Dave Bayer wrote:
I've been playing with parsing, for text filtering applications such as an alternate GHC literate preprocessor. Here, it pays to have one's monad handle both the input and output streams out of sight. Using ShowS-valued monads, one can express most grammatical constructs as simple composition. I'm sure many people have had this idea, but the resulting parsers that I write end up much shorter than any demo code I've seen. For example, the "code is indented, comments are flush, periods delimit block comments" literate preprocessor that I depend on daily (leaving out the heredoc code) is just
dot, comment, dotLine, commentLine, codeLine, dotBlock, delit ∷ Parser
dot = char '.' comment = place "-- "
codeLine = white ∘ till (heredoc ∨ whiteLine) word commentLine = comment ∘ line
dotLine = comment ∘ dot ∘ whiteLine dotBlock = dotLine ∘ till (dotLine ∨ eof) (whiteLine ∨ commentLine)
delit = till (skip (many whiteLine) ∘ eof) (whiteLine ∨ dotBlock ∨ codeLine ∨ commentLine)
This looks intriguing! Can you elaborate on how this works? Regards, Heinrich Apfelmus -- http://apfelmus.nfshost.com