Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!usc!wuarchive!uunet!bu.edu!nntp-read!jbw
From: jbw@bucsf.bu.edu (Joe Wells)
Newsgroups: comp.emacs
Subject: Re: require & provide
Message-ID: <JBW.90Sep30001726@bucsf.bu.edu>
Date: 30 Sep 90 04:17:26 GMT
References: <9009261610.AA01539@scenic.scenic.wa.com>
Sender: news@bu.edu.bu.edu
Organization: Boston University Computer Science Department
Lines: 26
In-reply-to: pauld@scenic.wa.com's message of 26 Sep 90 16:10:52 GMT

In article <9009261610.AA01539@scenic.scenic.wa.com> pauld@scenic.wa.com (Paul Barton-Davis) writes:

   ... Typical syntax is:

   (require 'foo)
   (provide 'bar)
   .... (elisp code for new mode etc.)

   Although this is almost always OK (most distributed code tends to work
   just fine like this), it seems conceptually wrong - 'bar has not yet
   been provided at the point claimed. It seems more appropriate to use

   (require 'foo)
   ... (elisp code)
   (provide 'bar)

   This way, any other packages that (require 'bar) are not corrupted by,
   for instance, terminating a load of bar.el, or the occasional error
   whilst loading.

Strangely enough, "provide" and "fset" (ie. "defun") operations are undone
if a failure occurs while loading the definition of an autoloaded
function, but the same is not done for "require".  Hmm.

-- 
Joe Wells <jbw@bu.edu>