Jules Bean wrote:

On 29 Sep 2004, at 14:07, Conor T McBride wrote:


However, the sooner we can escape from xemacs, the better.


What's your plan on that front?

I think perhaps I over-pluralised the pronoun. The sooner I can escape hacking in xemacs the better. I would like to see a gui, but I can see the argument for a number of interfaces, including text-only interactive editing, and also batch mode, or command-driven.

However, as you have things structured at the moment, you appear to seize total control of the keyboard...

That's because, if I'm going to write a weird editor for box-shaped things, I'd rather write it in Haskell than emacs lisp. However, that's no justification for the current lack of a proper abstraction barrier between the elaborator and the front end. `Proof-states-as-judgments, tactics-as-admissible-rules' is a reasonable idea; `Buffers-as-judgments, keystrokes-as-admissible-rules' is going too far...

so actually you don't really tap into the editor's power very conveniently (menus still work, but true emacs users don't use menus :-P).

If someone (possibly with some help) knew a bit more about emacs regions, and controlling which bits of the buffer are read-only, it would certainly be possible to do much more of the editing on the emacs side, and to restore proper emacs keystrokes. That might be no bad thing.

So do I gather that you're planning to write a gui front end of your own? In wxHaskell perhaps?

As Paul said, that's certainly something we're pursuing. The key thing is to establish the language which is used to communicate between the components. My plan is that we use syntax trees (monochrome and unlabelled in one direction, coloured and labelled in the other) as the interchange language: the elaborator should be ignorant of parsing, prettyprinting, the contents of sheds and even the mapping between user identifiers and internal names.

But the main obstacle is time.

Cheers

Conor
--
http://www.cs.rhul.ac.uk/~conor

Reply via email to