- 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 view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/CAGEwAA%3DumacWP1W0CbyT40a9oPO6B_ejDcqc7HbZkck2Y_5kBg%40mail.gmail.com.
