Le 2011-08-02 à 03:22, Alek Paunov a écrit : > I am using Cloud 9 IDE [1] locally and probably will try to achieve the same > next year (if the core team not have accomplished the task in the following > months :-)) using XML documents as parse trees, LPEG based parser and LPEG VM > stacks recovery as I tried to outline in the previous thread: > > http://www.mail-archive.com/peg@lists.csail.mit.edu/msg00379.html
Do you plan to use XML documents in memory or only for persistence? I think that keeping everything in XML would not be very efficient in my case. My goal is to provide a suitable replacement for building a custom AST in semantic actions (à la Yacc). I use hash tables that maps the start location of nonterminals to their end location. I have one such table for every "node" which are nonterminals marked with a "fat arrow" in the grammar (<= instead of <-). This is the only annotation allowed in my PEG grammar (the rest is identical to the ASCII grammar in Bryan Ford's paper). I may also add "skip" tables to speed up the traversal of "non-nodes". One interesting property that I maintain is that even if I drop entries in the table to save memory (or if I don't memoize at all), I can still traverse the tree with the same interface (triggering a reparse under the hood). > [1] https://github.com/ajaxorg/cloud9, http://cloud9ide.com _______________________________________________ PEG mailing list PEG@lists.csail.mit.edu https://lists.csail.mit.edu/mailman/listinfo/peg
