[Haskell-cafe] Why GHC is written in Happy and not a monadic parser library?

[Malcolm Wallace]

On 3 Aug 2013, at 21:03, Jason Dagit wrote:

> Another con of using parsec that I forgot to mention in my previous
> email is that with Parsec you need to be explicit about backtracking
> (use of try). Reasoning about the correct places to put try is not
> always easy and parsec doesn't help you with the task. In my
> experience, this is the main bug that people run into when using
> parsec.

Although the original question did not mention parsec explicitly, I find it disappointing that many people immediately think of it as the epitome of monadic combinator parsing.  The power of good marketing, eh?  There are so many other good parsing libraries out there.  Parsec happened to cure some known space-leaks in rival libraries about the time of its release (2000 or so), but the main reason it is popular is simply because it was distributed alongside ghc for a long time.

Curiously enough, the complaint you make about parsec is exactly the same as the observation that drove the development of my own set of combinators - polyparse.  But the concept of commitment to a partial parse (to prevent backtracking) was already present somewhat in Röjemo's applicative parsers way back in 1994.  He had both `ap` and `apCut` (the naming of "cut" borrowed from Prolog I suppose).  Space performance and the elimination of space leaks was totally his focus: he mentions being able to recompile the compiler itself in just 3Mb of RAM.

Regards,
    Malcolm