[Cuis-dev] No regex in parser

[Casey Ransberger]

Thanks for this!

> On Sep 25, 2019, at 11:13 PM, Thierry Goubier via Cuis-dev <cuis-dev at lists.cuis.st> wrote:
> 
> Hi Casey,
> 
> Le jeu. 26 sept. 2019 à 00:52, Casey Ransberger via Cuis-dev
> <cuis-dev at lists.cuis.st> a écrit :
>> 
>> Holy crap, below
>> 
>>> On Sep 25, 2019, at 3:40 PM, Thierry Goubier via Cuis-dev <cuis-dev at lists.cuis.st> wrote:
>>> 
>>> A chomskyian setup is 20 lines of code plus a table for LR(1), so it
>>> is very hard to handwrite a parser that is smaller than that.
>> 
>> DAMN SON. Show me these 20 lines of code. Show me this table.
> 
> Attached. A LR(1) is a push down automaton, with two synchronized
> stacks and just four possibles actions: shift, reduce, accept, error.
> Never backtracks, is deterministic, needs only a single look-ahead
> (and anything requiring more look-ahead can be converted to LR(1)).
> 
> It uses the tables generated by SmaCC, converted to a simpler,
> unoptimized form that requires only a single read from the state
> table. SmaCC, like others LR(1) generators, compress the tables
> requiring multiple reads to get the action and next state information.
> 
>> I shall endeavor to understand.
> 
> It's deceptively simple, and not simple. When I teach it, students
> have a hard time with the underlying theory...
> 
> The scanner part is even simpler, and is 3x faster than the compiled
> scanner that SmaCC generates.
> 
> Thierry
> 
>> —Casey
>> --
>> Cuis-dev mailing list
>> Cuis-dev at lists.cuis.st
>> https://lists.cuis.st/mailman/listinfo/cuis-dev
> <step.cs>-- 
> Cuis-dev mailing list
> Cuis-dev at lists.cuis.st
> https://lists.cuis.st/mailman/listinfo/cuis-dev