For the 10th pattern, there should be a way to catch the parse error and backtrack, then push a } and go again. I suppose the correct way to do this with MegaParsec is to make whatever the equivalent of satisfy is for working with custom token streams also perform this transformation as it goes. But I guess GHC is doing this layout insensitivity transformation online with the parsing. Initially I was trying to implement that between lexing and parsing and gave up because of that. The 10th pattern for L says to behave differently if the next token causes a parse error. Last time I read that wiki page was before reading the algorithm in the (98) Report. Is there a standard way to ask watchers what they want to see / if they have thoughts on how something should be done? I have a wiki page up with my current status, but that's about it. I also haven't worked on a big GitHub project outside of my organization before. Initially, it seems fast and more extensible. The information for token position is completely recorded in Alex's output stream, but the explicit whitespace and newlines are gone.Д«efore I go ahead using Parsec, I'm wondering if using MegaParsec is a good idea. However it doesn't reject infixl 1Д«ecause the indentation sensitivity behavior I've written doesn't check newlines (and it would be somewhat involved, but not necessarily difficult, to do so because the lexer eats the whitespace). I've been able to get it to properly parse forms like elem :: a -> -> Bool However Parsec lacks built-in indentation sensitive parsing functionality.

Initially I was planning on using Parsec, because I'm fairly familiar with it. I've built up the infrastructure for the frontend AST and for the lexer, and am working on the parser. I'm sure many of you saw my recent post about my Write You a Haskell project.

Haskell parsec