[fpc-pascal] parser combinator library
noreply at z505.com
noreply at z505.com
Tue Feb 28 20:56:03 CET 2017
On 2017-02-24 15:00, Bernd wrote:
> Eight years ago someone asked whether there is a parser combinator
> library for free pascal, nothing like that existed at that time and
> also does not seem to exist up to the present day.
> While I was reading about parser combinators in functional programming
> languages (during my 42nd attempt to learn Haskell) I thought to
> myself why not try to implement something like that in Object Pascal,
> just so see how far we can push the boundaries of this imperative
> object oriented language.
> This is what I have come up with so far:
> Since we don't have lambdas I choose the next closest approach to
> emulate them with object instances instead. This leads to a lot of
> boiler plate in the definition of the elementary parsers and
> combinators but fortunately it can all be hidden away in the library
> and the usage of the combinators looks quite neat:
> // define the grammar
> EXPR := Num or _PARENS;
> MULFUNC := Sym('mul') and EXPR and EXPR;
> ADDFUNC := Sym('add') and EXPR and EXPR;
> INNER := MULFUNC or ADDFUNC or Num;
> PARENS := Sym('(') and INNER and Sym(')');
> Please also note the unorthodox usage of and/or operators to invoke
> the combinators :-)
Cool! Parsing is my favourite area of computing science..
And my 42 attempt at learning haskell I gave up because it seemed not to
be able to do a simple char by char parser like any procedural language
But, now that you've mentioned this, my interest has peaked in both
Object Pascal and Haskell.
> Please post improvements or variations of this theme, especially I am
> interested in how to properly build up a syntax tree in the most
> generic and reusable manner, this is something I have not yet
> completely understood (because I am myself still quite unfamiliar with
> this whole parsing business) currently all my parsers only return
> arrays of strings that I can optionally post-process with an optional
> hook function after a parser has completed.
I think parsing is still something computing science has to study more.
There are many undiscovered techniques. The biggest issue I have seen is
large long case statements where you start duplicating code in the case
statement. Rob Pike has done some work on this, but I've had little time
to research his solution proposed.
More information about the fpc-pascal