Kimler Adventure Pages (http://randsco.com)

After we moved Randsco to a dedicated Linux server, I've been able to (again) use UNIX line commands via shell access. Sweet. Here's a growing reference of Linux ditties (currently just a one-line recursive, global search & replace)

LINUX: Ruling more each day!

It's been ages (1999) since I've worked on a Unix operating system (Sun Microsystem - Solaris). Since our recent move to a Linux web-server, I've had fun dusting off my rusty memory of line commands.

Gosh, 'Nix is such a more robust operating system than DOS or XP!

In any case, I had the need to do some fancier-than-normal stuff and thought it would be a good idea to jot the commands down, in case I needed them again. So, here's my (slowly growing) list of tested, linux-based one-offs.

• One-line Global Search & Replace (sed, xargs, grep)

- -

(1) Global Search and Replace

After moving web-servers, I needed to replace a string -  http://randsco.com/directory/file.html  - with something else. It only took a few manual edits, before I realized how many files I needed to edit. What I needed was a way to perform a global search & replace, but do it across every file in the directory tree.

After hitting Google, I realized that there were a bunch of script out there, that did the trick, but this seemed to be more work than necessary? Finally, I found what I was looking for: a one-line command that performed a global (and recursive) string search and replace.

This command was short, sweet and seemed to do what I wanted. It looks through every file under the directory in which it is run, searching for "oldtext" and replacing it with "newtext" and creates a backup "file.bak" for every file it changes.

I made a few test files and directories and it seemed to work. Once I did get it to work, I bent it a tad, to turn off the backup option (the cowboy that I'm learning to be).

It worked as advertised, but after running it, I realized a couple of things. (a) The command touches every file, regardless of whether it contains "oldtext" and (b) using a forward slash delimiter for SED made searching for forward slashes onerous.

Since I'd just transferred the whole of my old website to a new host, preserving the date-stamp wasn't an issue, but it might be, later on. So I searched for a solution and found this - which I have yet to test:

I'm not sure what the single quotes and e flag do for the sed argument (necessary?), but it's reported to work as advertised, where I found it.

Another afterthought: Though the first above commands worked fine for me, because the search string included forward slashes, the final construct looked a tad odd with all the backslash escapes (i.e., 's/http:\/\/randsco.com\/ ... /' ). I remembered that sed provides for any delimiter (you don't have to stick with forward slashes). For example, I think I could use pound signs as delimiters and the sed line would look like: 's#http://randsco.com/ ... #'  (a tad cleaner).

Source