/*
*
* postmd - matrix display program for PostScript printers.
*
* A simple program that can be used to display a matrix as a gray scale image on
* a PostScript printer using the image operator. Much of the code was borrowed
* from postdmd, the bitmap display program DMD screen dumps. May help if you have
* a large matix (of floating point numbers) and want a simple way to look for
* patterns.
*
* Matrix elements are a series of floating point numbers arranged in the input
* file in row major order. The actual matrix elements can be preceeded by a simple
* header that sets things like the matrix dimensions, interval list, and possibly
* a window into the matrix that we'll use for display. The dimension statement is
* perhaps the most important. If present it determines the number of rows and
* columns in the matrix. For example, either of the following defines a 50x50
* matrix,
*
* dimension 50
* dimension 50x50
*
* If no dimension statement appears in the input file, the matrix is assumed to
* be square, and the number of rows (and columns) is set to the square root of
* the number of elements in the input file.
*
* Each matrix element is mapped into an integer in the range 0 to 255 (actually
* 254) and PostScript's image operator then maps that number into a gray scale
* appropriate for the particular printer. The mapping from the floating point
* matrix elements to integers is accomplished using an interval list that can be
* set using the -i option. The format of the interval string is,
*
* num1,num2,num3,...,numn
*
* where each num is a floating point number. The list must be given in increasing
* numerical order. A list of n numbers partitions the real line into 2n+1 regions
* given as,
*
* region1 element < num1
* region2 element = num1
* region3 element < num2
* region4 element = num2
* .
* .
* .
* region2n element = numn
* region2n+1 element > numn
*
* Every number in a region is mapped one integer in the range 0 to 254, and that
* number, when displayed on a printer using the image operator, prints as a square
* filled with a gray shade that reflects the integer that was chosen. 0 maps to
* black and 255 maps to white (which by default will not be used).
*
* The default gray scale gets darker as the region number increases, but can be
* changed by supplying a gray scale list with the -g option or in the optional
* matrix header. The color map is again a comman or space separated list that
* looks like,
*
* color1,color2, ... ,color2n+1
*
* where color1 applies to region 1 and color2n+1 applies to region2n+1. Each
* number in the list should be an integer between 0 and 255. If less than 2n+1
* colors are given default assignments will be used for missing regions.
*
* The size of the matrix that we can display reasonably well is a function of the
* number of elements in the interval list, paper size, and printer resolution.
* For example a 300dpi printer using 8.5x11 inch paper gives us an image area of
* about 2400x2400 pixels. An interval list of two numbers generates five separate
* regions and will therefore need that many different shades of gray. Since we're
* not using white we'll need to partion our image area into 4x4 pixel squares,
* and that means a 600x600 matrix is about as big as we can go. In practice that's
* optimistic, but the argument illustrates some of the limitations.
*
* A submatrix can be selected to display by windowing into the matrix. The window
* list can be given using the -w option or can be set in the optional header that
* can preceed each matrix. The list should be a comma or space separated list
* that looks like,
*
* lower-column, lower-row, upper-column, upper-row
*
* where each element in the list must be a positive integer. Rows and columns in
* the input matrix start at 1. The dimension of the displayed window will be from
* lower-column to upper-column and from lower-row to upper-row inclusive.
*
* The encoding produced by the program is essentially identical to what's done
* by postdmd. See the comments at the beginning of that program if you need more
* details. The prologue also shares much of the same code.
*
* The PostScript prologue is copied from *prologue before any of the input files
* are translated. The program expects that the following PostScript procedures
* are defined in that file:
*
* setup
*
* mark ... setup -
*
* Handles special initialization stuff that depends on how this program
* was called. Expects to find a mark followed by key/value pairs on the
* stack. The def operator is applied to each pair up to the mark, then
* the default state is set up.
*
* pagesetup
*
* page pagesetup -
*
* Does whatever is needed to set things up for the next page. Expects
* to find the current page number on the stack.
*
* bitmap
*
* columns rows bitmap -
*
* Prints the image that's read as a hex string from standard input. The
* image consists of rows lines, each of which includes columns elements.
* Eight bits per pixel are used to encode the matrix elements.
*
* labelmatrix
*
* matrixname matrixlimits labelmatrix -
*
* Prints string matrixname just below the lower left corner of the image
* and prints string martixlimits near the lower right corner. Outlines
* the entire image with a (one pixel wide) box and then draws tick marks
* along the top and left sides of the image. One tick mark is printed
* for every ten elements.
*
* legend
*
* n1 ... nN N c1 m1 ... cM mM total regions legend -
*
* Prints the legend as a bar graph below the matrix image. n1 ... nN are
* strings that represent the interval list. c1 m1 ... cm mM are pairs
* that consist of a region's color and the statistics count. Actually
* the c's are trivial procedures that just leave a one character string
* on the stack when they're executed by image - which is the way the
* bar graph is drawn.
*
* done
*
* done
*
* Makes sure the last page is printed. Only needed when we're printing
* more than one page on each sheet of paper.
*
* Many default values, like the magnification and orientation, are defined in
* the prologue, which is where they belong. If they're changed (by options), an
* appropriate definition is made after the prologue is added to the output file.
* The -P option passes arbitrary PostScript through to the output file. Among
* other things it can be used to set (or change) values that can't be accessed by
* other options.
*
*/
int formsperpage = 1; /* page images on each piece of paper */
int copies = 1; /* and this many copies of each sheet */
int bytespp = 6; /* bytes per pattern - on output */
int dostats = ON; /* permanent statistics flag */
int nxtstat = ON; /* and the one for the next matrix */
char *interval = DFLTILIST; /* string representations of the interval */
char *colormap = NULL; /* color map */
char *window = NULL; /* and window lists */
char *matrixname = "pipe.end"; /* name for the next plot */
Ilist ilist[128]; /* active interval list and color map */
int next = 0; /* one past the last element in ilist[] */
int regions; /* an index assigned to the last region */
int wlist[4]; /* upper left and lower right corners */
int page = 0; /* last page we worked on */
int printed = 0; /* and the number of pages printed */
int dfltrows = 0; /* default rows */
int dfltcols = 0; /* and columns - changed by -d option */
int rows; /* real number of rows */
int columns; /* and columns in the matrix */
int patcount = 0; /* will be set to columns * rows */
double element; /* next matrix element */
char *raster = NULL; /* next raster line */
char *rptr; /* next free byte in raster */
char *eptr; /* one past the last byte in raster */
FILE *fp_in = stdin; /* read from this file */
FILE *fp_out = stdout; /* and write stuff here */
FILE *fp_acct = NULL; /* for accounting data */
/*
*
* Bitmap display program for matrices. Only one matrix is allowed per input file,
* and each one will be displayed on a page by itself. Input files consist of an
* optional header followed by floating point numbers that represent the matrix
* elements - in row major order.
*
*/
argc = agc; /* other routines may want them */
argv = agv;
prog_name = argv[0]; /* really just for error messages */
init_signals(); /* sets up interrupt handling */
header(); /* PostScript header comments */
options(); /* handle the command line options */
setup(); /* for PostScript */
arguments(); /* followed by each input file */
done(); /* print the last page etc. */
account(); /* job accounting data */
int ch; /* return value from getopt() */
int old_optind = optind; /* for restoring optind - should be 1 */
/*
*
* Scans the option list looking for things, like the prologue file, that we need
* right away but could be changed from the default. Doing things this way is an
* attempt to conform to Adobe's latest file structuring conventions. In particular
* they now say there should be nothing executed in the prologue, and they have
* added two new comments that delimit global initialization calls. Once we know
* where things really are we write out the job header, follow it by the prologue,
* and then add the ENDPROLOG and BEGINSETUP comments.
*
*/
while ( (ch = getopt(argc, argv, optnames)) != EOF )
if ( ch == 'L' )
prologue = optarg;
else if ( ch == '?' )
error(FATAL, "");
optind = old_optind; /* get ready for option scanning */
/*
*
* Reads and processes the command line options. Added the -P option so arbitrary
* PostScript code can be passed through. Expect it could be useful for changing
* definitions in the prologue for which options have not been defined.
*
*/
while ( (ch = getopt(argc, argv, optnames)) != EOF ) {
switch ( ch ) {
case 'a': /* aspect ratio */
fprintf(stdout, "/aspectratio %s def\n", optarg);
break;
case 'b': /* bytes per pattern - on output */
bytespp = atoi(optarg);
break;
/*
*
* Makes sure all the non-option command line arguments are processed. If we get
* here and there aren't any arguments left, or if '-' is one of the input files
* we'll process stdin.
*
*/
if ( argc < 1 )
matrix();
else { /* at least one argument is left */
while ( argc > 0 ) {
matrixname = *argv;
if ( strcmp(*argv, "-") == 0 ) {
fp_in = stdin;
matrixname = "pipe.end";
} else if ( (fp_in = fopen(*argv, "r")) == NULL )
error(FATAL, "can't open %s", *argv);
matrix();
if ( fp_in != stdin )
fclose(fp_in);
argc--;
argv++;
} /* End while */
} /* End else */
/*
*
* Finished with all the input files, so mark the end of the pages, make sure the
* last page is printed, and restore the initial environment.
*
*/
int count; /* pattern repeats this many times */
long total; /* expect this many patterns */
/*
*
* Reads a matrix from *fp_in, translates it into a PostScript gray scale image,
* and writes the result on stdout. For now only one matrix is allowed per input
* file. Matrix elements are floating point numbers arranged in row major order
* in the input file. In addition each input file may contain an optional header
* that defines special things like the dimension of the matrix, a window into
* the matrix that will be displayed, and an interval list.
*
* If we're reading from stdin we first make a copy in a temporary file so we can
* can properly position ourselves after we've looked for the header. Originally
* wasn't always making a copy of stdin, but I've added a few things to what's
* accepted in the header and this simplifies the job. An alternative would be
* to always require a header and mark the end of it by some string. Didn't like
* that approach much - may fix things up later.
*
*/
if ( fp_in == stdin ) /* make a copy so we can seek etc. */
copystdin();
rows = dfltrows; /* new dimensions for the next matrix */
columns = dfltcols;
buildilist(interval); /* build the default ilist[] */
addcolormap(colormap); /* add the colormap - if not NULL */
setwindow(window); /* and setup the initial matrix window */
nxtstat = dostats; /* want statistics? */
getheader(); /* matrix dimensions at the very least */
dimensions(); /* make sure we have the dimensions etc. */
int fd_out; /* for the temporary file */
int fd_in; /* for stdin */
int buf[512]; /* buffer for reads and writes */
int count; /* number of bytes put in buf */
/*
*
* If we're reading the matrix from stdin and the matrix dimension isn't set by
* a dimension statement at the beginning of the file we'll copy stdin to a
* temporary file and reset *fp_in so reads come from the temp file. Simplifies
* reading the header (if present), but is expensive.
*
*/
if ( temp_file != NULL ) /* been here already */
unlink(temp_file);
char buf[512]; /* temporary string space */
char *cmap = NULL; /* remember header colormap list */
long pos; /* for seeking back to first element */
/*
*
* Looks for the optional header information at the beginning of the input file,
* reads it if it's there, and sets *fp_in to be just past the header. That should
* be the beginning of the matrix element list. The recognized header keywords are
* dimension, interval, colormap (or grayscale), window, name, and statistics. All
* are optional, but may be useful in a spooling environment when the user doesn't
* doesn't actually run the translator.
*
* The dimension statement specifies the number of rows and columns. For example
* either of the following two lines define a 50 by 50 element matrix,
*
* dimension 50
* dimension 50x50
*
* The first integer is the number of rows and the second, if given, is the number
* of columns. If columns are missing from the dimension statement we assume the
* matrix is square.
*
* interval can be used to redefine the interval list used for mapping floating
* point numbers into integers in the range 0 to 254. The string following the
* interval keyword has the same format as the -i option. For example to set the
* interval list to -1, 0, and 1 you can add the line,
*
* interval -1,0,1
*
* The numbers are floats given in increasing order, and separated by commas or
* blanks. The last interval list in a header takes precedence.
*
* colormap can be used to redefine the grayscale list. The string following
* the colormap keyword has the same format as the -g option. For example
*
* colormap 0,50,100,150,200,250
* or grayscale 0,50,100,150,200,250
*
* The window keyword can be used to select a submatrix. The numbers following
* window are the upper left and lower right matix coordinates. May not be
* implemented yet but shouldn't be difficult. For example
*
* window 10 10 40 40
*
* selects the submatrix with corners at (10, 10) and (40, 40). The edges of the
* window are included in the display.
*
* The name keyword can be used to define the title of the display. For example,
*
* name Plot Of Matrix 1
*
* prints the string "Plot Of Matrix 1" at the top of the page. Everything up to
* the next newline is taken as the name string.
*
*/
char buf[100]; /* temporary storage for the elements */
long count = 0; /* number of elements in the matrix */
long pos; /* matrix elements start here */
/*
*
* Need to know the dimensions of the matrix before we can go any farther. If
* rows and columns are still 0 we'll read the entire input file, starting from
* the current position, count the number of elements, take the square root of it,
* and use it as the number of rows and columns. Then we seek back to the start
* of the real matrix, make sure columns is set, and allocate enough memory for
* storing each raster line. After we're certain we've got the number of rows and
* columns we check the window coordinates, and if they're not legitimate they're
* reset to cover the entire matrix.
*
*/
if ( rows == 0 ) {
pos = ftell(fp_in);
while ( fscanf(fp_in, "%s", buf) != EOF )
count++;
rows = sqrt((double) count);
fseek(fp_in, pos, 0);
} /* End if */
if ( columns <= 0 ) columns = rows;
if ( raster != NULL ) free(raster);
if ( (rptr = raster = malloc(columns)) == NULL )
error(FATAL, "no memory");
static char *templist = NULL; /* a working copy of the list */
char *ptr; /* next number in *templist */
int i; /* loop index - for checking the list */
/*
*
* Reads string *list and builds up the ilist[] that will be used in the next
* matrix. Since strtok() modifies the string it's parsing we make a copy first.
* The format of the interval list is described in detail in the comments at the
* beginning of this program. Basically consists of a comma or space separated
* list of floating point numbers that must be given in increasing numerical order.
* The list determines how floating point numbers are mapped into integers in the
* range 0 to 254.
*
*/
if ( templist != NULL ) /* free the space used by the last list */
free(templist);
while ( isascii(*list) && isspace(*list) )
list++;
for ( ptr = list, regions = 3; *ptr != '\0'; ptr++ ) {
if ( *ptr == ',' || *ptr == '/' || isspace(*ptr) )
regions += 2;
while ( isascii(*ptr) && isspace(*ptr) ) ptr++;
} /* End for */
static char *templist = NULL; /* a working copy of the color list */
char *ptr; /* next color in *templist */
int i = 0; /* assigned to this region in ilist[] */
/*
*
* Assigns the integers in *list to the color field for the regions defined in
* ilist[]. Assumes ilist[] has already been setup.
*
*/
if ( list != NULL ) {
if ( templist != NULL )
free(templist);
templist = savestring(list);
ptr = strtok(templist, ",/ \t\n");
while ( ptr != NULL ) {
ilist[i++].color = atoi(ptr) % 256;
ptr = strtok(NULL, ",/ \t\n");
} /* End while */
} /* End if */
char *list; /* corners of window into the matrix */
{
static char *templist = NULL; /* a working copy of the window list */
char *ptr; /* next window coordinate in *templist */
int i = 0; /* assigned to this region in wlist[] */
/*
*
* Sets up an optional window into the matrix.
*
*/
wlist[0] = wlist[1] = 1;
wlist[2] = wlist[3] = 0;
if ( list != NULL ) {
if ( templist != NULL )
free(templist);
templist = savestring(list);
ptr = strtok(templist, ",/ \t\n");
while ( ptr != NULL ) {
wlist[i++] = atoi(ptr);
ptr = strtok(NULL, ",/ \t\n");
} /* End while */
} /* End if */
double element; /* floating point matrix element */
{
int i; /* loop index */
/*
*
* Maps element into an integer in the range 0 to 255, and returns the result to
* the caller. Mapping is done using the color map that was saved in ilist[]. Also
* updates the count field for the region that contains element - not good!
*
*/
for ( i = 1; i < next && ilist[i].val < element; i += 2 ) ;
char *p1, *p2; /* starting and ending columns */
int n; /* set to bytes per pattern */
int i; /* loop index */
/*
*
* Takes the scanline that's been saved in *raster, encodes it according to the
* value that's been assigned to bytespp, and writes the result to *fp_out. Each
* line in the output bitmap is terminated by a 0 on a line by itself.
*
*/
char *p1; /* first patterns starts here */
int n; /* and extends this many bytes */
{
char *p2; /* address of the second pattern */
/*
*
* Compares the two n byte patterns *p1 and *(p1+n). FALSE if returned is they're
* different or extend past the end of the current raster line.
*
*/
p2 = p1 + n;
for ( ; n > 0; n--, p1++, p2++ )
if ( p2 >= eptr || *p1 != *p2 )
return(FALSE);
