Fail to integrate sqrt(x^2)/x

[rws]

sage: integrate(sqrt(x^2)/x,x)
...
RuntimeError: ECL says: Error executing code in Maxima: expt: undefined: 0 to a negative exponent.

sage: integrate(sqrt(x^2)/x,x,algorithm='fricas')
sqrt(x^2)
sage: integrate(sqrt(x^2)/x,x,algorithm='giac')
x*sign(x)
sage: integrate(sqrt(x^2)/x,x,algorithm='sympy')
sqrt(x^2)

See ​Maxima bug 3657.

[chapoton]

Here is a fix (bandaid). One should report upstream to maxima.

---

New commits:

​744d626

fix some details in integration, make one more integral work

[chapoton]

bots are morally green, please review

[kcrisman]

Thanks, Frédéric. Can I ask whether the changes other than the new doctest and the addition of RunTimeError are nontrivial? I don't think so, but there were a lot of prettification changes.

[kcrisman]

In vanilla Maxima:

(%i4) domain:complex;
(%o4)                               complex
(%i5) integrate(sqrt(x^2)/x,x);

expt: undefined: 0 to a negative exponent.
 -- an error. To debug this try: debugmode(true);

[kcrisman]

However, before giving positive review, I'd suppose we'd want a way to check that this one was fixed - maybe # known bug below it where we require algorithm='maxima'?

[chapoton]

All the other changes are purely white space removal or addition, for the sake of flake8 conmpliance.

I guess one could a "known bug" doctest, yes.

[kcrisman]

Ah yes. I haven't used that, personally, but I'm sure it complains a lot. Unfortunately, it just makes tickets harder to review at times. I won't raise this on -devel because I know how annoyingly much extra work it would be, but having two different commits for that sort of thing is helpful to reviewers.

[chapoton]

Do you want the "known bug" doctest ? This does not seem to be really necessary to me. We are not responsible for maxima bugs.

[kcrisman]

It would be nice, because we typically do this in other cases. But I guess since in this case it is an actual exception raised, as opposed to a wrong result we had to work around, it is not necessary.

But if Volker complains about failing doctests, I am trusting your morally green patchbot :-) Just upgraded OS (still several versions behind) and so won't be building new Sage for a little while until I have time to check that command line tools are working properly.