Bug in derivatives of some functions to the index var

[rws]

sage: gen_laguerre(n,a,x).diff(a)
-gen_laguerre(n - 1, a + 1, x)
sage: gen_laguerre(n,a,x).diff(x)
-gen_laguerre(n - 1, a + 1, x)

sage: derivative(hermite(n,x),x)
2*n*hermite(n - 1, x)
sage: derivative(hermite(n,x),n)
2*n*hermite(n - 1, x)

so something is badly wrong with recognizing the diff. parameter. Note only hermite is implemented in Pynac.

[mafra]

I have a patch for this issue (at least for gen_laguerre, my copy of sage doesn't have a derivative method for hermite).

The bug was in handling the diff_param choice, the 'else' branch had the return value for when diff_param was x, but that means that anything that didn't match the ifs before would return the derivative wrt x from the 'else' branch.

The attached patch fixes this (and adds proper handling to laguerre as well)

[mafra]

Patch to fix issue with gen_laguerre derivative

[tscrim]

@mafra you should create and post a git branch of your patch to this ticket. Then one that is done, you can put your (real) name as the author and set it to needs review.

[mafra]

New commits:

​e6311e0

Fix derivative of gen_laguerre wrt non-implemented variables

[rws]

That is fine, thanks. Can you please add a doctest for the fixed behaviour? We add tests for every branch of possible computation.

BTW Pynac fixes also get Python doctests, partly in the ​https://github.com/pynac/sage branches (which get merged with the Pynac upgrade ticket into Sage), and partly with separate Sage tickets.

[rws]

The hermite fix is in ​https://github.com/pynac/pynac/commit/20222d00281b2ab30970a407a8fb3e8fee93e07a

[rws]

However, see #21655 for a later enhancement.

[git]

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

​5b1b51b	Fix derivative of gen_laguerre wrt non-implemented variables
​24b25a4	Add doctest to trac 21644
​2decb06	Merge branch 'u/mafra/bug_in_derivatives_of_some_functions_to_the_index_var' of git://trac.sagemath.org/sage into t/21644/bug_in_derivatives_of_some_functions_to_the_index_var

[mafra]

Replying to rws:

That is fine, thanks. Can you please add a doctest for the fixed behaviour? We add tests for every branch of possible computation.

Ok, I did that now. But I was not sure if I should have rebased the new patch into the previous one or not, and when I decided to push I got an error message saying that I needed to 'pull' first. After pulling an automatic merge was done, so now there are three commits in this branch.

Should I have rebased the doctest patch into the previous one? Would that require a 'forced' git push?

[rws]

The merge commit would not have been necessary if you had done git co t/21644/bug_in_derivatives_of_some_functions_to_the_index_var on your machine, added the doctest commt, and finally git trac push. But it's no heavy disaster.