rewrite diff

[latot]

Well, this time its only rewrite of diff to python, why? time of execution:
with the actual diff

>> f=(x-1)^10;
>> tic;diff(f);toc;
Elapsed time is 0.0878189 seconds.

with python:

>> f=(x-1)^10;
>> tic;diff(f);toc;
Elapsed time is 0.0254591 seconds.

i have some questions, reading this: https://es.mathworks.com/help/symbolic/diff.html
and this tests:

%! assert (isequal (diff(f,sym(2)), g))
%! assert (isequal (diff(f,sym(2),x), g))
...
%! assert (isequal (diff(f,1), g))
%! assert (isequal (diff(f,1,x), g))

if i read right diff(f,sym(2),x) should diff f in 2 grades in x, and then derive it again in x? because in the example diff(f,sym(2),x) its the same as diff(f,sym(2))...
i never used SMT so i don't know if that its right, but its doesn't feel it match with the doc.

Well Cya.

[latot]

as a note, i don't change that tests yet, for that reason travis fails.

[cbm755]

Re: diff(f, 2, x): I dimly recall SMT used to do it this way, and at least in R2014b it is still mentioned in help diff. I haven't checked if its supported in R2016---at least it doesn't seem to be mentioned in the link you sent. Maybe we could remove that "feature": you could file a new issue for that.

[cbm755]

I read release notes, I don't see it being removed from SMT... so maybe we leave it too for now.

[cbm755]

I think the bit about diff(f, 2, x) is resolved.

Re: speed: I don't think we should do this sort of thing unless we write a re-useable "symvar" call in Python. Its non trivial and there are also subtly like #550.

Shall I close this?