Well, like I posted, you can convert to Diff notation, just not diff notation. I think the problem lies in the terms you're taking the derivative with respect to. If you look at the result in Diff notation, it substitutes variables t1 and t2 for x-y and z-y respectively and then evaluates t1 and t2 at those points. In the non-inert case (diff notation). it will try to evaluate at x-y and z-y and Maple can't carry out that derivative directly.

What does it look like printed? I find that looking at graphs on the computer can be misleading, especially with some of the previewers. I only noticed the aliasing at certain zoom levels which suggests it may just be an artifact of looking at it on the screen. Oh, in the future you'll want to put a link to the files in your post. I only found the links by looking at the recent files so you might want to edit the post to include links.

If you want to convert D notation to Diff, just use convert. convert(Dexpr,Diff); results in, -(eval(Diff(f(t1, z-y), t1), t1 = x-y))-(eval(Diff(f(x-y, t2), t2), t2 = z-y)) It doesn't appear that Maple wants to convert to the non-inert (diff) form, probably because of the derivatives with respect to the differences.

Oh, that's what I was looking for. I'll just have to apply T(x) to each term where I have T(x) applied to a derivative. The only trick is detecting which terms this has to apply to in the general case. The simplest thing to do I think is make the diff inert inside the T and then find all inert diff terms. I've put together something I think will work. I just haven't tried it on larger problems. View 62_linalg_experiment2.mw on MapleNet or Download 62_linalg_experiment2.mw
View file details The only flaw is that in general a is a vector and so when you try to evaluate diff(g(a),a) with a specific vector a, Maple chokes. So, as long as g(a) and h(a) are substituted in and the derivative is fine, all is well.