How to delete a line containing pattern which is first found after a specific pattern

Question

I am writting a shell script to delete an entry from a specific group. eg: file name is dest.xml

<domain id="1" group_name="group1">
    <node id="ABC">
    <node id="PQR">
    <node id="XYZ">
</domain>
<domain id="2" group_name="group2">
    <node id="PQR">
    <node id="XYZ">
</domain>
<domain id="3" group_name="group3">
    <node id="ABC">
    <node id="PQR">
 </domain>

I want to delete an entry from above file (filename is dest.xml) with node id="PQR" from group_name="group1" (it should not get deleted from group2 and group3). I can do it by reading a file sequentially n then delete it from a specific group. But if the file is to large (>10k lines) then it takes time.

Is there any simple way to this this?

Answer 1

I wrote a quick python script, do not know if it's simple enough.

Place this script in the same folder with your dest.xml.

#!/usr/bin/python
import re
FILENAME = 'dest.xml'
GROUPNAME = 'group1'
NODEID = 'PQR'

with open(FILENAME) as f:
    in_group = False
    for line in f:
        line = line.strip()
        group_pattern = 'group_name="{0}">'.format(GROUPNAME)
        end_group_pattern = '</domain>'
        node_pattern = '<node id="{0}">'.format(NODEID)
        if re.search(group_pattern, line):
            in_group = True
        if re.search(end_group_pattern, line):
            in_group = False
        if re.search(node_pattern, line) and in_group:
            pass
        else:
            print line

And now is the awk version.

#!/usr/bin/awk -f
BEGIN {
    GROUPNAME = "group1"
    NODEID = "PQR"
    in_group = 0
    group_pattern =  ".*group_name=\"" GROUPNAME "\""
    end_group_pattern = "</domain>"
    node_pattern = "<node id=\"" NODEID "\">"
}
$0 ~ group_pattern {
   in_group = 1
}
$0 !~ node_pattern || in_group == 0 {
    print $0
}
$0 ~ end_group_pattern {
    in_group = 0
}

Run this awk script with argument is your file name dest.xml. Seem that it's simpler than the python version.

Answer 2

I would use awk if the format doesn't diverge from the sample you provided:

awk -F'[<>="[:blank:]]+' '
  $2 == "domain" {group = $(NF-1)}
  !(group == "group1" && $2 == "node" && $(NF-1) == "PQR")
  ' < dest.xml > new-dest.xml

to remove that "PQR" node in the "group1" domain.

$ diff -u dest.xml new-dest.xml
--- dest.xml    2013-02-22 07:01:48.732227421 +0000
+++ new-dest.xml        2013-02-22 07:02:16.111512820 +0000
@@ -1,6 +1,5 @@
 <domain id="1" group_name="group1">
     <node id="ABC">
-    <node id="PQR">
     <node id="XYZ">
 </domain>
 <domain id="2" group_name="group2">

If you mean you wish to remove that node from the XML file in place, then that's not really possible. You'd need at least to rewrite the part that is after that node to shift the data as many bytes backward.

Alternatively, you could replace that node with blanks which would mean you could get away with replacing those bytes only.

perl -ne '
  if (/<domain.*group_name="(.*?)"/) {
    $in = $1 eq "group1"
  } elsif ($in && /<node id="PQR"/) {
    s/./ /g;
    seek STDOUT,tell(STDIN)-length$_,0;
    print
  }' < dest.xml 1<> dest.xml

Add a ;exit after the print above if there's only one such node and you want to stop processing as soon as you've found it.