Path: utzoo!attcan!uunet!lll-winken!lll-tis!ames!mailrus!tut.cis.ohio-state.edu!husc6!bbn!gatech!mcnc!rti!jb
From: jb@rti.UUCP (Jeff Bartlett)
Newsgroups: comp.lang.forth
Subject: Re: Glossaries
Summary: searching order
Keywords: Absolutely correct
Message-ID: <2368@rti.UUCP>
Date: 13 Jul 88 21:39:07 GMT
References: <8807121636.AA12801@jade.berkeley.edu> <3044@juniper.uucp>
Organization: Research Triangle Institute, RTP, NC
Lines: 28

In article olorin@juniper.uucp (David Weinstein) writes:
> In article A-PIRARD@BLIULG11.BITNET (Andre PIRARD) writes:
> >I have always been amazed by glossaries being sorted alphabetically.
> >This makes them a real pain to read. Isn't the place of LOOP just after DO,
> >or even better, both described simultaneously with multiple headers.
> >That way, a glossary is much more educational with everything in context.
> >And alphabet is the realm of an index, isn't it?
> 
> Agreed. And my favorite, worn, and much battered copy refernce ("Starting
> Forth" by Leo Brodie [Forth 79 edition]) does just that in Appendix 4.
> 
> Pity it isn't always done that way.
> 
An implementation of '79 that I once ported had the words in 'search order'.
Since the dictionary was a linked list, the base vocabulary was ordered by
frequency so that the more frequent words would be found earlier.   This
cut down on the average number of links to traverse during a LOAD.

A version of forth for VM/CMS at NCSU had sixteen lists through the dictionary
and the words were hashed to find which list to traverse.  This was a port
that started from an Rockwell AIM 65 evaluation kit and changed to 32 bits.
It also had full access to the system.  Later applications included a full
screen editor and LISP.   (Oooo, a prefix language implemented in a postfix
language, brain aerobics :-)

Jeff Bartlett
Center for Digital System Research
Research Triangle Institute
jb@rti.rti.com	...mcnc!rti!jb