Adding labels to PostScript figures

Here is a more concrete example of the problem. We start with a graph of the function y = x² for x = -1 to x=1 (click here to see the actual PostScript image):

We want to add to this a few labels to produce (click here to see the actual PostScript image):

The point is that the labels ought to be in TEX, as they are here.

We shall offer three ways to solve this problem. They differ in their answer to the questions: Where does text end and graphics begin? To what extent should one expect to poke around inside a graphics file? How much work is a well designed label in a figure worth?

• Overlaying the graphics file with TEX.
  • Advantages: It works with any conceivable graphics format. It guarantees that changes of fonts in the text itself will automatically lead to changes in the labels.
  • Disadvantages: Changes of size in the text are necessarily propagated to the labels. You have to move both files around together, and if you make changes in one you have to remember to make corresponding changes in the other. Difficult to write the correct macros to begin with, since TEX always seems to add and subtract space secretly.
• Making changes directly inside the original graphics file.
  • Advantages: It is very efficient, and gives the most control.
  • Disadvantages: You won't want to do this unless you know how to read and program in PostScript.
• Importing the figure file and PostScript files produced by TEX and dvips into a single larger file.
  • Advantages: Gives a reasonable amount of control over how the labels are embedded into the figure. Involves only a very small amount of PostScript programming to do this. The same technique applies to merge any two one-page PostScript figures into one.
  • Disadvantages: Produces somewhat bloated files. Probably involves at least some PostScript programming to embed the labels correctly into the figure. Requires more (but simpler) text-handling than the others.