Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site sjuvax.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!princeton!astrovax!sjuvax!armstron
From: armstron@sjuvax.UUCP (L. Armstrong)
Newsgroups: net.lang
Subject: Complex/Simple scanners - More hashing!
Message-ID: <2394@sjuvax.UUCP>
Date: Wed, 16-Oct-85 08:56:23 EDT
Article-I.D.: sjuvax.2394
Posted: Wed Oct 16 08:56:23 1985
Date-Received: Sat, 19-Oct-85 05:49:14 EDT
Distribution: net
Organization: St. Joseph's University, Phila. PA.
Lines: 31

*** YUM YUM - EAT EM' UP ***

I have a question that has bugging me for quite some time lately, and I was
wondering what some of the opinions of others were on the subject.

My question is this:
	Do you think it would be faster for a compiler to have a more
	complex scanner that could recognize key-words/reserved identifiers
	immediately (i.e., simply by keeping a record of each inputted 
	character) and not have to hash and search for those key-words
	each time one is encountered, or would it be better to keep the 
	scanner simple, recognizing each key-word as simply an identifier,
	and then search the symbol table to determined wether or not
	your identifier was reserved.

	It is not immediately obvious to me which one would be quicker,
	although I have a tendency to think it may be the former.  
	Lex allows for one to design a scanner of the first type 
	very simply.



						Thanks,
						-Len
-- 

NAME		Len Armstrong
UUCP		{astrovax | bpa | burdvax | allegra }!sjuvax!armstron
ORGANIZATIONS	RCA Advanced Technology Labs
		St. Joseph's University
PHONES		(WORK) (609) 866-6647