Last month we finally wrapped up the long journey toward creating a useful shell script with the hi-low game. Alright, "useful" might be a bit of a stretch, but if you've read through all the columns leading up to this point, you should have a good understanding of the basics of creating and debugging a shell script, a skill that will prove invaluable as you travel further down the Linux and Unix path.

In this column I'm going to present a short shell script that does something darn useful for those of you who secretly are also running Mac OS X, but even if you're not, it's going to be an interesting script to learn.

Parsing XML "plist" Files
A common task you'll want to accomplish in shell scripts is parsing and extracting specific data from an information file. Sometimes they're well formed, like the /etc/passwd file, but other times their format is more complex or obscure. Mac OS X (which is built upon a FreeBSD core) has lots of these obfuscated data files created in XML format.

A typical file, the bookmarks file for Apple's Safari browser, stores an individual bookmark this way:

<dict>
<key>URIDictionary</key>
<dict>
<key></key>
<string>http://www1.dailycamera.com/ </string>
<key>lastVisitedDate</key>
<string>134970860.4</string>
<key>title</key>
<string>Camera</string> </dict>
<key>URLString</key>
<string>http://www1.dailycamera.com/ </string>
<key>WebBookmarkType</key>
<string>WebBookmarkTypeLeaf</string>
<key>WebBookmarkUUID</key>
<string>CB24B-861-1D8-AE1-000A97EC4</string>
</dict>

Don't panic. The only thing you need to notice here is that the URL appears immediately after the line URLString, and that the name of the bookmark entry appears immediately after the title.

Extracting Lines from a File with Grep
You probably already know that you can use the grep command to extract lines that match a specific pattern. However, if you're running Linux, you have a more powerful version of grep - GNU grep - which lets you extract a specified number of lines before or after each matching line too. Perfect!

The first step in writing this script is to use grep to extract the lines that match the two fieldnames specified and the two lines immediately following each match. This is done with the -A1 flag:

bm="$HOME/Library/Safari/Bookmarks.plist"

grep -A1 -E '(>URLString<|>title<)' $bm

grep -A1 -E '(>URLString<|>title<)' $bm

(I've assigned the variable "bm" to the full pathname for convenience.) Notice that I'm also using a simple regular expression to match lines that have the pattern >URLString< or >title<. Use the -E flag to convince grep that you really want to use a regular expression.

We're getting there. The problem now is that we have both the lines that contain the information we want and the lines that match the fieldnames. Another job for grep, this time inverting the test to show only the lines that don't match the specified pattern:

grep -A1 -E '(>URLString<|>title<)' $bm |
grep -v -E '(>URLString<|>title<)'

Almost done, actually. Here's an example of how the output looks now:

<string>Camera</string>
<string>http://www1.dailycamera.com/</string>
<string>CD</string>
<string>http://www.coloradodaily.com/</string>
<string>Gnews</string>
<string>http://news.google.com/</string>
<string>NYT</string>
<string>http://www.nytimes.com/</string>
<string>WSJ</string>
<string>http://online.wsj.com/home/us</string>

All that's left is to clean up the format a bit.

Chopping Lines with the Cut Command
One command that I use quite frequently in shell script programming, though many Linux folk have never heard of it, is cut. Specify a delimiter and what field or fields you'd like, and it'll extract just those fields from the input stream. For example, the /etc/passwd file has a number of different data fields separated by colons. To extract just the third field is a very simple command: cut -d: -f3

Got it? Now, let's look at how we can use cut to strip off anything prior to the first ">" and subsequent to the second "<" in each matching line.

cut -d\> -f2 | cut -d\< -f1

Not the most elegant or graceful solution, but definitely quick and dirty, with an emphasis on quick. The first command tosses out anything prior to the first ">" symbol, then the second shows only what's on the line prior to the first occurrence of "<".

For the first line above, <string> Camera</string>, the first cut would produce Camera</string> and the second would produce Camera, exactly as we hoped.

Are we done? Not quite, because while it's useful to be able to produce an output of bookmark name, URL, bookmark name, URL, it'd be much nicer to produce an HTML format output that can then be viewed in any Web browser. To do this, however, is a bit more tricky and involves learning how you can hook a structured block of scripting code into a pipeline.

And that, I'm afraid, will have to wait until next month. See you then!