sed first n occurrences result set between 2 tags / patterns

Question

I have a big XML file and I get all occurrences between 2 tags:

Here is what I do:

sed -n '/<tag>/,/<\/tag>/p' file.xml

And I need to filter to get only first N occurrences. I've tried with l param but it was not enough :(

So any one knows how to get N matched occurrences from all resultset?

For Example. Here the xml file content :

<?xml version="1.0" encoding="UTF-8"?>
<root>
<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>
<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>
<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>
<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>
</root>

 sed -n '/<tag>/,/<\/tag>/p' file.xml 

returns all elements.

So the goal is to filter to get the First n matched patterns ( Elements are multi-line) If n = 2 then result =:

<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>
<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>

Answer 1

Try:

xmllint --xpath '//tag[position()<=2]' file.xml

Or:

xmlstarlet sel -t -c '//tag[position()<=2]' file.xml

Or:

xmlstarlet sel -t -m '//tag[position()<=2]' -c . -n file.xml

If you wanted to do it with sed only you could do something like:

sed -n '
  1{x;s/^/../;x;}; # initialise counter with two tokens
  /<tag>/,/<\/tag>/ {
    p; /<\/tag>/{
      x;s/.//;/./!q;x; # remove a token and quit if hold space empty
    }
  }' file.xml

That is, use the hold space as a counter of remaining sections to display (using dot characters).

Answer 2

You really should be using a parser for this, but, just so you know, sed -n '/<tag>/,/<\/tag>/p' file.xml gets you all elements because you print them all. That command works by addressing all lines between a line containing <tag> and the next line in input that contains </tag>. Since that makes pretty much all of your lines, just printing them doesn't show up much of a difference. Something like the following might be a little nearer to the mark:

sed -n '\|<tag>|{:n
    \|</tag>|!{N;bn}
    y|\n| |;p
}'

It addresses <tag> lines and checks them for </tag>. If they don't contain the closing string it pulls in another line - and it does so repeatedly until the pattern space contains <tag>.*</tag>[^\n]*$.

Then I just translate all \newline characters in pattern space into spaces.

Here it is again:

sed -n '\|<tag>|{:n;\|</tag>|!{N;bn};y|\n| |;p}' <<\DATA
<?xml version="1.0" encoding="UTF-8"?>
<root>
<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>
<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>
<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>
<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>
</root>
DATA

OUTPUT:

<tag>  <t1>john</t1>  <t2>john</t2>  <t3>john</t3> </tag>
<tag>  <t1>john</t1>  <t2>john</t2>  <t3>john</t3> </tag>
<tag>  <t1>john</t1>  <t2>john</t2>  <t3>john</t3> </tag>
<tag>  <t1>john</t1>  <t2>john</t2>  <t3>john</t3> </tag>

Now you might do:

sed -n '\|<tag>|{:n
    \|</tag>|!{N;bn}
    y|\n| |;p
}' ./file | 
sed 's|> |>\n|g;2q'

...which gets me:

<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>
<tag>
 <t1>john</t1>
 <t2>john</t2>
 <t3>john</t3>
</tag>

Answer 3

I think this is what you want,

sed -n '/<tag>/,/<\/tag>/p' file.xml | head -10

Try the below command to get the first two lines which starts with <tag>,

$ sed -n '/^<tag>/p' file.xml | head -2
<tag><t1>john</t1></tag>
<tag><t1>john</t1></tag>

Answer 4

Pipe through head -N. After the first N occurrences, head exits and thus sed will stop.

Answer 5

As far as I know, sed matches are always greedy i.e. /<tag>/,/<\/tag>/ will match from the first instance of <tag> to the last instance of <\tag> - including any other XML objects between.

If your version of awk supports multi-character record separators, you might be able to do something like

awk -v n=2 'BEGIN{RS="</tag>\n";ORS=RS} NR<=n'

but really a more robust solution would be to use a dedicated XML parser - for example a very minimal implementation using python's minidom

#!/usr/bin/python

from xml.dom import minidom

xmldoc = minidom.parse('file.xml')
taglist = xmldoc.getElementsByTagName('tag')
for i in range(2) :
        print taglist[i].toxml()

Answer 6

Well, eventually I'll answer myself to my own question.

The solution I've found works in 2 ( maybe 3 ) Steps :

1 - Getting all Required Element by :

sed -n '/<tag>/,/<\/tag>/p' file.xml > selectedItems.xml

2 - Getting the N-th position of the last item by

POS = grep -n '</tag>' ./selectedItems.xml | head -n [POS] | tail -n 1

3 - Getting First N Required Items :

sed -n 1,[POS]p selectedItems.xml > selectedItems.xml

Of course it is possible to do all steps without splitting, but It wont be such clear.

P.S. To be sure that the position corresponds to the real N-th position in the tree ( when xml file is formed all in-line ) I used to use:

xmllint --format ./myxmlfile.xml