The one thing that drove home for me the efficacy of the dyad ;:, is that an implementation <;.1 y as a sequential machine was within a factor of 2 (something like that) of the primitive <;.1 y. I hand-coded the latter in C and devoted much effort to it.
On Mon, Feb 27, 2012 at 9:49 AM, Raul Miller <rauldmil...@gmail.com> wrote: > I went through the wiki, with a text search on sequential machine, > looking for pages that might be of use here. > > The first page I examined (working up from the bottom on my list of > plausible candidates) was > > > http://www.jsoftware.com/jwiki/Guides/Language%20FAQ/J%20BNF?highlight=%28sequential%29%7C%28machine%29 > > This led me to it's "See Also" link: > http://www.jsoftware.com/pipermail/chat/2007-November/000678.html > > And, I noticed there a statement by Oleg Kobchenko: > > J uses ... interpreter, which is a context-sensitive > bottom-up table-driven parser/executor: > http://www.jsoftware.com/help/dictionary/dicte.htm > > (which is obviosly not recursive-descent). > > And, from my point of view, this parser is obviously recursive descent. > > That said, the typical recursive descent parser is a left to right > parser. This is what the "LR" in various parser acronyms stands for. > J's parser is a right to left parser, so the parallel acronym would be > "RL". [Technically, "LR" means "Left to right, producing a Rightmost > derivation".] > > Secondly, most recursive descent parsers have "one token lookahead". > I have thought in the past that J uses "four token lookahead", but > this is not the case. J uses one token lookahead (only for > assignment). Instead, J uses four token "lookbehind". > > Conceptually speaking, the J parser pops four tokens off the stack. > But this is just a description of the internal state of the parser, > and does not impact its status as a recursive descent parser. > > A person might argue that "RL" and "LR" are not logically equivalent, > but I do not think that's relevant here. Ultimately, the parser sees > a sequence of tokens and from an algorithmic point of view the > meanings of "left" and "right" are implemented in another module (the > one that breaks the input stream into words). But the lexer -- the > module that breaks the input stream into words -- is not considered > when classifying the parser. > > Anyways, I just had to get that off of my chest. > > Some pages which look maybe interesting for learning about sequential > machines include (possibly in partially reversed order of relevance): > > http://www.jsoftware.com/jwiki/Essays/Huffman%20Coding > > http://www.jsoftware.com/jwiki/Essays/Regex%20Lexer > > http://www.jsoftware.com/jwiki/Guides/Parsing > > http://www.jsoftware.com/jwiki/SequentialMachine > > http://www.jsoftware.com/jwiki/JWebServer/HttpParser > > And, of course, the Sequential Machine lab is probably a very good > resource to start with. > > -- > Raul > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
