A Clojure platform for generative programming.
At the moment there are two representation schemes for Push programs: plush and bb8. Both of these genomic schemes address the problem of representing a tree-like Push program (tree-like because it can contain arbitrary code blocks surrounded by matching parentheses) into a linear genome that can be translated back into the same program. Each comes at this problem from a different direction, though.
The plush genome is concerned with where and how many parenthesized code blocks are opened and closed, in the course of a depth-first traversal of the program tree. When the linear genome is transcribed into a Push program, a buffer of potential branches is created, and consumed as additional genes are translated.
By convention in the original Clojush , plush genomes refer to deep information about the syntax of the Push language itself, constructing branches in the program whenever instructions that would be expected to use those branches appear in the code. So for example because the Push instruction :exec-swap uses two :exec arguments, the Clojush plush convention would be to automatically and immediately open two consecutive branches following the insertion of an :exec-swap instruction. The implementation here permits this behavior through a backwards-compatible interface described below.
The bb8 genome takes a more abstract and generative approach. Each bb8 gene contains instructions that indicate how to move a “cursor” to a new position on the developing Push program tree, and how to insert a new item in that cursor’s new position. Unlike plush genomes, the tree structure of the program is explicitly captured; branches are inserted by inserting an empty code block in the developing tree, which can then be “filled” by later transcribed genes.