Re: Purpose of jde-autoloads
Przemysław Wojnowski <[email protected]> Fri, 14 Aug 2015 17:22:07 +0200
| Newsgroups | gmane.emacs.jdee.devel |
|---|---|
| Message-ID | <[email protected]> |
W dniu 14.08.2015 o 00:45, Stephen Leake pisze: > To make that single require work this way means that jde.el needs a > require for every other jde file; that's circular. Why circural? IMHO it is a directed graph with jde.el at the top and other "modules" below - nothing should require jde, except user. > I just spent some effort reducing exactly that problem, in order to fix > existing byte-compile errors. The code is still more cross-connected > than it should be; hence all the FIXME: refactor comments. Yes. Clearly concerns should be separated. > There is a way to get what you are asking for here; there could be a > jde-load-all.el that is just a list of requires. This is what I was thinking about, but as jde.el. :-) > In general, the Emacs style is to load things lazily, but in fairly > large chunks. That works better for people on older/slower machines. Such design assumptions should be revisited once in a while, because conditions change. How was "older/slow machine" defined? Do they still exist? > In particular, anything that starts or uses an external process on > load should be avoided and/or autoloaded. > On the other hand, it can be difficult to tell when something needs an > autoload, so the tendency is to add an autoload "just in case". So some > sort of policy on when they should be added, and tests/checks for this, > would be good. Good point with external processes. Now, IIUC Emacs autoload is like JVM class loading only on the first access to it. The difference is that in JVM it is by default for all classes and done automatically, but in elisp a programmer have to manually specify such code. > Is there something clearly broken with the current autoloads? Nothing I know of. :-) ------------------------------------------------------------------------------