Command-line Parsing Combinators

Half-baked talk, February 11, 2005 by Thomas Hallgren

• Slides
• Source code

Abstract

The Programatica Haskell front-end (PFE) has a command line interface that provides a lot of functionality and I have been looking for a nice way to program it. By nice, I mean that

• It should be possible describe the command line syntax in a modular/reusable/extensible way. PFE, like CVS, has a command line syntax with a number of subcommands, each with its own set of arguments and options. Other people should be able to add new subcommands to the command line interface by adding new modules and reusing existing modules unchanged.
• The implementation should be self-documenting, i.e., the implementation of the command-line interface should look similar to the user documentation.
• The documentation of the command line syntax should always be in sync with the implementation. To ensure this, I want the command line parser and the documentation to be generated from the same source.

Being a functional programmer, my choice is to use parsing combinators, since they satisfy most of the above goals. The only goal not satisfied by available parsing combinators is the ability to extract documentation as well as a parser from the same source, so in the talk, I describe my own parsing combinators that solve this problem (under some simplifying assumptions). They are currently used in the implementation of PFE.