Two quick questions: - Is there a possible priority inversion because of the separate ordering into the categories of non-recursive, directly recursive, or mutually recursive? - How does this behave with both left and right recursion in the same rule?
I'm usually one to spout about a particular "desired" semantic but it's not always clear that what I prefer is possible in all circumstances without too much overhead. Sent from my mobile device. On Jun 23, 2012 10:51 AM, "Jean-Luc Delatre" <j...@kevembuangga.com> wrote: > Le Sat, 23 Jun 2012 10:24:18 -0700, > Peter Goodman <peter.good...@gmail.com> a écrit : > > > Can you describe your method? > > It's summarily described in the README > http://www.kevembuangga.com/lrtt/README > I can add the following: > Productions in a given rule are categorized as > - non recursive > - mutual recursive > - directly left recursive > They are tried in the above order (even if there is ZERO non recursive) > As soon as a production is fully matched every recursive case > is tried again and again until it fails, if more recursive cases are > still at the bottom of the list they are tried also in the same manner. > When no more recursive production can match for the rule the whole stack > of recursive calls is kept to be run, first in descending order > (to run the semantic actions in front of the recursion points) > then in ascending order to follow up with whatever comes after > the recursion points. > The limitation is that once a recursive production has been exhausted > it will not be tried again even is some other one further down the list > succeeded and reinstated the conditions for a successful match of > a production upward in the list. > This could be changed by restarting at the very beginning in each case > (at some cost) but this is debatable: > What is the INTENDED semantic that we want? > > Best regards, > jld > >
_______________________________________________ PEG mailing list PEG@lists.csail.mit.edu https://lists.csail.mit.edu/mailman/listinfo/peg
