Vector constructor fails on empty list with no ring given

[Rob Beezer]

sage: vector(QQ,[])
()

sage: vector([])
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)

/home/sage/sage-4.6.1.rc1/devel/sage-main/<ipython console> in <module>()

/sage/sage-4.6.1.rc1/local/lib/python2.6/site-packages/sage/modules/free_module_element.so in sage.modules.free_module_element.vector (sage/modules/free_module_element.c:3229)()

/sage/sage-4.6.1.rc1/local/lib/python2.6/site-packages/sage/modules/free_module_element.so in sage.modules.free_module_element.prepare (sage/modules/free_module_element.c:3675)()

TypeError: unable to find a common ring for all elements

Default ring should be ZZ ideally.

---

Apply both

• trac_10595-vector-constructor-empty-list.patch
• 10595-referee.patch

[Rob Beezer]

Patch depends on #10537, #10364.

[Jeroen Demeyer]

This patch gives a massive number of doctest failures.

[Rob Beezer]

The following existing (undocumented) behavior was broken by this patch, so needs to be reinstated or worked around. I thought this went through all a full test suite, but I guess not.

sage: z=ComplexField().an_element()
sage: z
1.00000000000000*I
sage: vector(z)
(0.000000000000000, 1.00000000000000)

[Dima Pasechnik]

Replying to rbeezer:

The following existing (undocumented) behavior was broken by this patch, so needs to be reinstated or worked around. I thought this went through all a full test suite, but I guess not.

 sage: z=ComplexField().an_element()
 sage: z
 1.00000000000000*I
 sage: vector(z)
 (0.000000000000000, 1.00000000000000)

jeez, what an awful kludge...

[Rob Beezer]

Replying to dimpase:

jeez, what an awful kludge...

Agreed. Worse that it was not documented, I think. And it causes about 80 doctest failures in one module, so there is probably no getting rid of it. So I'll figure out how it was happening in the first place, return the behavior, and document it.

Rob

[Dima Pasechnik]

Replying to rbeezer:

Replying to dimpase:

jeez, what an awful kludge...

Agreed. Worse that it was not documented, I think. And it causes about 80 doctest failures in one module, so there is probably no getting rid of it. So I'll figure out how it was happening in the first place, return the behavior, and document it.

IMHO, if this "feature" is limited to complex numbers only, it better be eradicated. It should be easy to track down and fix the library code that uses it.

Dima

Rob

[Rob Beezer]

Replying to dimpase:

IMHO, if this "feature" is limited to complex numbers only, it better be eradicated. It should be easy to track down and fix the library code that uses it.

I agree - in principle. But I also have my hands full with some other changes in behavior. ;-)

I'll do some tests soon to gauge the magnitude of the chore and we can go from there.

Rob

[Rob Beezer]

Here's the extent of converting complex numbers to vectors via the vector() constructor. Of course many of the failures are consequences of some earlier failure.

        sage -t  -force_lib pha2/devel/sage/doc/en/bordeaux_2008/elliptic_curves.rst # 3 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/crypto/mq/sbox.py # 5 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/crypto/mq/sr.py # 1 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/schemes/elliptic_curves/ell_egros.py # 8 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/schemes/elliptic_curves/period_lattice.py # 61 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/schemes/elliptic_curves/modular_parametrization.py # 8 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/schemes/elliptic_curves/ell_rational_field.py # 30 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/schemes/elliptic_curves/constructor.py # 7 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/schemes/elliptic_curves/heegner.py # 81 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/schemes/elliptic_curves/ell_point.py # 16 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/schemes/generic/toric_divisor.py # 36 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/plot/colors.py # 1 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/algebras/quatalg/quaternion_algebra.py # 3 doctests failed
        sage -t  -force_lib pha2/devel/sage/sage/geometry/cone.py # 1 doctests failed

[Dima Pasechnik]

Replying to rbeezer:

Replying to dimpase:

IMHO, if this "feature" is limited to complex numbers only, it better be eradicated. It should be easy to track down and fix the library code that uses it.

I agree - in principle. But I also have my hands full with some other changes in behavior. ;-)

I'll do some tests soon to gauge the magnitude of the chore and we can go from there.

how about the following:

make vector(blah) check whether blah.vector() can be called, and call it, if it is there.

add vector() member function to the corresponding complex number class.

Dima

Rob

[Rob Beezer]

Replying to dimpase:

how about the following:

make vector(blah) check whether blah.vector() can be called, and call it, if it is there.

add vector() member function to the corresponding complex number class.

Yes, that's nice. Or even just call the member function every place it is needed, without adding more to the general vector constructor. That'd still require a fair bit of editing, but it would be mindless.

Rob

[John Palmieri]

apply on top of other patch

[John Palmieri]

Here's a patch which seems to fix the problem for me. I think the issue was this: some Sage objects (like complex numbers) can be converted to type "Sequence", and then those can be converted to vectors. But if those objects had no "len", then the patch would cause problems, so the "referee" patch tries to call "len", but ignores TypeErrors.

By the way, the "vector" function already checks to see if v._vector_() is defined, so another option would be to define a _vector_ method for complex numbers. This would fix many of the doctest errors, but not all of them, because some of them pass other objects (like generators) to "vector".

[John Palmieri]

To the patchbot: Apply both

• trac_10595-vector-constructor-empty-list.patch
• 10595-referee.patch

[Rob Beezer]

Referee patch does the job. Thanks, John. Dima and John have checked off on the original patch, and I'll check off on the referee patch.

Complex number behavior, and a bit more, will be documented on #10977.