putcole.c

/*  This file, putcole.c, contains routines that write data elements to    */
/*  a FITS image or table, with float datatype.                            */

/*  The FITSIO software was written by William Pence at the High Energy    */
/*  Astrophysic Science Archive Research Center (HEASARC) at the NASA      */
/*  Goddard Space Flight Center.                                           */

#include <limits.h>
#include <string.h>
#include <stdlib.h>
#include "fitsio2.h"

/*--------------------------------------------------------------------------*/
int ffppre( fitsfile *fptr,  /* I - FITS file pointer                       */
            long  group,     /* I - group to write(1 = 1st group)           */
            LONGLONG firstelem, /* I - first vector element to write(1 = 1st)  */
            LONGLONG nelem,     /* I - number of values to write               */
            float *array,    /* I - array of values that are written        */
            int  *status)    /* IO - error status                           */
/*
  Write an array of values to the primary array. Data conversion
  and scaling will be performed if necessary (e.g, if the datatype of
  the FITS array is not the same as the array being written).

  This routine cannot be called directly by users to write to large
  arrays with > 2**31 pixels (although CFITSIO can do so by passing
  the firstelem thru a LONGLONG sized global variable)
*/
{
    long row;
    float nullvalue;

    /*
      the primary array is represented as a binary table:
      each group of the primary array is a row in the table,
      where the first column contains the group parameters
      and the second column contains the image itself.
    */

    if (fits_is_compressed_image(fptr, status))
    {
        /* this is a compressed image in a binary table */

        fits_write_compressed_pixels(fptr, TFLOAT, firstelem, nelem,
            0, array, &nullvalue, status);
        return(*status);
    }

    row=maxvalue(1,group);

    ffpcle(fptr, 2, row, firstelem, nelem, array, status);
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffppne( fitsfile *fptr,  /* I - FITS file pointer                       */
            long  group,     /* I - group to write(1 = 1st group)           */
            LONGLONG firstelem, /* I - first vector element to write(1 = 1st)  */
            LONGLONG nelem,     /* I - number of values to write               */
            float *array,    /* I - array of values that are written        */
            float nulval,    /* I - undefined pixel value                   */
            int  *status)    /* IO - error status                           */
/*
  Write an array of values to the primary array. Data conversion
  and scaling will be performed if necessary (e.g, if the datatype of the
  FITS array is not the same as the array being written).  Any array values
  that are equal to the value of nulval will be replaced with the null
  pixel value that is appropriate for this column.

  This routine cannot be called directly by users to write to large
  arrays with > 2**31 pixels (although CFITSIO can do so by passing
  the firstelem thru a LONGLONG sized global variable)
*/
{
    long row;
    float nullvalue;

    /*
      the primary array is represented as a binary table:
      each group of the primary array is a row in the table,
      where the first column contains the group parameters
      and the second column contains the image itself.
    */

    if (fits_is_compressed_image(fptr, status))
    {
        /* this is a compressed image in a binary table */

        nullvalue = nulval;  /* set local variable */
        fits_write_compressed_pixels(fptr, TFLOAT, firstelem, nelem,
            1, array, &nullvalue, status);
        return(*status);
    }

    row=maxvalue(1,group);

    ffpcne(fptr, 2, row, firstelem, nelem, array, nulval, status);
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffp2de(fitsfile *fptr,   /* I - FITS file pointer                     */
           long  group,      /* I - group to write(1 = 1st group)         */
           LONGLONG  ncols,      /* I - number of pixels in each row of array */
           LONGLONG  naxis1,     /* I - FITS image NAXIS1 value               */
           LONGLONG  naxis2,     /* I - FITS image NAXIS2 value               */
           float *array,     /* I - array to be written                   */
           int  *status)     /* IO - error status                         */
/*
  Write an entire 2-D array of values to the primary array. Data conversion
  and scaling will be performed if necessary (e.g, if the datatype of the
  FITS array is not the same as the array being written).

  This routine does not support writing to large images with
  more than 2**31 pixels.
*/
{
    /* call the 3D writing routine, with the 3rd dimension = 1 */

    ffp3de(fptr, group, ncols, naxis2, naxis1, naxis2, 1, array, status);

    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffp3de(fitsfile *fptr,   /* I - FITS file pointer                     */
           long  group,      /* I - group to write(1 = 1st group)         */
           LONGLONG  ncols,      /* I - number of pixels in each row of array */
           LONGLONG  nrows,      /* I - number of rows in each plane of array */
           LONGLONG  naxis1,     /* I - FITS image NAXIS1 value               */
           LONGLONG  naxis2,     /* I - FITS image NAXIS2 value               */
           LONGLONG  naxis3,     /* I - FITS image NAXIS3 value               */
           float *array,     /* I - array to be written                   */
           int  *status)     /* IO - error status                         */
/*
  Write an entire 3-D cube of values to the primary array. Data conversion
  and scaling will be performed if necessary (e.g, if the datatype of the
  FITS array is not the same as the array being written).

  This routine does not support writing to large images with
  more than 2**31 pixels.
*/
{
    long tablerow, ii, jj;
    long fpixel[3]= {1,1,1}, lpixel[3];
    LONGLONG nfits, narray;
    /*
      the primary array is represented as a binary table:
      each group of the primary array is a row in the table,
      where the first column contains the group parameters
      and the second column contains the image itself.
    */
           
    if (fits_is_compressed_image(fptr, status))
    {
        /* this is a compressed image in a binary table */
        lpixel[0] = (long) ncols;
        lpixel[1] = (long) nrows;
        lpixel[2] = (long) naxis3;
       
        fits_write_compressed_img(fptr, TFLOAT, fpixel, lpixel,
            0,  array, NULL, status);
    
        return(*status);
    }

    tablerow=maxvalue(1,group);

    if (ncols == naxis1 && nrows == naxis2)  /* arrays have same size? */
    {
      /* all the image pixels are contiguous, so write all at once */
      ffpcle(fptr, 2, tablerow, 1L, naxis1 * naxis2 * naxis3, array, status);
      return(*status);
    }

    if (ncols < naxis1 || nrows < naxis2)
       return(*status = BAD_DIMEN);

    nfits = 1;   /* next pixel in FITS image to write to */
    narray = 0;  /* next pixel in input array to be written */

    /* loop over naxis3 planes in the data cube */
    for (jj = 0; jj < naxis3; jj++)
    {
      /* loop over the naxis2 rows in the FITS image, */
      /* writing naxis1 pixels to each row            */

      for (ii = 0; ii < naxis2; ii++)
      {
       if (ffpcle(fptr, 2, tablerow, nfits, naxis1,&array[narray],status) > 0)
         return(*status);

       nfits += naxis1;
       narray += ncols;
      }
      narray += (nrows - naxis2) * ncols;
    }
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffpsse(fitsfile *fptr,   /* I - FITS file pointer                       */
           long  group,      /* I - group to write(1 = 1st group)           */
           long  naxis,      /* I - number of data axes in array            */
           long  *naxes,     /* I - size of each FITS axis                  */
           long  *fpixel,    /* I - 1st pixel in each axis to write (1=1st) */
           long  *lpixel,    /* I - last pixel in each axis to write        */
           float *array,     /* I - array to be written                     */
           int  *status)     /* IO - error status                           */
/*
  Write a subsection of pixels to the primary array or image.
  A subsection is defined to be any contiguous rectangular
  array of pixels within the n-dimensional FITS data file.
  Data conversion and scaling will be performed if necessary 
  (e.g, if the datatype of the FITS array is not the same as
  the array being written).
*/
{
    long tablerow;
    LONGLONG fpix[7], dimen[7], astart, pstart;
    LONGLONG off2, off3, off4, off5, off6, off7;
    LONGLONG st10, st20, st30, st40, st50, st60, st70;
    LONGLONG st1, st2, st3, st4, st5, st6, st7;
    long ii, i1, i2, i3, i4, i5, i6, i7, irange[7];

    if (*status > 0)
        return(*status);

    if (fits_is_compressed_image(fptr, status))
    {
        /* this is a compressed image in a binary table */

        fits_write_compressed_img(fptr, TFLOAT, fpixel, lpixel,
            0,  array, NULL, status);
    
        return(*status);
    }

    if (naxis < 1 || naxis > 7)
      return(*status = BAD_DIMEN);

    tablerow=maxvalue(1,group);

     /* calculate the size and number of loops to perform in each dimension */
    for (ii = 0; ii < 7; ii++)
    {
      fpix[ii]=1;
      irange[ii]=1;
      dimen[ii]=1;
    }

    for (ii = 0; ii < naxis; ii++)
    {    
      fpix[ii]=fpixel[ii];
      irange[ii]=lpixel[ii]-fpixel[ii]+1;
      dimen[ii]=naxes[ii];
    }

    i1=irange[0];

    /* compute the pixel offset between each dimension */
    off2 =     dimen[0];
    off3 = off2 * dimen[1];
    off4 = off3 * dimen[2];
    off5 = off4 * dimen[3];
    off6 = off5 * dimen[4];
    off7 = off6 * dimen[5];

    st10 = fpix[0];
    st20 = (fpix[1] - 1) * off2;
    st30 = (fpix[2] - 1) * off3;
    st40 = (fpix[3] - 1) * off4;
    st50 = (fpix[4] - 1) * off5;
    st60 = (fpix[5] - 1) * off6;
    st70 = (fpix[6] - 1) * off7;

    /* store the initial offset in each dimension */
    st1 = st10;
    st2 = st20;
    st3 = st30;
    st4 = st40;
    st5 = st50;
    st6 = st60;
    st7 = st70;

    astart = 0;

    for (i7 = 0; i7 < irange[6]; i7++)
    {
     for (i6 = 0; i6 < irange[5]; i6++)
     {
      for (i5 = 0; i5 < irange[4]; i5++)
      {
       for (i4 = 0; i4 < irange[3]; i4++)
       {
        for (i3 = 0; i3 < irange[2]; i3++)
        {
         pstart = st1 + st2 + st3 + st4 + st5 + st6 + st7;

         for (i2 = 0; i2 < irange[1]; i2++)
         {
           if (ffpcle(fptr, 2, tablerow, pstart, i1, &array[astart],
              status) > 0)
              return(*status);

           astart += i1;
           pstart += off2;
         }
         st2 = st20;
         st3 = st3+off3;    
        }
        st3 = st30;
        st4 = st4+off4;
       }
       st4 = st40;
       st5 = st5+off5;
      }
      st5 = st50;
      st6 = st6+off6;
     }
     st6 = st60;
     st7 = st7+off7;
    }
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffpgpe( fitsfile *fptr,   /* I - FITS file pointer                      */
            long  group,      /* I - group to write(1 = 1st group)          */
            long  firstelem,  /* I - first vector element to write(1 = 1st) */
            long  nelem,      /* I - number of values to write              */
            float *array,     /* I - array of values that are written       */
            int  *status)     /* IO - error status                          */
/*
  Write an array of group parameters to the primary array. Data conversion
  and scaling will be performed if necessary (e.g, if the datatype of
  the FITS array is not the same as the array being written).
*/
{
    long row;

    /*
      the primary array is represented as a binary table:
      each group of the primary array is a row in the table,
      where the first column contains the group parameters
      and the second column contains the image itself.
    */

    row=maxvalue(1,group);

    ffpcle(fptr, 1L, row, firstelem, nelem, array, status);
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffpcle( fitsfile *fptr,  /* I - FITS file pointer                       */
            int  colnum,     /* I - number of column to write (1 = 1st col) */
            LONGLONG  firstrow,  /* I - first row to write (1 = 1st row)        */
            LONGLONG  firstelem, /* I - first vector element to write (1 = 1st) */
            LONGLONG  nelem,     /* I - number of values to write               */
            float *array,    /* I - array of values to write                */
            int  *status)    /* IO - error status                           */
/*
  Write an array of values to a column in the current FITS HDU.
  The column number may refer to a real column in an ASCII or binary table, 
  or it may refer to a virtual column in a 1 or more grouped FITS primary
  array.  FITSIO treats a primary array as a binary table
  with 2 vector columns: the first column contains the group parameters (often
  with length = 0) and the second column contains the array of image pixels.
  Each row of the table represents a group in the case of multigroup FITS
  images.

  The input array of values will be converted to the datatype of the column 
  and will be inverse-scaled by the FITS TSCALn and TZEROn values if necessary.
*/
{
    int tcode, maxelem2, hdutype, writeraw;
    long twidth, incre;
    long ntodo;
    LONGLONG repeat, startpos, elemnum, wrtptr, rowlen, rownum, remain, next, tnull, maxelem;
    double scale, zero;
    char tform[20], cform[20];
    char message[FLEN_ERRMSG];

    char snull[20];   /*  the FITS null value  */

    double cbuff[DBUFFSIZE / sizeof(double)]; /* align cbuff on word boundary */
    void *buffer;

    if (*status > 0)           /* inherit input status value if > 0 */
        return(*status);

    buffer = cbuff;

    /*---------------------------------------------------*/
    /*  Check input and get parameters about the column: */
    /*---------------------------------------------------*/
    if (ffgcprll( fptr, colnum, firstrow, firstelem, nelem, 1, &scale, &zero,
        tform, &twidth, &tcode, &maxelem2, &startpos,  &elemnum, &incre,
        &repeat, &rowlen, &hdutype, &tnull, snull, status) > 0)
        return(*status);
    maxelem = maxelem2;

    if (tcode == TSTRING)   
         ffcfmt(tform, cform);     /* derive C format for writing strings */

    /*
       if there is no scaling and the native machine format is not byteswapped
       then we can simply write the raw data bytes into the FITS file if the
       datatype of the FITS column is the same as the input values.  Otherwise,
       we must convert the raw values into the scaled and/or machine dependent
       format in a temporary buffer that has been allocated for this purpose.
    */
    if (scale == 1. && zero == 0. && 
       MACHINE == NATIVE && tcode == TFLOAT)
    {
        writeraw = 1;
        if (nelem < (LONGLONG)INT32_MAX) {
            maxelem = nelem;
        } else {
            maxelem = INT32_MAX/4;
        }
     }
    else
        writeraw = 0;

    /*---------------------------------------------------------------------*/
    /*  Now write the pixels to the FITS column.                           */
    /*  First call the ffXXfYY routine to  (1) convert the datatype        */
    /*  if necessary, and (2) scale the values by the FITS TSCALn and      */
    /*  TZEROn linear scaling parameters into a temporary buffer.          */
    /*---------------------------------------------------------------------*/
    remain = nelem;           /* remaining number of values to write  */
    next = 0;                 /* next element in array to be written  */
    rownum = 0;               /* row number, relative to firstrow     */

    while (remain)
    {
        /* limit the number of pixels to process a one time to the number that
           will fit in the buffer space or to the number of pixels that remain
           in the current vector, which ever is smaller.
        */
        ntodo = (long) minvalue(remain, maxelem);      
        ntodo = (long) minvalue(ntodo, (repeat - elemnum));

        wrtptr = startpos + ((LONGLONG)rownum * rowlen) + (elemnum * incre);

        ffmbyt(fptr, wrtptr, IGNORE_EOF, status); /* move to write position */

        switch (tcode) 
        {
            case (TFLOAT):
              if (writeraw)
              {
                /* write raw input bytes without conversion */
                ffpr4b(fptr, ntodo, incre, &array[next], status);
              }
              else
              {
                /* convert the raw data before writing to FITS file */
                ffr4fr4(&array[next], ntodo, scale, zero,
                        (float *) buffer, status);
                ffpr4b(fptr, ntodo, incre, (float *) buffer, status);
              }

              break;

            case (TLONGLONG):

                ffr4fi8(&array[next], ntodo, scale, zero,
                        (LONGLONG *) buffer, status);
                ffpi8b(fptr, ntodo, incre, (long *) buffer, status);
                break;

            case (TBYTE):
 
                ffr4fi1(&array[next], ntodo, scale, zero, 
                        (unsigned char *) buffer, status);
                ffpi1b(fptr, ntodo, incre, (unsigned char *) buffer, status);
                break;

            case (TSHORT):

                ffr4fi2(&array[next], ntodo, scale, zero,
                        (short *) buffer, status);
                ffpi2b(fptr, ntodo, incre, (short *) buffer, status);
                break;

            case (TLONG):

                ffr4fi4(&array[next], ntodo, scale, zero,
                        (INT32BIT *) buffer, status);
                ffpi4b(fptr, ntodo, incre, (INT32BIT *) buffer, status);
                break;

            case (TDOUBLE):
                ffr4fr8(&array[next], ntodo, scale, zero,
                       (double *) buffer, status);
                ffpr8b(fptr, ntodo, incre, (double *) buffer, status);
                break;

            case (TSTRING):  /* numerical column in an ASCII table */

                if (cform[1] != 's')  /*  "%s" format is a string */
                {
                  ffr4fstr(&array[next], ntodo, scale, zero, cform,
                          twidth, (char *) buffer, status);

                  if (incre == twidth)    /* contiguous bytes */
                     ffpbyt(fptr, ntodo * twidth, buffer, status);
                  else
                     ffpbytoff(fptr, twidth, ntodo, incre - twidth, buffer,
                            status);

                  break;
                }
                /* can't write to string column, so fall thru to default: */

            default:  /*  error trap  */
                snprintf(message, FLEN_ERRMSG, 
                       "Cannot write numbers to column %d which has format %s",
                        colnum,tform);
                ffpmsg(message);
                if (hdutype == ASCII_TBL)
                    return(*status = BAD_ATABLE_FORMAT);
                else
                    return(*status = BAD_BTABLE_FORMAT);

        } /* End of switch block */

        /*-------------------------*/
        /*  Check for fatal error  */
        /*-------------------------*/
        if (*status > 0)  /* test for error during previous write operation */
        {
          snprintf(message,FLEN_ERRMSG,
          "Error writing elements %.0f thru %.0f of input data array (ffpcle).",
             (double) (next+1), (double) (next+ntodo));
         ffpmsg(message);
         return(*status);
        }

        /*--------------------------------------------*/
        /*  increment the counters for the next loop  */
        /*--------------------------------------------*/
        remain -= ntodo;
        if (remain)
        {
            next += ntodo;
            elemnum += ntodo;
            if (elemnum == repeat)  /* completed a row; start on next row */
            {
                elemnum = 0;
                rownum++;
            }
        }
    }  /*  End of main while Loop  */


    /*--------------------------------*/
    /*  check for numerical overflow  */
    /*--------------------------------*/
    if (*status == OVERFLOW_ERR)
    {
        ffpmsg(
        "Numerical overflow during type conversion while writing FITS data.");
        *status = NUM_OVERFLOW;
    }

    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffpclc( fitsfile *fptr,  /* I - FITS file pointer                       */
            int  colnum,     /* I - number of column to write (1 = 1st col) */
            LONGLONG  firstrow,  /* I - first row to write (1 = 1st row)        */
            LONGLONG  firstelem, /* I - first vector element to write (1 = 1st) */
            LONGLONG  nelem,     /* I - number of values to write               */
            float *array,    /* I - array of values to write                */
            int  *status)    /* IO - error status                           */
/*
  Write an array of complex values to a column in the current FITS HDU.
  Each complex number if interpreted as a pair of float values.
  The column number may refer to a real column in an ASCII or binary table, 
  or it may refer to a virtual column in a 1 or more grouped FITS primary
  array.  FITSIO treats a primary array as a binary table
  with 2 vector columns: the first column contains the group parameters (often
  with length = 0) and the second column contains the array of image pixels.
  Each row of the table represents a group in the case of multigroup FITS
  images.

  The input array of values will be converted to the datatype of the column
  if necessary, but normally complex values should only be written to a binary
  table with TFORMn = 'rC' where r is an optional repeat count. The TSCALn and
  TZERO keywords should not be used with complex numbers because mathmatically
  the scaling should only be applied to the real (first) component of the
  complex value.
*/
{
    /* simply multiply the number of elements by 2, and call ffpcle */

    ffpcle(fptr, colnum, firstrow, (firstelem - 1) * 2 + 1,
            nelem * 2, array, status);
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffpcne( fitsfile *fptr,  /* I - FITS file pointer                       */
            int  colnum,     /* I - number of column to write (1 = 1st col) */
            LONGLONG  firstrow,  /* I - first row to write (1 = 1st row)        */
            LONGLONG  firstelem, /* I - first vector element to write (1 = 1st) */
            LONGLONG  nelem,     /* I - number of values to write               */
            float *array,    /* I - array of values to write                */
            float  nulvalue, /* I - value used to flag undefined pixels     */
            int  *status)    /* IO - error status                           */
/*
  Write an array of elements to the specified column of a table.  Any input
  pixels equal to the value of nulvalue will be replaced by the appropriate
  null value in the output FITS file. 

  The input array of values will be converted to the datatype of the column 
  and will be inverse-scaled by the FITS TSCALn and TZEROn values if necessary
*/
{
    tcolumn *colptr;
    LONGLONG  ngood = 0, nbad = 0, ii;
    LONGLONG repeat, first, fstelm, fstrow;
    int tcode, overflow = 0;

    if (*status > 0)
        return(*status);

    /* reset position to the correct HDU if necessary */
    if (fptr->HDUposition != (fptr->Fptr)->curhdu)
    {
        ffmahd(fptr, (fptr->HDUposition) + 1, NULL, status);
    }
    else if ((fptr->Fptr)->datastart == DATA_UNDEFINED)
    {
        if ( ffrdef(fptr, status) > 0)               /* rescan header */
            return(*status);
    }

    colptr  = (fptr->Fptr)->tableptr;   /* point to first column */
    colptr += (colnum - 1);     /* offset to correct column structure */

    tcode  = colptr->tdatatype;

    if (tcode > 0)
       repeat = colptr->trepeat;  /* repeat count for this column */
    else
       repeat = firstelem -1 + nelem;  /* variable length arrays */

    if (abs(tcode) >= TCOMPLEX)
    { /* treat complex columns as pairs of numbers */
        repeat *= 2;
    }
    
    /* if variable length array, first write the whole input vector, 
       then go back and fill in the nulls */
    if (tcode < 0) {
      if (ffpcle(fptr, colnum, firstrow, firstelem, nelem, array, status) > 0) {
        if (*status == NUM_OVERFLOW) 
	{
	  /* ignore overflows, which are possibly the null pixel values */
	  /*  overflow = 1;   */
	  *status = 0;
	} else { 
          return(*status);
	}
      }
    }

    /* absolute element number in the column */
    first = (firstrow - 1) * repeat + firstelem;

    for (ii = 0; ii < nelem; ii++)
    {
      if (array[ii] != nulvalue)  /* is this a good pixel? */
      {
         if (nbad)  /* write previous string of bad pixels */
         {
            fstelm = ii - nbad + first;  /* absolute element number */
            fstrow = (fstelm - 1) / repeat + 1;  /* starting row number */
            fstelm = fstelm - (fstrow - 1) * repeat;  /* relative number */

            /* call ffpcluc, not ffpclu, in case we are writing to a
	       complex ('C') binary table column */
            if (ffpcluc(fptr, colnum, fstrow, fstelm, nbad, status) > 0)
                return(*status);

            nbad=0;
         }

         ngood = ngood +1;  /* the consecutive number of good pixels */
      }
      else
      {
         if (ngood)  /* write previous string of good pixels */
         {
            fstelm = ii - ngood + first;  /* absolute element number */
            fstrow = (fstelm - 1) / repeat + 1;  /* starting row number */
            fstelm = fstelm - (fstrow - 1) * repeat;  /* relative number */

            if (tcode > 0) {  /* variable length arrays have already been written */
              if (ffpcle(fptr, colnum, fstrow, fstelm, ngood, &array[ii-ngood],
                status) > 0) {
		if (*status == NUM_OVERFLOW) 
		{
		  overflow = 1;
		  *status = 0;
		} else { 
                  return(*status);
		}
              }
	    }
            ngood=0;
         }

         nbad = nbad +1;  /* the consecutive number of bad pixels */
      }
    }

    /* finished loop;  now just write the last set of pixels */

    if (ngood)  /* write last string of good pixels */
    {
      fstelm = ii - ngood + first;  /* absolute element number */
      fstrow = (fstelm - 1) / repeat + 1;  /* starting row number */
      fstelm = fstelm - (fstrow - 1) * repeat;  /* relative number */

      if (tcode > 0) {  /* variable length arrays have already been written */
        ffpcle(fptr, colnum, fstrow, fstelm, ngood, &array[ii-ngood], status);
      }
    }
    else if (nbad) /* write last string of bad pixels */
    {
      fstelm = ii - nbad + first;  /* absolute element number */
      fstrow = (fstelm - 1) / repeat + 1;  /* starting row number */
      fstelm = fstelm - (fstrow - 1) * repeat;  /* relative number */
      ffpcluc(fptr, colnum, fstrow, fstelm, nbad, status);
    }
    
    if (*status <= 0) {
      if (overflow) {
        *status = NUM_OVERFLOW;
      }
    }

    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffr4fi1(float *input,          /* I - array of values to be converted  */
            long ntodo,            /* I - number of elements in the array  */
            double scale,          /* I - FITS TSCALn or BSCALE value      */
            double zero,           /* I - FITS TZEROn or BZERO  value      */
            unsigned char *output, /* O - output array of converted values */
            int *status)           /* IO - error status                    */
/*
  Copy input to output prior to writing output to a FITS file.
  Do datatype conversion and scaling if required.
*/
{
    long ii;
    double dvalue;

    if (scale == 1. && zero == 0.)
    {       
        for (ii = 0; ii < ntodo; ii++)
        {
            if (input[ii] < DUCHAR_MIN)
            {
                *status = OVERFLOW_ERR;
                output[ii] = 0;
            }
            else if (input[ii] > DUCHAR_MAX)
            {
                *status = OVERFLOW_ERR;
                output[ii] = UCHAR_MAX;
            }
            else
                output[ii] = (unsigned char) input[ii];
        }
    }
    else
    {
        for (ii = 0; ii < ntodo; ii++)
        {
            dvalue = (input[ii] - zero) / scale;

            if (dvalue < DUCHAR_MIN)
            {
                *status = OVERFLOW_ERR;
                output[ii] = 0;
            }
            else if (dvalue > DUCHAR_MAX)
            {
                *status = OVERFLOW_ERR;
                output[ii] = UCHAR_MAX;
            }
            else
                output[ii] = (unsigned char) (dvalue + .5);
        }
    }
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffr4fi2(float *input,      /* I - array of values to be converted  */
            long ntodo,        /* I - number of elements in the array  */
            double scale,      /* I - FITS TSCALn or BSCALE value      */
            double zero,       /* I - FITS TZEROn or BZERO  value      */
            short *output,     /* O - output array of converted values */
            int *status)       /* IO - error status                    */
/*
  Copy input to output prior to writing output to a FITS file.
  Do datatype conversion and scaling if required.
*/
{
    long ii;
    double dvalue;

    if (scale == 1. && zero == 0.)
    {           
        for (ii = 0; ii < ntodo; ii++)
        {
            if (input[ii] < DSHRT_MIN)
            {
                *status = OVERFLOW_ERR;
                output[ii] = SHRT_MIN;
            }
            else if (input[ii] > DSHRT_MAX)
            {
                *status = OVERFLOW_ERR;
                output[ii] = SHRT_MAX;
            }
            else
                output[ii] = (short) input[ii];
        }
    }
    else
    {
        for (ii = 0; ii < ntodo; ii++)
        {
            dvalue = (input[ii] - zero) / scale;

            if (dvalue < DSHRT_MIN)
            {
                *status = OVERFLOW_ERR;
                output[ii] = SHRT_MIN;
            }
            else if (dvalue > DSHRT_MAX)
            {
                *status = OVERFLOW_ERR;
                output[ii] = SHRT_MAX;
            }
            else
            {
                if (dvalue >= 0)
                    output[ii] = (short) (dvalue + .5);
                else
                    output[ii] = (short) (dvalue - .5);
            }
        }
    }
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffr4fi4(float *input,      /* I - array of values to be converted  */
            long ntodo,        /* I - number of elements in the array  */
            double scale,      /* I - FITS TSCALn or BSCALE value      */
            double zero,       /* I - FITS TZEROn or BZERO  value      */
            INT32BIT *output,  /* O - output array of converted values */
            int *status)       /* IO - error status                    */
/*
  Copy input to output prior to writing output to a FITS file.
  Do datatype conversion and scaling if required.
*/
{
    long ii;
    double dvalue;

    if (scale == 1. && zero == 0.)
    {       
        for (ii = 0; ii < ntodo; ii++)
        {
            if (input[ii] < DINT_MIN)
            {
                *status = OVERFLOW_ERR;
                output[ii] = INT32_MIN;
            }
            else if (input[ii] > DINT_MAX)
            {
                *status = OVERFLOW_ERR;
                output[ii] = INT32_MAX;
            }
            else
                output[ii] = (INT32BIT) input[ii];
        }
    }
    else
    {
        for (ii = 0; ii < ntodo; ii++)
        {
            dvalue = (input[ii] - zero) / scale;

            if (dvalue < DINT_MIN)
            {
                *status = OVERFLOW_ERR;
                output[ii] = INT32_MIN;
            }
            else if (dvalue > DINT_MAX)
            {
                *status = OVERFLOW_ERR;
                output[ii] = INT32_MAX;
            }
            else
            {
                if (dvalue >= 0)
                    output[ii] = (INT32BIT) (dvalue + .5);
                else
                    output[ii] = (INT32BIT) (dvalue - .5);
            }
        }
    }
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffr4fi8(float *input,      /* I - array of values to be converted  */
            long ntodo,        /* I - number of elements in the array  */
            double scale,      /* I - FITS TSCALn or BSCALE value      */
            double zero,       /* I - FITS TZEROn or BZERO  value      */
            LONGLONG *output,  /* O - output array of converted values */
            int *status)       /* IO - error status                    */
/*
  Copy input to output prior to writing output to a FITS file.
  Do datatype conversion and scaling if required.
*/
{
    long ii;
    double dvalue;

    if (scale == 1. && zero ==  9223372036854775808.)
    {       
        /* Writing to unsigned long long column. Input values must not be negative */
        /* Instead of subtracting 9223372036854775808, it is more efficient */
        /* and more precise to just flip the sign bit with the XOR operator */

        for (ii = 0; ii < ntodo; ii++) {
            if (input[ii] < -0.49) {
              *status = OVERFLOW_ERR;
              output[ii] = LONGLONG_MIN;
            }
	    else if (input[ii] > 2.* DLONGLONG_MAX)
            {
                *status = OVERFLOW_ERR;
                output[ii] = LONGLONG_MAX;
            } else {
              output[ii] =  ((LONGLONG) input[ii]) ^ 0x8000000000000000;
            }
        }
    }
    else if (scale == 1. && zero == 0.)
    {       
        for (ii = 0; ii < ntodo; ii++)
        {
            if (input[ii] < DLONGLONG_MIN)
            {
                *status = OVERFLOW_ERR;
                output[ii] = LONGLONG_MIN;
            }
            else if (input[ii] > DLONGLONG_MAX)
            {
                *status = OVERFLOW_ERR;
                output[ii] = LONGLONG_MAX;
            }
            else
                output[ii] = (long) input[ii];
        }
    }
    else
    {
        for (ii = 0; ii < ntodo; ii++)
        {
            dvalue = (input[ii] - zero) / scale;

            if (dvalue < DLONGLONG_MIN)
            {
                *status = OVERFLOW_ERR;
                output[ii] = LONGLONG_MIN;
            }
            else if (dvalue > DLONGLONG_MAX)
            {
                *status = OVERFLOW_ERR;
                output[ii] = LONGLONG_MAX;
            }
            else
            {
                if (dvalue >= 0)
                    output[ii] = (LONGLONG) (dvalue + .5);
                else
                    output[ii] = (LONGLONG) (dvalue - .5);
            }
        }
    }
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffr4fr4(float *input,      /* I - array of values to be converted  */
            long ntodo,        /* I - number of elements in the array  */
            double scale,      /* I - FITS TSCALn or BSCALE value      */
            double zero,       /* I - FITS TZEROn or BZERO  value      */
            float *output,     /* O - output array of converted values */
            int *status)       /* IO - error status                    */
/*
  Copy input to output prior to writing output to a FITS file.
  Do datatype conversion and scaling if required.
*/
{
    long ii;

    if (scale == 1. && zero == 0.)
    {       
      memcpy(output, input, ntodo * sizeof(float) ); /* copy input to output */
    }
    else
    {
        for (ii = 0; ii < ntodo; ii++)
            output[ii] = (float) ((input[ii] - zero) / scale);
    }
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffr4fr8(float *input,      /* I - array of values to be converted  */
            long ntodo,        /* I - number of elements in the array  */
            double scale,      /* I - FITS TSCALn or BSCALE value      */
            double zero,       /* I - FITS TZEROn or BZERO  value      */
            double *output,    /* O - output array of converted values */
            int *status)       /* IO - error status                    */
/*
  Copy input to output prior to writing output to a FITS file.
  Do datatype conversion and scaling if required.
*/
{
    long ii;

    if (scale == 1. && zero == 0.)
    {       
        for (ii = 0; ii < ntodo; ii++)
                output[ii] = (double) input[ii];
    }
    else
    {
        for (ii = 0; ii < ntodo; ii++)
            output[ii] = (input[ii] - zero) / scale;
    }
    return(*status);
}
/*--------------------------------------------------------------------------*/
int ffr4fstr(float *input,     /* I - array of values to be converted  */
            long ntodo,        /* I - number of elements in the array  */
            double scale,      /* I - FITS TSCALn or BSCALE value      */
            double zero,       /* I - FITS TZEROn or BZERO  value      */
            char *cform,       /* I - format for output string values  */
            long twidth,       /* I - width of each field, in chars    */
            char *output,      /* O - output array of converted values */
            int *status)       /* IO - error status                    */
/*
  Copy input to output prior to writing output to a FITS file.
  Do scaling if required.
*/
{
    long ii;
    double dvalue;
    char *cptr;
    
    cptr = output;

    if (scale == 1. && zero == 0.)
    {       
        for (ii = 0; ii < ntodo; ii++)
        {
           sprintf(output, cform, (double) input[ii]);
           output += twidth;

           if (*output)  /* if this char != \0, then overflow occurred */
              *status = OVERFLOW_ERR;
        }
    }
    else
    {
        for (ii = 0; ii < ntodo; ii++)
        {
          dvalue = (input[ii] - zero) / scale;
          sprintf(output, cform, dvalue);
          output += twidth;

          if (*output)  /* if this char != \0, then overflow occurred */
            *status = OVERFLOW_ERR;
        }
    }

    /* replace any commas with periods (e.g., in French locale) */
    while ((cptr = strchr(cptr, ','))) *cptr = '.';
    
    return(*status);
}