No worries! Too bad the info was too late to help.
One clarification: my implementation wasn't memoizing, so we "disable the
memoization cache" (clear it, rather) just as part of responsible ressource
management -- it does not have semantic incidence.
And one question: how do you manage the cases where a rule is both left-
and right-recursive (typically a binary operator) using only the
memoization cache (since you say you do not use a block set -- which in my
implem is used to prevent right-recursion when we specify
left-associativity).
Here is a very crude example:
E = E + E
E = Integer
input: 1 + 2 + 3
- parse E at the start of input
- cache miss, match `Integer(1)`
- retry, cache hit: match `Integer (1) + E`
+ recursively parse E at the position of "2"
+ cach miss: match `Integer(2)`
+ retry, cache hit: parse `Integer(2) + E`
* recursively parse E at the position of "3"
* cache miss: match `Integer(3)`
* retry, cache hit: match `Integer(3)`
+ final match for E at the position of "2": `Integer(2) + Integer(3)`
- final match for E at the start of input: `Integer(1) + (Integer(2) +
Integer(3))`
So if right-recursion is not blocked, this yields a right-associative parse.
I guess you have some magic up your sleeve, but I'd be interested to know
what it is.
Cheers,
Nicolas LAURENT
On 14 May 2017 at 19:34, Juancarlo Añez <apal...@gmail.com> wrote:
>
> On Sun, May 14, 2017 at 6:38 AM, Nicolas Laurent <
> nicolas.laur...@uclouvain.be> wrote:
>
>> For another approach: http://norswap.com/pubs/sle2015.pdf
>>
>> It's a bit hard to say if the semantics are different from the succint
>> description,
>> but I don't expect them to be.
>>
>
> It's exactly the same algorithm, Nicolas, with these minor differences:
>
> - I found no need to turn off the memoization cache
> - The memoization cache is used instead of a *block* set
> - Associativity is always left when left recursing
>
> I'm (doubly) sorry I missed your paper in my search: for the omission
> per-se, and because it would have saved me a lot of thinking time.
>
> My apologies (to both of us). TatSu will credit your paper on its next
> release.
>
> Cheers!
>
> --
> Juancarlo *Añez*
>
_______________________________________________
PEG mailing list
PEG@lists.csail.mit.edu
https://lists.csail.mit.edu/mailman/listinfo/peg
