Re: C-pointers from Mathematica
- To: mathgroup at smc.vnet.net
- Subject: [mg113075] Re: C-pointers from Mathematica
- From: Bruno Silva <bfasilva at gmail.com>
- Date: Tue, 12 Oct 2010 04:30:41 -0400 (EDT)
Albert and Daniel,
Thank you for your replies.
The string pattern matching trick really does the job (and showed me several
mathematica typical expressions that I was not aware of... thanks for that
too!).
Concerning pointer arithmetic, I haven't been programming since long but it
was an habit I acquired in the past (if I well remember, that would allow to
manage allocated memory dynamically... correct me if my wrong).
Thanks and good luck,
Bruno
On Sat, Oct 9, 2010 at 11:33 AM, Albert Retey <awnl at gmx-topmail.de> wrote:
> Hi,
>
> > I want to use Mathematica to transform a nonlinear PDE to its equivalent
> > finite difference expression under a particular discretization scheme,
> and
> > finally export the result to a C source file.
> > I manage to do the first part of the job but the problem arrives when
> > translating Mathematica variable to C-program pointers.
> >
> > I want to transform T[x] -> * (profile + i), meaning that the value of
> "T"
> > at point "x" (inside Mathematica) is equivalent to the "i-th" value in
> > pointer "profile" (inside file.c).
> > Because Mathematica doesn't allow me to use the C-pointer sign * that
> way, I
> > thought replacing T[x] by a string would solve the problem (
> > T[x]->"*(profile+i)" ) but this does not work because, as soon as I apply
> > CForm to the result, the string symbol "..." is kept inside the
> expression I
> > would like to write in file.c :
> >
> > T[x]/.T[x] -> "*(profile+i)"
> > CForm[%]
> >
> > Does anyone know how to overcome this difficulty?
>
> I have done similar things, and for everything where the syntax is just
> not available in Mathematica, I found it easiest to work with string
> pattern matching of the resulting code. Here is an example:
>
> expr = T[5] - T[4]
>
> StringReplace[
> ToString[CForm[expr]],
> {
> "T(" ~~ x : (Except[")"] ..) ~~ ")" :> "*(profile + " <> x <> ")"
> }
> ]
>
> On the other hand I don't really see that why you want to address your
> array with pointer arithmetic, from all I remember from C you could just
> as well use array syntax, which would probably be easier generated from
> a Mathematica expression -- and create C code that is easier to
> read/debug. Here is an example of that approach (you need to change T[x]
> to T[[x]] to get T[x] instead of T(x) in CForm):
>
> CForm[expr /. T[x_] :> T[[x]]]
>
> You might want to suppress the warning messages with Quiet or by
> preventing the evaluation of the expression with something like this:
>
> CForm[Apply[HoldForm,{expr}] /. T[x_] :> T[[x]]]
>
>
> hth,
>
> albert
>
>