site-packages/dolfin/fem/formmanipulations.py correctly insists on Function but is poorly documented

[Nico Schlömer]

When using derivative(F, u), Dolfin insists for u to be a Function. When looking at the code though, I see that the only factual requirement for u is to provide the .function_space() which is also the case for, e.g., TestFunction objects and maybe some others. This is how I got bugged by this: I provided a TestFunction and was disturbed by an error message and a "third argument" (?).

The derivative() method would profit from using the try-catch logic. Also, the error message must be improved, and the third argument better explained in the documentation.

[Martin Sandve Alnæs]

1) Why would you want to pass a TestFunction? It would fail at the ufl level if dolfin allowed it.

2) The "third argument" is du (see the signature), i.e. the variation: derivative(F, u, du) := d/dt F(u+t*du)|_t=0

3) The documentation is bad. I consider this a documentation bug only.

[Nico Schlömer]

1) Why would you want to pass a TestFunction? It would fail at the ufl level if dolfin allowed it.

Ah, I thought it was whatever symbol is used in F to indicate the argument. -- I believe I don't quite understand the interface yet. Documentation++.

[Martin Sandve Alnæs]

It's explained in some detail in the UFL chapter in the book, or in the new UFL paper: http://arxiv.org/abs/1211.4047

[Prof Garth Wells]

Could this be fixed so the Issue can be closed? Seems like an easy fix.

[Martin Sandve Alnæs]

At least improved in ufl commit 09586b568738546c94984279e598939fde100ec9

[Prof Garth Wells]

Removing milestone: 1.4 (automated comment)