RE: problem with fem-fenics

[Marco Vassallo]

> Date: Mon, 22 Jul 2013 17:18:32 +0200

> Subject: Re: problem with fem-fenics
> From: address@hidden
> To: address@hidden
> CC: address@hidden
>
> On Mon, Jul 22, 2013 at 4:29 PM, Marco Vassallo
> <address@hidden> wrote:
> > I would really appreciate also if you could tell me if I can see somewhere
> > this code that you have produced in the past.
>
> Here you have an example done in Python that manipulates strings, it
> is slow but it works. Python is very well designed to do this kind of
> things.
> http://ubuntuone.com/7I2kMksCiSCysEqelr5WUN
> Ask me if there is something that is not clear.

Hi Jpi,
Thanks for the example!
Could you also tell me how can I run a small example which uses your code?
Because I'm having some trouble using it :-)
In fact I think that if I see the output I can also understand
better how things work.

> If you want to do it in Octave, I would suggest doing some C++ parser
> using for example tinyxml (1 or 2, though 2 seems more adequate)
> http://www.grinninglizard.com/tinyxml2/index.html or any other DOM
> structure. You could use Octave structs, but I think that could be
> quite painful and un-maintainable. You could also define your own
> class, though again, seems like re-inventing the wheel. Remeber that
> you are creating and interface, therefore maintainability should have
> high priority, because things will change!
>

Looking at your example, I agree with you that Python seems to be really
better for this problem,
so I will try to use it instead of Octave or C++.

Thanks again

marco

> Basically you need the static data (all of your file that will
> never-ever change dynamically). That can be hard-coded into your
> parser.
> Then you need to have an object/function that fills in the xml DOM
> with the dynamic data. Then you have a XML2C method/function that is
> in charge of creating the file with the right name and code. A method
> that finds out how to compile the file and generates the corresponding
> command would also be useful. For these you need to check for features
> of the OS, I wouldn't recommend doing this natively form Octave but
> rather use python or M4 or any other language meant to do this.
>
> Finally, do not over do it. Think of ecological-balance, the problem
> description file is small, therefore your first version of the
> parser/generator should be small as well. Just spend some time
> thinking on maintainability and potential extension.
>
> Another final comment, :D, among us there are true computer scientist
> that have worked on automatic-code generation
> Here is a link to the discussion I had some time ago (at the end I did
> it with python)
> http://octave.1599824.n4.nabble.com/Automatic-code-generation-td3990003.html
> Try asking them directly (keeping list in CC), for ideas suggestions.
>
> I hope this helps.
>
> JPi