Hacker News Reader

Comment by melling | Hacker News Reader

Guile? Great idea. Can someone fix Emacs so that Guile (Scheme) is the default language? Stallman was pretty shortsighted with elisp.

http://www.emacswiki.org/emacs/GuileEmacs

View on HN

I wouldn't blame Stallman for being shortsighted with elisp. He began work on GNU Emacs in 1984. Scheme wasn't standardized until 1990.

I agree that Scheme would be a better choice of an extension language, and honestly I think Stallman would probably agree. It's more that changing the extension language would cause a lot of compatibility problems. I hope the Guile Emacs guys can pull it off.

Yes, it was standardized 24 years ago. I think Stallman's argument was that too much code was already written in elisp. For example, calc is a huge elisp app. He didn't want to lose apps like this:

http://www.gnu.org/software/emacs/manual/html_mono/calc.html

I don't know much about Guile, but can't it compile elisp too?

But you would need dynamic scope in Scheme to mix elisp/Guile code in Emacs?

Edit: Ah, this was discussed in your link at the top of this comment thread.

Not sure if calc already existed in 1984.

But Stallman argued (in one of the early emacs papers) that dynamic scope is much more convenient for a editor than lexical scoping (as implemented in scheme or Common Lisp)

Dynamic scope makes it easier to change some variable per buffer. With lexical scoping you would need to simulate dynamic scope somehow to be able to change state per buffer.

Common Lisp supports both dynamic and lexical scope.

Emacs-like editors have been implemented in it, and there's no fundamental reason that a true Emacs couldn't be.

Scheme is more of a toy; I'd not prefer it to Common Lisp for a production system, whether that system is an OS or an editing environment.

Scheme was plenty standardized by the mid-80s when RMS forked (James) Gosling Emacs and replaced its Mocklisp with a real LISP. Very few if any pay attention to the 1990 IEEE standard; for one thing, you've got to pay money to get it (currently $138 for a PDF, $110 if you're a member), vs. the free RnRS series.

RMS could in theory have adopted a Scheme, or even a lexicaly scoped version of mainline LISP, but by then he was so experienced with normal, pre-Common Lisp mainline LISP (Maclisp and Lisp Machine Lisp) I'd wonder if it even occurred to him.

There have been efforts in that direction for a long time: http://www.emacswiki.org/emacs-en/GuileEmacsHistory . The current push seems to have reached the "80% done, 80% to go" stage.

Not really. Look at the "Current State" section here:

http://www.emacswiki.org/emacs-en/GuileEmacs

> As of the end of GoogleSummerOfCode 2014, the Elisp engine of Emacs is fully replaced with that of libguile, and most things Just Work™. Some performance regression of dynamic scoping (lexically scoped code should be fine), and broken corner cases (like crazy Elisp macros that directly output legacy Elisp bytecode) remain.

From its earliest days, one of the selling points of Guile was that eventually it should be possible to use it to run Emacs Lisp. This is one of the reasons RMS chose it as the GNU's official extension language. For some of the other reasons, google "Tcl war".

Reply on news.ycombinator.com