We have acknowledged our debt to Kernighan's PIC [4], and hopefully made clear how FPIC differs. There are quite a few languages for specifying pictures. We should particularly mention Timothy Van Zandt's PSTricks [9], a collection of TeX macros, because FPIC is implemented using them (a BitMap is actually just a sequence of PSTricks macro calls). Another is Kristoffer Rose's xyPic [8] package.
The closest relatives of this work are Chailloux and Suarez's mlPicTeX [2] and Simon Peyton Jones and Sigbjorn Finne's "simple structured graphics model" [7]. Both are embeddings of picture-drawing primitives in a functional language (ML and Haskell, respectively). In Peyton Jones and Finne's work, the type Picture contains abstract syntax trees of picture primitives; a program produces such a tree, and then a renderer traverses this tree and produces the picture.
We have emphasized in this paper the search for an appropriate definition of Picture, and we consider this an important step in the language design, but this is a philosophical issue. (Peyton Jones and Finne may also have considered this issue and then implemented the language as they did; they do not mention it. Chailloux and Suarez say nothing about what the type Picture is in their system.)
The substantive difference between FPIC and these other two systems is that FPIC has a naming facility for points and pictures that they lack. This comes directly from PIC. We think this is a significant difference, both because the facility is in fact used heavily in our examples (as it was in PIC) and because it represents the most interesting challenge in the language design.