PostScript, The Forgotten Art of Programming

The first step is to torture your raw data until you get a simple table. In practice you use Perl and friends for this step. For the sake of demonstration I will use a tiny table:

1993 9.0 8.6
1994 5.7 7.8
1995 6.4 7.1
1996 7.5 6.1
1997 8.4 5.9

This table has an x, a y and a z. What I wish to draw is a light gray histogram for x and y and a dark gray one for x and z. Normally, you know the minimum and maximum values in your table, or you just use an awk one-liner to determine those values.

Next step is get started using a template of the Perl script, histogram.pl, that will generate the PostScript file. This template is shown in Listing 1.

One remark about the last bounding box line. This is the size of A4 (European standard page size); for letter size you need 0 0 612 792. In a later stage we will change this line, so that the bounding box fits more tightly.

Run the script and save the output in histogram.ps. Start up Ghostview to view this file. Not much to see, right? Time to edit histogram.ps. It is easier to do a little experimentation with this file rather than making changes directly to the Perl file (especially in a later stage when you are actually processing your data). We are going to experiment with the axes; our changes are shown in Listing 2.

When you are pleased with the result, copy it into histogram.pl just after the save command and add the line 1 setlinewidth to restore the original line width. Now it is time to do the hard work: defining two subroutines Histo-y and Histo-z. Again, this normally requires some experimentation, so create the PostScript file and edit it. We will assume that each subroutine gets x,y and x,z respectively on the stack. We will give both histograms a border line. It often helps to put a couple of your data points on the stack as an experiment.

You can copy your subroutines just in front of the EndPrologue line of your Perl script as shown in Listing 3.

Just a few words: I warned you to avoid using variables, and I did not practice what I preached. Well, only in the case of huge tables do you do everything on the stack. Doing so is much harder and often not worth the effort—my time is more expensive than what I gain in speed. Furthermore, I do some of the computations in PostScript; usually, it does pay off to do this in your Perl script. Finally, you normally do not want to recompute the path; you save it. I just wanted to keep the example simple.

Now it is time to complete your Perl script and process your data by adding the lines:

while (<>) {
           chomp;
           ($x, $y, $z) = split;
           print "$x $y Histo-y $x $z Histo-z\n";
   }

Now you can run your script with the data as stdin to create a new histogram.ps. The final step is to determine a better bounding box. This is where Ghostview comes into play. Go to the leftmost and rightmost pixel of your picture and write down the x coordinates. Now do the same for the top and bottom of your picture, writing the y values. With these coordinates you can determine the bounding box (it does not have to be pixel fit) 83 85 400 405, and you can change it in your PostScript file. (Or in your Perl script; however, if you have a huge data file to process, recreating the PostScript file can take a while.)

Now you have a fully level-1 compliant PostScript file less than 2KB is size that you can actually understand. I have seen PostScript files generated by applications under MS-DOS that need 2MB for the same picture. The complete Perl script and output PostScript are included in the gzipped tar file on the ftp site as Listings 4 and 5. The output histogram is shown in Figure 1.

Epilogue

So from now on we do everything in PostScript, right? Wrong. If it goes faster using another application and the generated PostScript file is not too large, use that application. For many pictures I still use xfig or something similar. Use PostScript directly if your data set is big and importing your data into the application already requires a lot of work. If you are relatively new to PostScript, concentrate on x-y graphics and histograms. If you have gained some experience, read the Blue and Red Books. Most importantly, have fun.

References