Parsing examples

4 December 2014

My work on P1, E3, P3 and P4 was motivated in part by a desire to have simple tools that I could use in other areas of research. These tools are mostly source-to-source translators. For example, in the past I have had various ad-hoc tools to translate OCaml to LaTeX, or OCaml to HOL, HOL to Lem, Lem to OCaml etc.

The idea is to start making these tools maintainable and not one-off efforts to be used for a particular task then thrown away.

Previous versions of P1 etc have not been engineered well enough to make the resulting tools maintainable or usable. I am hoping to change that by investing more engineering effort into P1 etc.

So I have started a new repository on github https://github.com/tomjridge/example_grammars/ which contains grammars and grammar-based tools that I use.

To reduce grammar maintenance, I have added additional functionality to P1 etc:

  • Actions now use positional variables (x1,x2,...) rather than requiring a fun (x,(y,...)) -> nested tuple pattern. This makes maintenance easier and also allows p1 and p4 to use the same grammar files (p1 has a right-associative sequential combinator, while p4 has a left-associative seq. comb.). For example, the file https://github.com/tomjridge/example_grammars/blob/master/_todo/eee.p1 uses fun ... actions, whereas the file https://github.com/tomjridge/example_grammars/blob/master/_todo/eee2.p1x uses positional variables.

  • Explicitly-named symbols in actions are supported. In the following excerpt, the variables in the second alternative are w1,t,w2 rather than x1,x2,....

    OCAMLTYPEXPR -> "'" ?ident?            {{ Tex.tyvar ([x1;x2]|>ss_concat|>c) }}
  | "(" w1=?w? t=TYPEXPR w2=?w? ")"       {{ "("^(Tex.w w1)^t^(Tex.w w2)^")" }}

  • Multiple actions are now supported. For example, to map to tex you might have an action {tex{...}}, whilst to map to markdown, you might have an action {md{...}}. There is an example using multiple actions here: https://github.com/tomjridge/example_grammars/blob/master/_todo/eee3.p1x.acts

As a (somewhat largish) example, a grammar for a Lem/OCaml language is https://github.com/tomjridge/example_grammars/blob/master/_todo/ocaml.cppo Note that this can be used with p1 (for correctness/simplicity) and p4 (for speed). This is possible because p1 and p4 have essentially the same interface. At the moment there is only a single set of actions (to map to LaTeX). I am currently using this to produce typeset Lem code for inclusion in a paper.