reading fp-values written in Fortran D-format in C

[Daniel Franke]

[cross-posting to get views from both sides of the pond]

Hi all.

I have some trouble reading floating-point numbers written with the Fortran
D-descriptor:

$> cat dout.f90
write( *, FMT="(d13.5)") 3.1415
end

$> gfortran dout.f90 -o dout && ./dout
0.31415D+01

Reading above number via [f]scanf in C:

$> cat din.c
#include <stdio.h>
int main() {
double x;
fscanf(stdin, "%lf", &x);
fprintf(stdout, "%lf\n", x);
return 0;
}

$> gcc din.c -o din && ./dout | din
0.314150

While I can update the application that writes files in above format, I
still have to deal with existing files. Besides converting these files via
sed, is there anything in C (short of reimplementing/patching scanf) that
would allow me to read numbers in this format properly?

Thanks

Daniel
--
comp.lang.c.moderated - moderation address: cl...@plethora.net -- you must
have an appropriate newsgroups line in your header for your mail to be seen,
or the newsgroup name in square brackets in the subject line. Sorry.

[Glen Herrmannsfeldt]

Daniel Franke wrote:

> I have some trouble reading floating-point numbers written
> with the Fortran D-descriptor:

(snip)

> While I can update the application that writes files in above format, I
> still have to deal with existing files. Besides converting these files via
> sed, is there anything in C (short of reimplementing/patching scanf) that
> would allow me to read numbers in this format properly?

Similar to the suggestion often given for Fortran I/O, read the line
into a buffer with fgets, convert D to E, possibly with a STRCHR loop,
and then use sscanf to convert it.

That works pretty well if the file has a line (record) structure,
slightly less well if it doesn't.

-- glen

[Daniel Franke]

Glen Herrmannsfeldt wrote:

Glen,

thanks for your suggestion. Unfortunately, in my case it may work slightly
less well. Still worth to give a try :)

Regards

Daniel

P.S. Although you replied five days ago, I only got the message today.
Strange ...

[Dave Allured]

Daniel Franke wrote:
>
> [cross-posting to get views from both sides of the pond]
>
> Hi all.
>
> I have some trouble reading floating-point numbers written with the Fortran
> D-descriptor:
>
> $> cat dout.f90
> write( *, FMT="(d13.5)") 3.1415
> end
>
> $> gfortran dout.f90 -o dout && ./dout
> 0.31415D+01
>
> Reading above number via [f]scanf in C:
>
> $> cat din.c
> #include <stdio.h>
> int main() {
> double x;
> fscanf(stdin, "%lf", &x);
> fprintf(stdout, "%lf\n", x);
> return 0;
> }
>
> $> gcc din.c -o din && ./dout | din
> 0.314150
>
> While I can update the application that writes files in above format, I
> still have to deal with existing files. Besides converting these files via
> sed, is there anything in C (short of reimplementing/patching scanf) that
> would allow me to read numbers in this format properly?

Just read each input line as a string, change the D's to e's in the
string, then read the edited numbers from the string instead of with
fscanf. I do not know what the C function is to read floats from a
string; I am responding from the fortran group. I figure that should be
simple enough in C, though. HTH.

--Dave