Lazy power series: fix bad handling of base ring and categorification

[Florent Hivert]

Currently

sage: R = LazyPowerSeriesRing(QQ)
sage: type(R.gen().coefficient(0))
<type 'sage.rings.rational.Rational'>
sage: type(R.gen().coefficient(1))
<type 'int'>
sage: type(R.gen().coefficient(2))
<type 'int'>

it should be always Rational

I also fixed the following bug:

sage: from sage.combinat.species.series import LazyPowerSeriesRing
sage: L = LazyPowerSeriesRing(QQ)
sage: g = L.gen(); z = L.zero()
sage: s = z+g; s
Uninitialized lazy power series
sage: s.coefficients(10)
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
sage: s
Uninitialized lazy power series

It should be:

sage: s.coefficients(10)
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0]
sage: s
x + O(x^10)

[Florent Hivert]

The patch is not yet ready for review !

[Nicolas M. Thiéry]

Salut Florent,

There are some failing tests:

sage -t sage/combinat/tutorial.py  # 5 doctests failed
sage -t sage/tests/french_book/polynomes.py  # 7 doctests failed

Could you fix those or move the patch down the queue?

Thanks,

Nicolas

[Florent Hivert]

Fixed !

Florent

[Florent Hivert]

Fixed a multiline doctest to silence the patchbot. Ready for review.

Florent

[Darij Grinberg]

Any chance to get docstrings on _sum_generator_gen and/or sum_generator? Andrew and me are wondering what precisely these functions do in ​http://trac.sagemath.org/sage_trac/ticket/14542 . Thank you!

[Nicolas M. Thiéry]

I have just been through the path, and it looks good to go.

Just one thing: using the new syntax for example continuations. And since the whole file is touched anyway, we might as well do this change everywhere.

[Darij Grinberg]

There is a doctest failing, see ​http://patchbot.sagemath.org/log/13433/Fedora/18/x86_64/3.9.5-201.fc18.x86_64/desktop/2013-07-13%2021:47:59%20+0100?short . One way to fix it is to replace the _div_ method of the CycleIndexSeries class by a __div__ method, since I don't think that class belongs to any category which has a division method. On the other hand, this might not be what you want.

Ceterum censeo, _sum_generator_gen and sum_generator still need to be documented...

[Mike Hansen]

I've attached a review patch which takes care of these issues if someone would take a look at it.

[Mike Hansen]

I've attached a new patch which will apply on top of #14685. I've pushed these to the queue.

[Mike Hansen]

Apply trac_13433-lazy_power_serie_gen_fix-fh.patch trac_13433-review-mh.patch

[Darij Grinberg]

Alternative version of the BASE patch, which applies on my system. The review patch needs no changes, hence is not included.

[Darij Grinberg]

For patchbot:

apply

Apply trac_13433-base_patch-darijs_mod.patch​ trac_13433-review-mh.patch

[Darij Grinberg]

Mike, thank you for the docstring! I've understood it now. I've just attached my version of your review patch, which adds another example hopefully clarifying how to use sum_generator *in practice*. The example is a bit artificial because right now lazy power series can neither be divided nor subtracted(!!). I've also fixed another typo.

I must say I'm a bit surprised to see division (__div__) defined in generating_series.py but not in series.py. I am also surprised by the lack of _neg_ (or __neg__?) in both files, although one can use (-1) * u of course. Finally, subtraction of lazy power series does not seem to work (it ends in RuntimeError: maximum recursion depth exceeded in __instancecheck__).

For patchbot:

apply

Apply trac_13433-base_patch-darijs_mod.patch​ trac_13433-review-mh-darijs_mod.patch

[Darij Grinberg]

This patch breaks subtraction of lazy power series. I'm going to investigaate.

[Darij Grinberg]

OK, I have no idea what is broken and how to fix it, but it's clear that something is going wrong. All I know is that the bug is in the base file (trac_13433-base_patch-darijs_mod.patch​, or previously trac_13433-lazy_power_serie_gen_fix-fh.patch​).

[Mike Hansen]

A fix is in #15673

[Martin Rubey]

This is no longer relevant, since #32367.