Upgrade to Pynac-0.7.10

[rws]

In the new version:

• remove static PyObjects; (#23134)
• fix segmentation fault with coefficients() (#23545)
• fix compile error --with-giac=yes
• fix tgamma/lgamma return types
• (real^m)^(1/n) --> abs(real)^(m/n), m even, m/n integer (#14305)
• fix add::conjugate() (#23135)
• fix is_real(zeta(real))
• negative symbolic domain from assume(sym<0) (#21973)
• add missing calls to set_domain in symbol constructors (#23138)
• implement add.info(negative)
• exact comparison of overall_coefficients
• internal LONG numeric; nice speedup in some parts
• fast internal in-place arithmetic operators
• obsolete some internal usages of Python

​https://github.com/pynac/pynac/releases/download/pynac-0.7.10/pynac-0.7.10.tar.bz2

[rws]

New commits:

​1d1aeb8	23325: pkg version/chksum, remove obsolete patch
​cf62180	23325: fix py_get_parent_char()
​f976b8c	23325: fix doctests

[jdemeyer]

I don't agree with the change to py_get_parent_char: why assume characteristic zero if the parent cannot decide?

[jdemeyer]

If you revert the change to py_get_parent_char and all doctests pass, you can set this to positive review.

[rws]

That's a no, because this then needs investigation (doctest fail at src/sage/manifolds/differentiable/diff_form.py:796):

sage: M = Manifold(3, 'R3', '\RR^3', start_index=1)
sage: c_cart.<x,y,z> = M.chart()
sage: eps = M.diff_form(3, 'epsilon', r'\epsilon')
sage: eps[1,2,3] = 1
sage: a = M.one_form('A')
sage: a[:] = (x*y*z, -z*x, y*z)
sage: b = M.one_form('B')
sage: b[:] = (cos(z), sin(x), cos(y))
sage: ab = a.wedge(b)
sage: dab = ab.exterior_derivative()
sage: dab[[1,2,3]]/eps[[1,2,3]]
...
/home/ralf/sage/src/sage/structure/element.pyx in sage.structure.element.RingElement._div_ (build/cythonized/sage/structure/element.c:17655)()
   2453         return l
   2454 
-> 2455     cpdef _div_(self, other):
   2456         """
   2457         Default implementation of division using the fraction field.

/home/ralf/sage/local/lib/python2.7/site-packages/sage/manifolds/coord_func_symb.pyc in _div_(self, other)
    947         if other._express.is_zero():
    948             raise ZeroDivisionError("division of a coordinate function by zero")
--> 949         res = self._simplify(self._express / SR(other))
    950         if res.is_trivial_zero():  # NB: "if res == 0" would be too
    951                                    # expensive (cf. #22859)

/home/ralf/sage/src/sage/structure/element.pyx in sage.structure.element.Element.__div__ (build/cythonized/sage/structure/element.c:12950)()
   1638         cdef int cl = classify_elements(left, right)
   1639         if HAVE_SAME_PARENT(cl):
-> 1640             return (<Element>left)._div_(right)
   1641         if BOTH_ARE_ELEMENT(cl):
   1642             return coercion_model.bin_op(left, right, div)

/home/ralf/sage/src/sage/symbolic/expression.pyx in sage.symbolic.expression.Expression._div_ (build/cythonized/sage/symbolic/expression.cpp:24795)()
   3445                 raise ZeroDivisionError("Symbolic division by zero")
   3446             else:
-> 3447                 raise
   3448         finally:
   3449             sig_off()

/home/ralf/sage/src/sage/symbolic/expression.pyx in sage.symbolic.expression.Expression._div_ (build/cythonized/sage/symbolic/expression.cpp:24684)()
   3437                                o)
   3438             else:
-> 3439                 x = left._gobj / _right._gobj
   3440             return new_Expression_from_GEx(left._parent, x)
   3441         except Exception as msg:

/home/ralf/sage/src/sage/libs/pynac/pynac.pyx in sage.libs.pynac.pynac.py_get_parent_char (build/cythonized/sage/libs/pynac/pynac.cpp:9942)()
    818         return 0
    819 
--> 820     c = (<Element>o)._parent.characteristic()
    821 
    822     # Pynac only differentiates between

/home/ralf/sage/local/lib/python2.7/site-packages/sage/categories/rings.pyc in characteristic(self)
    461             from sage.rings.infinity import infinity
    462             from sage.rings.integer_ring import ZZ
--> 463             order_1 = self.one().additive_order()
    464             return ZZ.zero() if order_1 is infinity else order_1
    465 

/home/ralf/sage/src/sage/structure/element.pyx in sage.structure.element.RingElement.additive_order (build/cythonized/sage/structure/element.c:18049)()
   2470         Return the additive order of ``self``.
   2471         """
-> 2472         raise NotImplementedError
   2473 
   2474     def multiplicative_order(self):

NotImplementedError: 

sage: dab[[1,2,3]]
Scalar field on the 3-dimensional differentiable manifold R3
sage: type(_)
<class 'sage.manifolds.differentiable.scalarfield_algebra.DiffScalarFieldAlgebra_with_category.element_class'>
sage: eps[[1,2,3]]
Scalar field on the 3-dimensional differentiable manifold R3
sage: type(_)
<class 'sage.manifolds.differentiable.scalarfield_algebra.DiffScalarFieldAlgebra_with_category.element_class'>

sage: parent(dab[[1,2,3]])
Algebra of differentiable scalar fields on the 3-dimensional differentiable manifold R3
sage: parent(eps[[1,2,3]])
Algebra of differentiable scalar fields on the 3-dimensional differentiable manifold R3

While this is triggered by changes in Pynac that should be addressed it's not clear to me why the division of two scalar fields must go through SR at all.

[jdemeyer]

I will have a look.

[rws]

The reason why this is now failing is that the quotient here, which is 1 but not integer, is now recognized as is_one and triggers the ...pos_char call.

​https://github.com/pynac/pynac/blob/master/ginac/mul.cpp#L702

Removing the call altogether in preliminary testing only fails one doctest involving Mod and, frankly, I could care less. Full testing now.

[jdemeyer]

I think the best solution here is simply to implement characteric for the "Ring of coordinate functions".

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​fb52b8b

23325: revert libs/pynac commit

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​1a369cb	Merge branch 'develop' into t/23325/upgrade_to_pynac_0_7_9
​1fa0f69	Merge branch 'develop' into t/23325/upgrade_to_pynac_0_7_9
​725db22	Merge branch 'develop' into t/23325/upgrade_to_pynac_0_7_9
​3b7db8b	pynac-2.7.10 doctest fixes
​f5682bf	necessary code change to make pynac-2.7.10 work

[rws]

Correct is 0.7.10 not 2.7.10.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​bafd41a	Merge branch 'develop' into t/23325/upgrade_to_pynac_0_7_9
​a90ce6a	23582: Robustify doctest in hyperbolic_geodesic.py
​0879761	Merge branch 'u/rws/robustify_doctest_in_hyperbolic_geodesic_py' of git://trac.sagemath.org/sage into t/23325/upgrade_to_pynac_0_7_9

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​28ae958	23325: doctest fixes
​5fc85e5	23587: Silence Cython warnings with expression.pyx
​b0dd29b	Merge branch 'u/rws/silence_cython_warnings_with_expression_pyx' of git://trac.sagemath.org/sage into t/23325/upgrade_to_pynac_0_7_9

[chapoton]

May we please change the .configure line in spkg-install to

    ./configure --disable-static --prefix=${SAGE_LOCAL} --with-giac=no --libdir=
"$SAGE_LOCAL/lib" PYTHON=sage-python23

to ensure compatibility with python3 ?

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​ad2e96d

23325: explicitly look for sage-python23

[rws]

If you looked at the doctest changes you probably have noticed natural simplifications where quadratic numeric factors inside sqrt are drawn out of the root. I will have to revert this behaviour, unfortunately. There is code in the Sage code base that relies on that root content is not changed, because it allows manipulations leading to expressions that can later be better simplified by Maxima's simplification functions. Apparently none of these functions is able to normalize root content in complicated expressions. I will maybe implement it separately, similar to gamma_normalize, for inclusion in simplify_full but releasing pynac-0.7.10 is more urgent.

[rws]

I made a new branch. Please review.

---

New commits:

​86d0ef3	trac 23329 characteristic of coordinate chart function ring
​2e40ec5	Merge branch 'u/rws/robustify_doctest_in_hyperbolic_geodesic_py' of git://trac.sagemath.org/sage into tmp11
​0f51d45	23325: chksum,version
​3cc0708	23325: remove obsolete patch
​95cb0a5	23325: explicitly look for sage-python23
​7793f1c	23325: code change to make use of new pynac
​9520ada	23325: doctest fixes

[jdemeyer]

For ease of reviewing, I would prefer to wait until the dependencies have been merged.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​f52f602

23325: revert merge of 23582

[jdemeyer]

I don't think this is a valid way to revert a merge. Either you really revert the merge (by rewriting git history as if the merge never happened), or you keep the dependency on #23582.

[rws]

Nice opportunity for the rarely used git rebase -i.

---

New commits:

​a887a8b	23325: chksum,version
​98bab37	23325: remove obsolete patch
​fe100f7	23325: explicitly look for sage-python23
​8fe77a7	23325: code change to make use of new pynac
​62cebba	23325: doctest fixes

[jdemeyer]

Replying to rws:

Nice opportunity for the rarely used git rebase -i.

I use that all the time...

[jdemeyer]

New commits:

​b54f51b

Doc formatting

[vbraun]

Fails on 32-bit Linux

**********************************************************************
File "src/doc/de/thematische_anleitungen/sage_gymnasium.rst", line 191, in doc.de.thematische_anleitungen.sage_gymnasium
Failed example:
    sqrt(18)
Expected:
    3*sqrt(2)
Got:
    sqrt(18)
**********************************************************************
1 item had failures:
   1 of 294 in doc.de.thematische_anleitungen.sage_gymnasium
    [207 tests, 1 failure, 5.39 s]
**********************************************************************
File "src/sage/functions/orthogonal_polys.py", line 1668, in sage.functions.orthogonal_polys.Func_assoc_legendre_Q._eval_
Failed example:
    gen_legendre_Q(2,1,3)
Expected:
    -1/4*sqrt(-2)*(-36*I*pi + 36*log(4) - 36*log(2) - 25)
Got:
    -1/8*sqrt(-8)*(-36*I*pi + 36*log(4) - 36*log(2) - 25)
**********************************************************************
1 item had failures:
   1 of   2 in sage.functions.orthogonal_polys.Func_assoc_legendre_Q._eval_
    [378 tests, 1 failure, 7.79 s]
**********************************************************************
File "src/sage/rings/number_field/number_field_base.pyx", line 219, in sage.rings.number_field.number_field_base.NumberField.minkowski_bound
Failed example:
    B = K.minkowski_bound(); B
Expected:
    16/3*sqrt(3)/pi
Got:
    8/9*sqrt(108)/pi
**********************************************************************
File "src/sage/rings/number_field/number_field_base.pyx", line 230, in sage.rings.number_field.number_field_base.NumberField.minkowski_bound
Failed example:
    B = K.minkowski_bound(); B
Expected:
    sqrt(10)
Got:
    1/2*sqrt(40)
**********************************************************************
1 item had failures:
   2 of  19 in sage.rings.number_field.number_field_base.NumberField.minkowski_bound
    [74 tests, 2 failures, 2.09 s]
**********************************************************************
File "src/sage/rings/real_mpfr.pyx", line 3961, in sage.rings.real_mpfr.RealNumber.sqrt
Failed example:
    r.sqrt()
Expected:
    2*sqrt(1086)
Got:
    sqrt(4344)
**********************************************************************
1 item had failures:
   1 of  13 in sage.rings.real_mpfr.RealNumber.sqrt
    [1000 tests, 1 failure, 3.14 s]

[rws]

Okay, I'm installing virtualization (a first) and Ubuntu 32-bit to get the ability to test on 32-bit.

[rws]

STDERR: error: Server does not allow request for unadvertised object b54f51bab33a602448c1b3e2f719fb375661aac3

So I'll push to my previous branch u/rws/23325-1 (including Jeroen's fix) in the hope you can pick up its last commit, which contains a last-minute fix.

[rws]

New commits:

​a887a8b	23325: chksum,version
​98bab37	23325: remove obsolete patch
​fe100f7	23325: explicitly look for sage-python23
​8fe77a7	23325: code change to make use of new pynac
​62cebba	23325: doctest fixes
​5a2da58	23325: doc formatting
​5387ba5	23325: last-minute 32-bit fix

[fbissey]

I think it can just be put back to positive review. Especially if the last fix is the stuff for 32bits problems.