The fix was easy. Much harder was to find the root of the problem! If you are interested in intensive computations over QQbar I strongly advise you to look towards https://flintlib.org/ maintained by Fredrik Johansson. This library has *two* implementations of QQbar which are way more performant (and probably more reliable) than what is in SageMath. The more or less short term goal is to use flint implementations to replace the current QQbar implementation in SageMath.You could give it a try inside SageMath via https://github.com/mezzarobba/flint_gr_sage (feel free to open an issue there asking for installation help or describing the things you want to compute).
I do intend to make my commit part of SageMath QQbar, but it is not on the top priority list. Best Vincent On Mon, 6 May 2024 at 09:27, Hakan Granath <hakan.gran...@gmail.com> wrote: > > Great, thanks for working out and fixing the (quite non-trivial, it seems) > root cause of this issue! > > Just a curious question: would it be useful to add the functionality of your > commit 8de6ba5 to potentially simplify QQbar objects? Especially doing linear > algebra over QQbar I guess you get many trivial operations (like adding 0) > and optimizing them away might maybe improve performance? Just an idea, I > haven't done any benchmarking to support that hypothesis. > > Best regards, > > Håkan > > On Mon, May 6, 2024 at 12:15 AM Vincent Delecroix <20100.delecr...@gmail.com> > wrote: >> >> Turns out to be a serious bug in complex interval fields. The problem >> is hopefully fixed by https://github.com/sagemath/sage/pull/37941 >> which should make its way to the next sage release. >> >> Thanks again for your report. >> Vincent >> >> On Fri, 3 May 2024 at 18:13, vdelecroix <20100.delecr...@gmail.com> wrote: >> > >> > Thanks for your report! I simplified a bit your example and posted an >> > issue https://github.com/sagemath/sage/issues/37927. >> > >> > Vincent >> > >> > Le vendredi 3 mai 2024 à 15:05:03 UTC+2, Hakan Granath a écrit : >> >> >> >> Hi, >> >> >> >> I think sometimes matrices over QQbar give erroneous results (sorry for >> >> the messy example, if I try to simplify it the problem disappears): >> >> >> >> R.<y> = QQ[] >> >> v1 = QQbar.polynomial_root(AA.common_polynomial(y^2 + 1), CIF(RIF(RR(0)), >> >> RIF(-RR(1)))) >> >> v2 = QQbar.polynomial_root(AA.common_polynomial(y^2 + 1), CIF(RIF(RR(0)), >> >> RIF(RR(1)))) >> >> v3 = 4*v2 >> >> v4 = AA.polynomial_root(AA.common_polynomial(y^2 - 2), >> >> RIF(-RR(1.4142135623730951), -RR(1.4142135623730949))) >> >> v5 = AA.polynomial_root(AA.common_polynomial(y^2 - 2), >> >> RIF(RR(1.4142135623730949), RR(1.4142135623730951))) >> >> v6 = QQbar.polynomial_root(AA.common_polynomial(y^2 + 16), >> >> CIF(RIF(RR(0)), RIF(-RR(4)))) >> >> v7 = v6*v6 >> >> v8 = QQbar.polynomial_root(AA.common_polynomial(y^2 + 1), CIF(RIF(RR(0)), >> >> RIF(RR(1)))) >> >> >> >> M = matrix(QQbar, [[0, 0, 1, 0, 0, 0, 0, 0, 0, 0], [0, 1, 0, 0, 0, 0, 0, >> >> 0, 0, 0], [-4, 2*v1, 1, 64, -32*v1, -16, 8*v1, 4, -2*v1, -1], [4*v1, 1, >> >> 0, -192*v1, -80, 32*v1, 12, -4*v1, -1, 0], [2, 0, 0, -480, 160*v1, 48, >> >> -12*v1, -2, 0, 0], [-4, 2*v2, 1, 64, -32*v2, -16, 8*v2, 4, -2*v2, -1], >> >> [v3, 1, 0, -192*v2, -80, 32*v2, 12, -v3, -1, 0], [2, 0, 0, -480, 160*v2, >> >> 48, -12*v2, -2, 0, 0], [0, 0, 0, 8, QQbar(4*v4), 4, QQbar(2*v4), 2, >> >> QQbar(AA.polynomial_root(AA.common_polynomial(y^2 - 2), >> >> RIF(-RR(1.4142135623730951), -RR(1.4142135623730949)))), 1], [0, 0, 0, >> >> QQbar(24*v4), 20, QQbar(8*v4), 6, QQbar(2*v4), 1, 0], [0, 0, 0, 8, >> >> QQbar(4*v5), 4, QQbar(2*v5), 2, >> >> QQbar(AA.polynomial_root(AA.common_polynomial(y^2 - 2), >> >> RIF(RR(1.4142135623730949), RR(1.4142135623730951)))), 1], [0, 0, 0, >> >> QQbar(24*v5), 20, QQbar(8*v5), 6, QQbar(2*v5), 1, 0], [0, 0, 0, v7*v7*v7, >> >> v7*v7*v6, v7*v7, v6*v6*v6, v6*v6, v6, 1], [0, 0, 0, -4096, 1024*v8, 256, >> >> -64*v8, -16, 4*v8, 1]]) >> >> >> >> With this matrix I get: >> >> >> >> sage: M.right_kernel_matrix() >> >> [] >> >> >> >> but in fact the right kernel is generated by: >> >> >> >> sage: v = Matrix(QQbar, 10, 1, [-108, 0, 0, 1, 0, 12, 0, -60, 0, 64]) >> >> sage: M * v == 0 >> >> True >> >> >> >> This is with SageMath version 10.3 using Python 3.11.1 on Ubuntu 22.04.4. >> >> >> >> Best regards, >> >> >> >> Håkan Granath >> > >> > -- >> > 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/1c927686-bada-4943-8000-a6fdf6e71029n%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/CAGEwAAkgsAcFmt-6NKWvuh0va_UVy-jdSJVnPrSrjGE_CA3Nog%40mail.gmail.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/CAOnrL6%2BO0s%3D6poKYFo17xTLrrX4SdCFuj%2BQj7tpkjtQfL%2BvQNw%40mail.gmail.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/CAGEwAAnudV9KV1daN%3DY5YCC3btCf7%3D%3D9hc9rRTmuX0Ck%2BRs8ww%40mail.gmail.com.
