sed：在模式匹配后对N行执行替换

Question

I'm trying to use sed to perform a replace on only N lines after a specific pattern.

Looking at the sed man page it seems like it should be possible to do using addr1,+N syntax. But I can not get it to work.

 <data>
  <variable>DebuggerItem.Count</variable>
  <value type="int">2</value>
 </data>
 <data>
  <variable>Version</variable>
  <value type="int">1</value>
 </data>

The replace I am performing finds the numbers in this markup and increments them by 1. It is as follows.

sed -r 's/([ ]*<value type=)\"(int)\"(>)([0-9]*)(.*)/echo "\1\\"\2\\"\3$((\4+1))\5"/ge' filename

I now want to restrict that replacement to just the N lines after the one containing "DebuggerItem.Count". In my current instance N is 1, but I have another example where I would want to perform this for 5 lines after the matched one, so N would be 5.

It can probably be performed with a more complex replace, but ideally I would like to learn more about using addresses with sed.

Note: I had some trouble with the double quotes in the search replace as I needed to double escape them for the following echo. So the search replace is a little convoluted. My apologies.

Answer 1

I agree with @Birei; don't do this with sed . Nobody expects your XML tools to break if there's a newline added in a harmless place or nodes reordered, and that's exactly the sort of thing that will happen if you use line-based tools for the task. Things would break in horrible ways, and they will probably do it two months after you forgot what you built there and why.

There are plenty of tools that can parse and transform XML sanely. For example, with xmlstarlet your problem can be solved like this:

xmlstarlet ed -u '//data[variable="DebuggerItem.Count"]/value' -x '. + 1' filename.xml

Here //data[variable="DebuggerItem.Count"]/value is an XPath expression selecting the value subnode of a data node anywhere in the document whose variable subnode has a value of "DebuggerItem.Count" , and . + 1 . + 1 is another XPath expression meaning the value of the current node plus one.

See also here .

If you are hell-bent on using sed , and rest assured that it will come back to bite you, you can restrict the application of your s command to three lines after you found DebuggerItem.Count in the file like this:

#               new bit            from here it's the same as before.
#       vvvvvvvvvvvvvvvvvvvvvvvv  vvvvvvv
sed -r '/DebuggerItem\.Count/,+3 s/([ ]*<value type=)\"(int)\"(>)([0-9]*)(.*)/echo "\1\\"\2\\"\3$((\4+1))\5"/ge' foo.xml