fix NameError in mathematica_free integration

[github/bukzor]

It seems like the purpose was to normalize the variable of integration to x.

The line in question substitutes x for dvar rather than the variable of integration passed in, v.

​https://github.com/sagemath/sage/pull/37

[git]

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

​1b5e936

fix NameError in mathematica_free integration

[kcrisman]

Can you give an example of where this NameError occurs? It would make little sense to change this without somehow doctesting what it was that went wrong. In the most recent Sage devel version I get

sage: integral(sin(x)^2, x, algorithm='mathematica_free')
-1/2*cos(x)*sin(x) + 1/2*x

[github]

github/kcrisman wrote:

Hi! Let's continue the discussion at ​http://trac.sagemath.org/ticket/18212.

[kcrisman]

Can you give an example of where this NameError occurs? It would make little sense to change this without somehow doctesting what it was that went wrong. In the most recent Sage devel version I get

Figured it out.

sage: var('y')
y
sage: integral(sin(y)^2, y, algorithm='mathematica_free')
---------------------------------------------------------------------------
NameError: global name 'dvar' is not defined

[kcrisman]

I'm pushing a branch with doctests, and in doing this made a few more optional for internet so one doesn't have the optional=internet,sage problem as usual, then I replicated a different test elsewhere so it still gets tested ordinarily. Basically just needs review that this all works.

---

New commits:

​c254015	Merge branch 'u/github/ticket/18212' of git://trac.sagemath.org/sage into ticket/18212
​5078cf8	Doctest Trac 18212 for mma_free integration fix

[jdemeyer]

Why are these tests marked # optional?

sage: _ = var('x, y, z') # optional - internet
sage: f = sin(x^2) + y^z # optional - internet

[jdemeyer]

Same question here:

sage: from sage.symbolic.integration.external import mma_free_integrator # optional - internet

If importing the module requires internet, you're doing it wrong...

[jdemeyer]

And there is this oddity:

Test a few imports, without internet::

???

[kcrisman]

I'm pushing a branch with doctests, and in doing this made a few more optional for internet so one doesn't have the optional=internet,sage problem as usual, then I replicated a different test elsewhere so it still gets tested ordinarily.

I hope that this quote from comment:9 answers your questions. It's very annoying to do --optional=internet and then have tests fail because the "prereq" lines weren't evaluated. So that is my philosophy, to avoid --optional=sage,internet which I view as more than annoying to remember.

[jdemeyer]

Ah, I see now.

[jdemeyer]

If you really have problems with the fact that ./sage -t --optional=internet tests do not work, the proper fix would simply be to change the doctest framework (after sending a message to sage-devel of course). As far as I know, when we rewrote the doctest framework in #12415, we implemented --optional the way it is simply to emulate historical behaviour, I don't think it was a conscious choice.

[jdemeyer]

The fact that your patch actually needs to duplicate a doctest (once with # optional and once without) shows to me that your solution is not good.

[kcrisman]

Well, I only do that so that it's 100% sure to be doctested. Submit a change you like and I'll review it, this isn't that horrible in any case, the key is to fix the error.

[chapoton]

New commits:

​5d3207a

Merge branch 'u/kcrisman/ticket/18212' in 8.0.b5

[tscrim]

While I agree with Jeroen that we should have the minimal amount of # optional tags as needed, I see this as essentially bikeshedding. So I'm setting this to a positive review.