Re: Writing Compilers in Functional Languages

Neelakantan Krishnaswami wrote:
>>I not sure about this. Appel's book uses updatable references in many
>>places, such as for name generation and hash tables. These are not
>>easily translated into Haskell. The same is true for exceptions.
>>Additionally, though SML and Haskell are both functional languages,
>>the "traditional" programming style for these is quite different - you
>>would tend to use type classes, list comprehensions and monads in
>>Haskell, and these are not found in SML. So translating a program
>>directly from SML to Haskell will not be trivial and the result will
>>be very atypical Haskell.
>
> This is a good point, and I'm changing my mind.
>
> When I read Appel's book, I was already decent at functional
> programming, and I translated on the fly from his imperative style to
> a pure style (using state passing style in ML). But if Kkaa wants to
> write a compiler to learn Haskell, then this advice won't work.


If it is of any interest to you, I am precisely working on porting
ML/OCaml code concerning this book to Haskell. You can see a brief
summary of this here :
http://www.haskell.org/communities/11-2004/html/report.html (§ 8.4)


Porting code is quite a bit of work, but not so much as what has been
said. I have partly suppressed the side-effects used by Appel, partly
translated them using the ST monad, which is quite straightforward in
fact. The resulting code works (that's the nice part of a port, no
extended testing is necessary), and is absolutely not atypical
Haskell. ML and especially OCaml are closer to Haskell that what I
thought at the beginning (but the port took me more time than I
thought -- but this is always true in C.S. ). Code available if you
like (and helping hand appreciated, because it is far from complete).


Alain

Return to the comp.compilers page.
Search the comp.compilers archives again.