[Haskell-cafe] IDE?

[Claus Reinke]

> However, one of the reasons that Emacs is so great is the absolute wealth of
> libraries available for it. It's been aroud a long time and people like it a lot
> so there's pretty much an Emacs Lisp library to integrate _any_ tool, to help
> editing _any_ kind of source/configuration file etc. It would take many
> lifetimes to replicate this in a different language.

see also the history section in Luke Gorrie's Distel paper:

    http://fresh.homeunix.net/~luke/distel/distel-euc.pdf

    This was followed by Ermacs, a concurrent
    Emacs clone written completely in Erlang. Ermacs
    is fairly complete – it has major modes for
    Erlang and Scheme programming, a built-in Erlang
    shell, and support for efficiently editing large
    files. However, once the core editor was complete,
    it was obvious that GNU Emacs has an incredibly
    large set of wonderful features, and that extending
    Ermacs to include “enough” of them was
    completely out of the question.
    The lessons learned from Ermacs lead to Distel,..

how is Yi going to avoid that trap?

> I've thought a little about writing a Haskell -> Elisp compiler, so that people
> could contribute to Emacs by writing Haskell, but I've got doubts about its
> feasibility.

Luke's Distel seems to have been very successful for
erlang <-> emacs interaction, enabling erlang modes for emacs
to be written in erlang, with tools that actually understand
erlang. this seems to suggest that easy communication with
some good-haskell-system-api-based haskell code might
be a promising equivalent?

    http://fresh.homeunix.net/~luke/distel/

both emacs and vim can pass buffer segments and editing
session information to external (haskell) code, working as
text or file transformers, and both emacs and vim can be
controlled by such external code. so haskellers can already
use haskell to extend emacs/vim functionality, it just isn't
very convenient when it goes beyond transforming text.

i would much prefer if vim exposed a standard ffi, instead
of all those specific bindings to popular scripting languages
and tools, but even without a proper ffi, haskell code
cooperating with vim script can already be written. and
similarly for emacs, i suspect.

so it may just be a matter of doing less of the ide work in
elisp/vimscript, and of introducing an explicit interface
for inter-process communication between the editor frontend
and some haskell-system-api-based backend.

claus