improve doctest coverage of libs/mpmath/ext_main.pyx

[mvngu]

As the subject says. As of Sage 4.4, the doctest coverage of sage/libs/mpmath/ext_main.pyx is:

[mvngu@sage mpmath]$ sage -coverage ext_main.pyx 
----------------------------------------------------------------------
ext_main.pyx
ERROR: Please add a `TestSuite(s).run()` doctest.
SCORE ext_main.pyx: 0% (0 of 102)

Missing documentation:
 * __cinit__(ctx):
 * default(ctx):
 * _get_prec(ctx):
 * _set_prec(ctx, prec):
 * _set_dps(ctx, n):
 * _get_dps(ctx):
 * _get_prec_rounding(ctx):
 * mpf make_mpf(ctx, tuple v):
 * mpc make_mpc(ctx, tuple v):
 * _convert_param(ctx, x):
 * _wrap_libmp_function(ctx, mpf_f, mpc_f=None, mpi_f=None, doc="<no doc>"):
 * _wrap_specfun(cls, name, f, wrap):
 * __init__(self, mpf_f, mpc_f=None, mpi_f=None, doc="<no doc>"):
 * __call__(self, x, **kwargs):
 * __init__(self, name, f):
 * __call__(self, *args, **kwargs):
 * __richcmp__(self, other, int op):
 * __add__(self, other):
 * __sub__(self, other):
 * __mul__(self, other):
 * __div__(self, other):
 * __mod__(self, other):
 * __pow__(self, other, mod):
 * ae(s, t, rel_eps=None, abs_eps=None):
 * __hash__(self):
 * __repr__(self):
 * __str__(self):
 * real(self):
 * imag(self):
 * conjugate(self):
 * man(self):
 * exp(self):
 * bc(self):
 * __int__(self):
 * __long__(self):
 * __float__(self):
 * __complex__(self):
 * to_fixed(self, prec):
 * __getstate__(self):
 * __setstate__(self, val):
 * __init__(self, x=0, **kwargs):
 * __reduce__(self):
 * _get_mpf(self):
 * _set_mpf(self, v):
 * __nonzero__(self):
 * __hash__(self):
 * real(self):
 * imag(self):
 * conjugate(self):
 * man(self):
 * exp(self):
 * bc(self):
 * to_fixed(self, long prec):
 * __int__(self):
 * __float__(self):
 * __getstate__(self):
 * __setstate__(self, val):
 * __cinit__(self):
 * __neg__(s):
 * __pos__(s):
 * __abs__(s):
 * sqrt(s):
 * __richcmp__(self, other, int op):
 * __init__(self, func, name, docname=''):
 * __call__(self, prec=None, dps=None, rounding=None):
 * _mpf_(self):
 * __repr__(self):
 * __nonzero__(self):
 * __neg__(self):
 * __pos__(self):
 * __abs__(self):
 * sqrt(self):
 * to_fixed(self, prec):
 * __getstate__(self):
 * __setstate__(self, val):
 * __hash__(self):
 * __richcmp__(self, other, int op):
 * __init__(self, real=0, imag=0):
 * __cinit__(self):
 * __reduce__(self):
 * __setstate__(self, val):
 * __repr__(self):
 * __str__(s):
 * __nonzero__(self):
 * __complex__(self):
 * _get_mpc(self):
 * _set_mpc(self, tuple v):
 * real(self):
 * imag(self):
 * __hash__(self):
 * __neg__(s):
 * conjugate(s):
 * __pos__(s):
 * __abs__(s):
 * __richcmp__(self, other, int op):

Missing doctests:
 * convert(ctx, x, strings=True):
 * isnan(ctx, x):
 * isinf(ctx, x):
 * isint(ctx, x):
 * fsum(ctx, terms, bint absolute=False, bint squared=False):
 * fdot(ctx, A, B=None):
 * mag(ctx, x):

[fredrik.johansson]

Argh. Who wrote this code?

The actual code coverage, indirectly through mpmath's unit tests (which however aren't run automatically by Sage) is close to 99%, but writing Sage doctests slipped my mind entirely.

Some of these functions have doctests, but they're added at runtime when importing mpmath. I'm considering whether statically duplicating them is a good idea, or whether it's better to just add static placeholders.

Most of these are probably trivial to write, e.g.:

sage: import mpmath
sage: mpmath.mpf(1).__add__(1)
mpf('2.0')

But I can't say when I'll have time to do this.

[schilly]

While I started to write some tests, I stumbled about to_fixed(). What does it do? Would be

sage: import mpmath
sage: mpmath.mpf('88494.93').to_fixed(6)
5663675

be a doctest or a total abuse?!

[fredrik.johansson]

Thanks a *lot*, Harald!

to_fixed converts a number to binary fixed-point format. Perhaps for clarity something like the following could be used (1.25 is for exactness):

sage: mpmath.mpf(1.25).to_fixed(10)
1280
sage: int(1.25 * 2^10)
1280

[schilly]

ok, i've done some more but i don't know how to really test init, cinit, call and some others.

[schilly]

I have another question, is this a bug:

sage: mpmath.mpf(1) < 3
True
sage: 1 < mpmath.mpf(3)
False
sage: 4 == mpmath.mpf(4)
False

[schilly]

I don't know how to test the remaining ones, but shouldn't be that hard to finish. Only 16 are still to go.

[fredrik.johansson]

Thanks! Very nice work.

The comparisons are definitely a bug. I created #8924 for this.

[fredrik.johansson]

Here is a new patch with almost full coverage: ​http://sage.math.washington.edu/home/fredrik/mpmath_doctests.patch

(I did not use Harald's old patch as a base, though I think that one will mostly still be working. Perhaps they could be combined.)

The functions still missing doctests are pickling support for classes that shouldn't be pickable (not sure why they are there).

[was]

review this, not Harald's

[davidloeffler]

Apply mpmath_doctests.patch

(for patchbot)

[davidloeffler]

Wow -- this is a monumental piece of doctesting! It looks great. This will get us a long way towards the 90% doctest coverage goal.

[jdemeyer]

This obviously fails on 32-bit systems:

sage -t  --long -force_lib devel/sage/sage/libs/mpmath/ext_main.pyx
**********************************************************************
File "/export/home/buildbot/build/sage/hawk-1/hawk_full/build/sage-5.0.beta10/devel/sage-main/sage/libs/mpmath/ext_main.pyx", line 2023:
    sage: int(7.25 * 2**30)
Expected:
    7784628224
Got:
    7784628224L
**********************************************************************

[davidloeffler]

apply over previous patch

[davidloeffler]

Apply mpmath_doctests.patch, trac_8791-fix.patch

This is a one-line fix -- anyone willing to quickly sign off on it?

[jdemeyer]

Good!