From nellardo@xxxxxxxxxx.xxxx Sat Sep 18 16:28:36 1999 X-Digest-Num: 237 Message-ID: <44114.237.1295.959273825@eGroups.com> Date: Sat, 18 Sep 1999 19:28:36 -0400 (EDT) From: David Brookshire Conner From: "Bob LeChevalier (lojbab)" > > At 11:02 AM 9/18/99 +0100, Piermaria Maraziti wrote: > >From: Piermaria Maraziti > > > >At 11.36 18/09/1999 +0200, you wrote: > > >I don't see the point in having an extra word processor just for one > > >language. Agreed - that's why I suggested an extension to an existing editor. Since I know Framemaker and Emacs best (and neither are provided by the Evil Empire :-), I suggested extensions to those. > > > Also, I think there are far more important items on the > > >programming agenda: a good, portable glosser / automatic translator > > >of syntax-parsed lojban text. > > > >The idea is to do a WP with an integrated "good, portable glosser / > >automatic translator > >of syntax-parsed lojban text" - perhaps at least as for vwrsion "1.0", > >without the "automatic translator" part, excluding the interlinearizing > >capabilities that will help so much in translating (and learning)! Erm, I'd be careful about what "the" idea is. See below. > Well, even that is probably version 2.0. Robin Turner, who originally > mentioned it on conlang, suggested that being able to click on a word and > call up its place structure (which implies calling up its tanru breakdown > if it is not in the dictionary) is the thing most needed for Lojban > writing. Yeah, this would seem to be the case - place structure is probably the most opaque part of the language. Sure, natural languages have similar problems (transitive vs intransitive verbs come to mind), but I'll bet people learning English (or anyone trying to write in a non-native tongue) would appreciate usage hints comparable to lojban place structure. > Having a dictionary lookup for English words would seem an > obvious thing to include at the same level. Yes, especially with the part of speech identified and allowable adjacent parts of speech and their meaning. > Such lookups are presumably > trivial utility routines by now, so if you started from that open software > base that several have mentioned, we have something that is small enough > for someone to tackle in their spare time in a few days. Yep - the "hard" part is having the database of words to look up. But le gi'uste looks regular enough in format that it could probably be adapted pretty much as is. So, presuming Emacs as a base, here's the stepping stones: 1. an interactive function that looks up the word under point (the cursor) in various word databases - gismu, cmavo, etc. 2. An extension that provides tab-completion of partial words 3. A major mode that provides some basic functionality: regexps to use with outline-minor-mode (ni'oni'o and the like) notations for use with font-lock (what's a quoted piece of text (string to font-lock), "paren" matching, etc) bindings to 1. and 2. binding to pipe buffer or region to parser 4. A major mode with more functionality interactive syntax checking auto-suggestion and elision of cmavo spell-checking dictation (!) lujvo tools > The fancy version written from scratch that uses Java, and has the > functionality of the parser/glosser and the lujvo maker and typesetting > including Tolkienian stuff is a much bigger project - probably > worth doing, > but not likely to have someone writing it very soon. I just want to be clear on a few things: 1. Writing a text editor from scratch is an interesting student exercise, but I don't see that you'd gain much for lojban functionality over existing open source (or closed but extensible) text editors. 2. The typography Mark and I have been discussing is (to me, at least, and I imagine to Mark as well) a wholly separate issue from the text editor. The issue got raised in the context of the text editor, but type-setting and text editing are not the same thing (I write in Emacs, but use TeX to type-set - and I'm damn glad I don't typeset in Emacs). 3. Very specifically, the use of Tengwar for lojban is strictly an esthetic exercise for me. I make no claims that it has any real practical application beyond beauty (for a very conlang-ish kind of beauty at that). > You either have to > write it from scratch or translate the existing code. Just making existing > Turbo Pascal code portable has never been accomplished for any of our core > software (at least a dozen people started to translate LogFlash into C, > including using an auto-translator to help, and no one in 10 years ever got > a working program, though Eric Raymond got close. I suspect that the > glosser is bigger than LogFlash, though I haven't looked. Both are very > time consuming to debug and test). This doesn't surprise me - especially rewriting Pascal in C (something that strikes me as a largely misguided effort at best, and positively sadistic (or masochistic, depending on whether the task was assigned or voluntarily undertaken)). For various kinds of parsing and transformational applications, I'd suggest the programming language Haskell. It isn't wide-spread, but it runs everywhere, and is quite well-suited to these kinds of things. Brook --------- All wiyht. Rho sritched mg kegtops awound? --------- Fancy. Myth. Magic. http://www.concentric.net/~nellardo/