[sage-devel] Re: List of doctest failures in current Mandriva sagemath
2009/9/2 Martin Albrecht : > >> > doc/en/constructions/rings.rst +58 >> > sage: R = singular.ring(97, '(a,b,c,d)', 'lp') >> > sage: I = singular.ideal(['a+b+c+d', 'ab+ad+bc+cd', >> > 'abc+abd+acd+bcd', 'abcd-1']) >> > sage: R >> > Expected: >> > // characteristic : 97 >> > // number of vars : 4 >> > // block 1 : ordering lp >> > // : names a b c d >> > // block 2 : ordering C >> > Got: >> > // characteristic : 97 >> > // number of vars : 4 >> > // block 1 : ordering lp >> > // : names abcd >> > // block 2 : ordering C >> > * The sage spkg don't have a patch to separate the names, so I am >> > assuming it is a minor change in singular >> >> looks safe > > Yes, this was fixed in Singular recently, I assume Mandriva only needs to > update to the newest upstream release. I tried again a newer singular, but singular-3-1-0-4, which is the latest one api/abi compatible with sage. I am preferring to package upstream, and then apply sage patches as appropriate. I will check if I can make a simple patch make singular provide the result expected by sage. >> > rings/polynomial/toy_d_basis.py +171 >> > sage: from sage.rings.polynomial.toy_d_basis import gpol >> > sage: P. = PolynomialRing(IntegerRing(), 3, order='lex') >> > sage: f = x^2 - 1 >> > sage: g = 2*x*y - z >> > sage: gpol(f,g) >> > Expected: >> > x^2*y - y >> > Got: >> > x^2*y - x*z + y >> > * Not sure what is the cause, neither if this is an alternate correct >> > result...] >> >> Martin -- any thoughts? > > Here is what gpol does > > a1,a2 = g1.lc(),g2.lc()# a1 = 1, a2 = 2 > a, c1, c2 = xgcd(a1,a2) # (1,0,1) -> this is not unique > t1,t2 = g1.lm(), g2.lm() # x^2, x*y > t = t1.parent().monomial_lcm(t1,t2) # x^2*y > s1,s2 = t//t1, t//t2 # y, x > return c1*s1*g1 + c2*s2*g2 # 0*y*g1 + 1*x*g2 > > I guess xgcd changed (e.g. (1,-1,1)) and thus the result is different. So it > seems also correct. Many thanks for the review. About quaddouble, since only sagemath requires it, I packaged the sage spkg in Mandriva. But the package is kind problematic, as it has only a static library, and the sage spkg fails to build the fortran bindings in x86_64. Anyway, if sage stops using it, then it can be dropped from Mandriva later. Since there is still some time before Mandriva 2010.0, I am also updating the package to ship sage 4.1.1. > Cheers, > Martin Paulo --~--~-~--~~~---~--~~ To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~--~~~~--~~--~--~---
[sage-devel] Re: List of doctest failures in current Mandriva sagemath
> > doc/en/constructions/rings.rst +58 > >sage: R = singular.ring(97, '(a,b,c,d)', 'lp') > >sage: I = singular.ideal(['a+b+c+d', 'ab+ad+bc+cd', > > 'abc+abd+acd+bcd', 'abcd-1']) > >sage: R > > Expected: > >// characteristic : 97 > >// number of vars : 4 > >//block 1 : ordering lp > >// : namesa b c d > >//block 2 : ordering C > > Got: > >// characteristic : 97 > >// number of vars : 4 > >//block 1 : ordering lp > >// : namesabcd > >//block 2 : ordering C > > * The sage spkg don't have a patch to separate the names, so I am > > assuming it is a minor change in singular > > looks safe Yes, this was fixed in Singular recently, I assume Mandriva only needs to update to the newest upstream release. > > rings/polynomial/toy_d_basis.py +171 > >sage: from sage.rings.polynomial.toy_d_basis import gpol > >sage: P. = PolynomialRing(IntegerRing(), 3, order='lex') > >sage: f = x^2 - 1 > >sage: g = 2*x*y - z > >sage: gpol(f,g) > > Expected: > >x^2*y - y > > Got: > >x^2*y - x*z + y > > * Not sure what is the cause, neither if this is an alternate correct > > result...] > > Martin -- any thoughts? Here is what gpol does a1,a2 = g1.lc(),g2.lc()# a1 = 1, a2 = 2 a, c1, c2 = xgcd(a1,a2) # (1,0,1) -> this is not unique t1,t2 = g1.lm(), g2.lm() # x^2, x*y t = t1.parent().monomial_lcm(t1,t2) # x^2*y s1,s2 = t//t1, t//t2 # y, x return c1*s1*g1 + c2*s2*g2 # 0*y*g1 + 1*x*g2 I guess xgcd changed (e.g. (1,-1,1)) and thus the result is different. So it seems also correct. Cheers, Martin -- name: Martin Albrecht _pgp: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x8EF0DC99 _otr: 47F43D1A 5D68C36F 468BAEBA 640E8856 D7951CCF _www: http://www.informatik.uni-bremen.de/~malb _jab: martinralbre...@jabber.ccc.de --~--~-~--~~~---~--~~ To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~--~~~~--~~--~--~---
[sage-devel] Re: List of doctest failures in current Mandriva sagemath
2009/9/1 Paulo César Pereira de Andrade <
paulo.cesar.pereira.de.andr...@gmail.com>
>
> Hi,
>
> The package I am building uses newer versions of several components,
> and while I believe most of these tests probably are correct, I may be
> missing some patch, so, if you can confirm it is correct, or is an
> alternate correct response, please let me know.
> These are not the only errors, but are a sampling of similar errors:
>
> doc/en/constructions/polynomials.rst +45
>sage: print gap.eval("R:= PolynomialRing( GF(97))")
> Expected:
>PolynomialRing(..., [ x_1 ])
> Got:
>GF(97)[x_1]
> * this should be due to using gap 4.4.12, while sage uses gap 4.4.10
>
looks fine
>
> doc/en/constructions/rings.rst +58
>sage: R = singular.ring(97, '(a,b,c,d)', 'lp')
>sage: I = singular.ideal(['a+b+c+d', 'ab+ad+bc+cd',
> 'abc+abd+acd+bcd', 'abcd-1'])
>sage: R
> Expected:
>// characteristic : 97
>// number of vars : 4
>//block 1 : ordering lp
>// : namesa b c d
>//block 2 : ordering C
> Got:
>// characteristic : 97
>// number of vars : 4
>//block 1 : ordering lp
>// : namesabcd
>//block 2 : ordering C
> * The sage spkg don't have a patch to separate the names, so I am
> assuming it is a minor change in singular
>
looks safe
>
> doc/en/tutorial/tour_numtheory.rst +94
>sage: x = crt(2, 1, 3, 5); x
> Expected:
>-4
> Got:
>11
> * This is caused by using pari 2.3.4 while sage uses pari 2.3.3
>
safe
>
> libs/pari/gen.pyx +6781
>sage: y = QQ['yy'].0; _ = pari(y) # pari has variable ordering
> rules
>sage: x = QQ['zz'].0; nf = pari(x^2 + 2).nfinit()
>sage: nf.nfroots(y^2 + 2)
> Expected:
>[-zz, zz]
> Got:
>[Mod(-zz, zz^2 + 2), Mod(zz, zz^2 + 2)]
> * Again, due to using newer version of pari
>
ok
>
> matrix/matrix_double_dense.pyx +983
>sage: m = matrix(RDF, 2, range(6)); m
>[0.0 1.0 2.0]
>[3.0 4.0 5.0]
>sage: U, S, V = m.SVD()
>sage: U*S*V.transpose() # random low bits
>[7.62194127257e-17 1.0 2.0]
>[ 3.0 4.0 5.0]
>sage: U
>[-0.274721127897 -0.961523947641]
>[-0.961523947641 0.274721127897]
>sage: S
>[7.34846922835 0.0 0.0]
>[ 0.0 1.0 0.0]
>sage: V
> Expected:
>[-0.392540507864 0.824163383692 0.408248290464]
>[-0.560772154092 0.137360563949 -0.816496580928]
>[ -0.72900380032 -0.549442255795 0.408248290464]
> Got:
>[-0.392540507864 0.824163383692 -0.408248290464]
>[-0.560772154092 0.137360563949 0.816496580928]
>[ -0.72900380032 -0.549442255795 -0.408248290464]
>
last eigenvector is just off by sign?
> * This one gives significantly different result, but is not easy to do
> an alternate build with sage's version of quaddouble
>
> * I think I will switch to use sage's version. Sage uses quaddouble
> 2.2.p9 (patchlevel 9), while I packaged upstream
> quaddouble 2.2.7
>
> rings/real_rqdf.pyx +463
>sage: RQDF(2^60 + 9 )
> Expected:
>1.152921504606846985e18
> Got:
>1e+18
> * Again should be a quaddouble issue. But I can see that sage result
> is correct, while the quaddouble I am using is truncating the result.
>
We voted a lng time ago to completely remove quaddouble from Sage, since
it causes too much trouble for too little gain. This needs to happen asap.
Anyway, quaddouble will vanish from sage soon, so I wouldn't worry too much
about quaddouble issues.
>
> rings/polynomial/toy_d_basis.py +171
>sage: from sage.rings.polynomial.toy_d_basis import gpol
>sage: P. = PolynomialRing(IntegerRing(), 3, order='lex')
>sage: f = x^2 - 1
>sage: g = 2*x*y - z
>sage: gpol(f,g)
> Expected:
>x^2*y - y
> Got:
>x^2*y - x*z + y
> * Not sure what is the cause, neither if this is an alternate correct
> result...]
>
Martin -- any thoughts?
>
> calculus/calculus.py +1068
>sage: f = log(log(x))/log(x)
>sage: forget(); assume(x<-2); lim(f, x=0, taylor=True)
> Expected:
>limit(log(log(x))/log(x), x, 0)
> Got:
>0
> * This is when using newer maxima
>
>
That agrees with Mathematica at least:
In[4]:= Limit[Log[Log[x]]/Log[x],x->0]
Out[4]= 0
So it seems like it is just an improvement we'll get in Sage.
It sounds like Sage in Mandriva will be better than our Sage, until we
update a few packages.
William
--~--~-~--~~~---~--~~
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups
