Extract a substring from a line in a faill

PMichaud · 12-12-2009, 07:40 AM

I'm new to linux and bash, and all the awesome utilities linux has, so I'm struggling with the syntax here.

Basically, I have a configuration file that's roughly:

<?php

$var = 'dsfsdf'; //sdfsdf
define('KEY', 'ftyurghj');

//etc.

?>

My goal is to extract the values of $var and 'KEY'

The issues I'm having are A) I'm having a heck of a time escaping some characters like "="

I've done something like this:

sed -ne '/^\$var/p' file.php

And gotten the right line back, but when I go further and try

sed -ne '/^\$var \= /p' file.php

It starts complaining. I would ideally want it to match any whitespace so it could read:

$var ________ = __________ 'blah';

without choking.

and B) I don't know how to capture matches using grep or sed or whatever I need to use

sed -ne '/^\$var \= \'([.*])\'\;/p' file.php

So let's pretend that regex worked (even though it actually just chokes). I want the contents of the parens.

How should I approach this?

ghostdog74 · 12-12-2009, 07:57 AM

just the shell

Code:

while read -r line
do
    case "$line" in
        *\$var*)
            line=${line%%;*}
            val=${line##*=}
            ;;
        *define*)
            line=${line#*\'}
            key=${line%%\'*}
            ;;
    esac
    [ ! -z "$val" -a ! -z "$key" ] && echo "$val: $key" && exit
done <"file"

jschiwal · 12-12-2009, 07:58 AM

Code:

sed -n "/<\?php/,/^?>/{ /$var *= */s/\$var *= *'\([^']*\)'.*/\1/p 
                                                                  }" testfile

Escaping the single quote is potentially the trickiest part. I did it here by putting the sed line inside double quotes. The dollar sign also needs to be escaped in the LHS so it is taken literally. You can insert a second line inside the brackets to match define line.

PMichaud · 12-12-2009, 09:05 AM

Thanks for the help!

I'm trying to figure out what's going on using my crude knowledge of sed.

The bracket is on a new line, is that necessary?

When I try to use what you wrote, with new line, something strange happens. I paste it in, and without pressing enter, it seems to execute, and it lists all the files in the current directory in 2 columns with a ">" at the end. It repeats that list maybe 2 dozen times (all in less than a second). Don't know what that's about

When I remove the tabs and new line, I just don't get a match, so I execute, get nothing back, and am back at the prompt.

From what I can gather though, you're saying:

use sed, without printing the normal output. We're looking between <?php and ?> for a string matching "/$var *= */s/\$var *= *'$[^']*$'.*/", and we'll grab the first reference we get, and print it.

The pattern is confusing though.

/$var *= */s/\$var *= *'$[^']*$'.*/

Doesn't the $ need to be escaped? What about the =? I get the zero ro more spaces, and I get the reference matching zero or more characters that aren't ', but why does it try to match ".*" after that?

jschiwal · 12-16-2009, 01:29 PM

The "/<\?php/,/^?>/" at the beginning defines a range of lines. The brackets "{}" then enclose the sed commands to run when inside that range.

Yes, I should have escaped the "$" but it worked when I tested it and was to lazy to go back and escape it.

I put the closing bracket on another line to emphasize that you could add another line for the line with "define".