On Oct 18, 2006, at 4:14 PM, hzhatlboro wrote:
I think I have sent my complete program to this list at least twice before.
Your point being...?
The part I sent to you today is directly related to a problem I get at the moment.
Directly related, perhaps, but too long and complicated by at least a factor of ten to illustrate the problem.
As I said before, I am now trying to get one thing working urgently and then I could compare the results with what I got from a FEM package.
You have so far spent three "urgent" days not getting the answer to your real question. Even if you got your approach to work, it would "urgently" take thousands or millions of times longer than it needs to (I'm not kidding). I have a very strong suspicion that if we knew what your ultimate goal was, we could have come up with a succinct expression in about five minutes. It is quite possible that FiPy already has a method to do the calculation that you ultimately need. I strongly doubt that this business of examining the values of cells and of the neighbors of cells is even necessary. There *must* be some underlying mathematical question involved, and it's quite likely that that mathematical question has already been addressed by FiPy.
At the beginning of my question today, I said "For structured meshes....". I do know that FiPy, which is based on Finite Volume Method, is designed for unstructured meshes. The example I am currently working on was developed on regular geometry and solved by Finite Difference Method. That's why my questions may seem "unreasonable" for FiPy users.
No, your questions are "unreasonable" because they focus on too low a level. If you want to write low-level finite difference code, by all means do so, but FiPy and Python are not good tools for you. The fact is that, unless you are writing in raw C or Fortran, there are almost always superior ways to express things like what you're trying to do, whether it be in Python, Java, Matlab, or what have you. FiPy (and tools like it) sits at an even higher level, and so the ways that you use it should be more abstract, not less.
As far as I can see, I have helped to identify several issues for FiPy during my communications with you during the past.
Indeed you have. I do not wish to discourage your interest in FiPy; only to help you understand that the way that questions are asked has an enormous bearing on how useful (or forthcoming) the answers are. I think that everybody (myself included) can benefit from a periodic review of <http://www.catb.org/~esr/faqs/smart-questions.html>.
