Danny's Weblog

2005 Nov 06 [ Sun ]

How to repeat a long filespec in a command line

I have been looking for how to do this for *years*. You often have to change the name of a file in another directory. That's easy in Windows, but in Unix you have to key in the whole blasted filespec again (and hope you don't make a typo): mv /home/d/dannyw/blosxom/data/images/phnompenhtrip/crossing1.jpg... aaagh, I can't go on.

In a Linux journal article on the *bash* shell, I found this:

www.linuxjournal.com [http://www.linuxjournal.com/article/7385]

echo {one,two,three}fish
 onefish twofish threefish
echo /etc/httpd/conf/httpd.conf{,.bak}
 /etc/httpd/conf/httpd.conf /etc/httpd/conf/httpd.conf.bak

Those are *wiggly* brackets, and they create *multiple* parameters to your command line, so you can do something like:

mv /etc/httpd/conf/httpd.conf{,.bak}
mv /etc/httpd/conf/httpd.conf /etc/httpd/conf/httpd.conf.bak

I tried it under tcsh, and it still worked.

Perfecto! If only I can remember it!

2004 Jul 30 [ Fri ]

Link to teach-yourself-sh guide on the web

I feel embarrassed to admit this but when I need to do a shell script I often realize that I have forgotten some essential trick. Because the shell tends to use very short and/or common keywords, it's very difficult to search for terms like "if" or "return" to check syntax/usage. (I get mixed up between sh, csh and bash, also perl of course; sometimes I try syntax which *appears* to work but is actually doing something quite different from what I intended...)

So the most efective thing is probably to check out general beginner-level guides now and then. Criminy, I've done most of this stuff n times before, honestly – but I learn something new every time.

steve-parker.org [http://steve-parker.org/sh/sh.shtml] Nicely-formatted, and with clear useful examples. I noticed bigwebmaster.com and librenix.com both linked to it.

2004 Apr 26 [ Mon ]

Tips on using the Unix command line -- syntax for "locate" utility

The "locate" utility uses a prebuilt database to rapidly find filenames anywhere on the system. Clearly it depends on a recent database. Also, the database frequently is generated without following symbolic links (which might well be circular, or otherwise result in gigantic numbers of files0; and the many security issues mean it is often disabled for non-root users.

Oh well. If you *can* use it, you may run into this syntax issue.

I'd often tried to use commands of the form

locate stuff.*

but they didn't seem to work.

Commands of this form work:

locate stuff

and return stuff.c, stuff/some.txt, /home/danny/stuff.001/some.txt etc, but don't give much control to reject the stuff you don't need.

Of course the * needs to be hidden from the shell:

locate "stuff.*"

but that does not work either, because when a wildcard is used the non-wildcard characters have to match *exactly* – so in this case it will not even find "stuff.c" in the current directory (because "locate" stores all its locations in absolute form, and does not compare relative to the current directory).

Instead what you have to do is:

locate "*stuff.*"

Tips on using the Unix command line -- Fixing a broken shell

Whenever I get an account on a new Unix system it usually turns out that the action of the backspace key is messed up. The reason for this is that Emacs has a default configuration which is both terrible and impossible to understand, and the original installer spends weeks finding one issue after another, whether inside Emacs or at the shell level; when he winds up with a configuration that works for him, he never wants to mess with it, even if it's screwed up for everyone else.

The result is that I have to dink around with various shell config changes till I get it right, and sometimes I wind up with a hosed shell. This is something that worked for me using bash, when I found a "magical" snippet of code in the default shell config. I tried editing .bashrc: I added "stty ek" to run after the funny code which seemed to cause the problem: it resets erase and kill key handling to default, and seemed to do what I want for vi, at least.

Tips on using the Unix command line -- Acting on a list of files

I wanted to listen to a bunch of sound files, all of which were quite short. Eventually I used "find" to put a list of NL-separated sound filenames in audio.dat); the trick for converting filenames into *arguments* is one that took some head-scratching. It's the "$()" below:

cat $(cat audio.dat) > /dev/audio

This is actually an overcomplex example. I only needed to use "find" because there were sound files in multiple subdirs, and I wanted to pick up the full pathnames. Likewise, if you don't know how the linux /dev/audio device works, just catting a binary sound file to it will look a little confusing.

Here's a simpler example, though less useful. Suppose you wanted to concatenate a bunch of files whose filenames were in a file called orange.txt. You could do:

cat $(cat orange.txt) > compendium.txt

This example, it occurs to me, is probably also a little confusing. The "cat" program is being used in two different ways: just to copy orange.txt to standard output; and then to truly concatenate the files defined in orange.txt. Oh well; I suppose most people who can see the usefulness of this tip at all already grok cat.

(It also occurs to me I could have done a find -exec, but that's always so clumsy; this seems better. Also, because it works with a pre-generated file listing, it avoids long pauses in the sound output if "find" takes a while to find the next file.)

When I checked, csh did not like the "$()" syntax, but it worked in sh and bash.

---

Debug: hittotal: 17 startban: 0 dancookie: endbandate: banned: 0 tempdate: tert: jse: jsno jsh: 17