Symbolic integration issue with abs_integrate (Maxima)

[Michael Orlitzky]

I started working on #11483, but I'm seeing one regression. Namely, this
used to work, and there's a doctest for it:

sage: f = function('f'); a,b=var('a,b')
sage: h = definite_integral(f(x), x,a,b)
sage: h.diff(x) # indirect doctest
0

With the new abs_integrate package, Maxima will try evaluate the
relation (a < b) and crash. It doesn't seem to do this for any other
integrals, and all other doctests are passing:

sage -t "devel/sage-abs_integrate/sage/symbolic/integration
/integral.py"
#0: lambda([a,b],is(a < b) = true)(a=a,b=b)
#1: partitions_interval_p(l=[],endpts=[a,b])
#2: abs_defint(e=f(x),x=x,lo=a,hi=b)
...
RuntimeError: ECL says: Error executing code in Maxima: Unable to
evaluate predicate a < b

Now, I can fix this particular doctest by adding an assumption that a<b,
but that could break some existing code.

Another solution would be to catch this exception, and make
the assumption for the user. I'm not sure that's very nice, though, and
we would need to be sure that this error only gets thrown where a<b can
be assumed WLOG.

Finally, we could wait for Maxima to fix it. Supposedly, it already is:

http://sourceforge.net/tracker/index.php?func=detail&aid=3153168&group_id=4933&atid=104933

(Can someone give me a hint how I can upgrade my Maxima spkg to test?)

What's the best way to proceed?

[Dima Pasechnik]

> Can someone give me a hint how I can upgrade my Maxima spkg to test?)

> What's the best way to proceed?

you can try replacing everything in maxima-5.23.2.p0/src/ with the new Maxima release, then rebuild it.  Have at look at http://www.sagemath.org/doc/developer/producing_spkgs.html and http://www.sagemath.org/doc/developer/patching_spkgs.html

Dima

[Michael Orlitzky]

Ah, thanks. The bug is fixed in Maxima 5.25.1. I guess I'll open a
ticket to bump it if all tests still pass.

[kcrisman]

On Nov 28, 12:29 pm, Michael Orlitzky <mich...@orlitzky.com> wrote:
> On 11/28/11 05:44, Dima Pasechnik wrote:
>
> >> Can someone give me a hint how I can upgrade my Maxima spkg to test?)
>
> >> What's the best way to proceed?
>
> > you can try replacing everything in maxima-5.23.2.p0/src/ with the new
> > Maxima release, then rebuild it.  Have at look
> > athttp://www.sagemath.org/doc/developer/producing_spkgs.html

> > andhttp://www.sagemath.org/doc/developer/patching_spkgs.html

>
> Ah, thanks. The bug is fixed in Maxima 5.25.1. I guess I'll open a
> ticket to bump it if all tests still pass.

You might want to first look at http://trac.sagemath.org/sage_trac/ticket/11966
and base on that spkg, to ensure we also work on Lion.

[Michael Orlitzky]

On 11/28/11 12:57, kcrisman wrote:
>
> You might want to first look at http://trac.sagemath.org/sage_trac/ticket/11966
> and base on that spkg, to ensure we also work on Lion.
>

Yup, I'm using the latest alpha which has this applied already. The
patches still apply cleanly to the new Maxima, but I've no way to test
if it builds on Lion.

[Michael Orlitzky]

On 11/28/2011 01:20 PM, Michael Orlitzky wrote:
> On 11/28/11 12:57, kcrisman wrote:
>>
>> You might want to first look at http://trac.sagemath.org/sage_trac/ticket/11966
>> and base on that spkg, to ensure we also work on Lion.
>>

Trac #12094 now.

I went with Maxima 5.24.0 because there's another unrelated regression
in 5.25.1 that I didn't want to deal with at the same time.

This lets us fix the abs_integrate stuff, and a couple of other
symbolics bugs.