Re: converting int 2 real and vice versa

[John W. Eaton]

On  3-Sep-1998, Daniel Heiserer <address@hidden> wrote:

| Unfortunately I don't know what it is as long as I don't look at the
| value. Else I could specify on the fread what I want to read. Probably
| the best thing is writing a C-code translating all the stuff into
| doubles with the right conversion:
| // As far as I know that works for IEEE-be and IEEE-le 
| // No idea about cray formats
| // all the values are represented by a 4byte-word.
| // read the 4byte-word into a signed integer my_int
| // then make the decision
| if (my_int>limit || my_int<-limit){
|       my_doub=equivalence(my_int);    
| }
| else{
|       my_doub=(double) my_int;
| }
| It would be nice to do something like that in octave.
| 
| Thanks for your help. Any more ideas?

Here is a possibility.  Read all the data as integer values.  Select
the ones that need to be converted to floats, write them (as integers)
to a temporary file, reread them as floats and then insert them back
into the array.  Obligatory-Famous-Last-Words: I've not tested this,
but I think it should work.

  fid = fopen (data_file, "r");
  data = fread (fid, Inf, "int");
  fclose (fid);
  idx = find (data >= -limit && data <= limit);
  tmp_file = tmpnam ();
  fid = fopen (tmp_file, "w+");
  fwrite (fid, data(idx), "int");
  frewind (fid);
  tmp_data = fread (fid, Inf, "float");
  fclose (fid);
  unlink (tmp_file);
  data(idx) = tmp_data;


jwe