The reason is simple: you should use some substitution, which "substitutes" the tag and the surrounding brackets to nothing, leaving only the text between the tags - "2".

What you have written is a so-called "range" command: you told sed to do something - the command "p" in your case - to a range of lines which are starting with "<min-pool-size>" and ending with "</min-pool-size>". So everything in the one line constituting the range is being printed.

Regarding youe problem: lets reformulate your problem in a more general way: You have text in the form

Code:

<tag>some text</tag>

where "tag" is some supplied text from outside. you want to filter out everything save for "some text". First, ask yourself if the text you want to preserve could span more than one line, because there will be extra effort necessary if this could be the case. Is the following possible?

Code:

<tag>some
more text</tag>

And if it is do you want to preserve line breaks as in the original or do you want some one-line stream to be the result?

Let us start with the simplest: only one-liners. The problem here, like in all the other cases, is to put the inherited variable content into the sed-regexp. Save the following to a file called "test1.sh", give it execute-rights and call it like "test1.sh /path/to/inputfile":

Code:

typeset fIn="$1"
typeset tag="min-pool-size"
sed 's/<'"${tag}"'>\(.*\)<\/'"$tag"'>/\1/' $fIn

This will work in your example, but it will fail in my second variant. Notice that i quoted cautiously all the variables to be sure to end up with a continuous string as regexp after the shell is done expanding the variables.

In the next version we will take care of multiline contents but will let them remain multiline. We have three sorts of lines to deal with:

Code:

1) "some content<tag>content-to-preserve"
2) "lines between start- and end-tag"
3) "content-to-preserve</tag>some more content"

We will use the range facility, like you did, but in a somewhat more complicated fashion:

Code:

typeset fIn="$1"
typeset tag="min-pool-size"
sed -n '/<'"${tag}"'>/,/<\/'"$tag"'>/ {
               s/^.*<'"$tag"'>//
               s/<\/'"$tag"'>.*$//
               p
          }' $fIn

With the "-n" all output of sed is suppressed. This way we filter out all irrelevant text before and after our tags. In the first line of the regexp we declare the range - the first line where the start-tag appears, the last line where the end-tag appears and all lines in between. To these lines we apply all the commands in the curly braces one by one. Then we cut out everything before the start-tag including it itself in the first line (taking care of the type-1 lines above), everything after the end-tag (taking care of the type-3-lines above) and the type-2-lines are untouched. The last command, "p", prints everything inside the range left over from the cuttings - voila!

How to transform this output to a one-line stream is left as an exercize to the interested reader who by now should be eager to try his newly found insight in the workings of sed on a problem of his own. ;-))

I hope this helps.

bakunin