sed regex is replacing an entire file - regex

Question

I have hdfs-site.xml file which contains following information
<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<!-- Put site-specific property overrides in this file. -->
<configuration>
<property>
<name>dfs.replication</name>
<value>3</value>
</property>
<property>
<name>dfs.name.dir</name>
<value>/data/dfs/nn</value>
</property>
<property>
<name>dfs.data.dir</name>
<value>/data/dfs/dn,/mnt_test_volume/data/dfs/dn,/mnt_test_volume/data/dfs/dni,/mnt_test_v5olume/data/dfs/dn,/mnt_test_volume/d5ata/dfs/dgn</value>
</property>
<property>
<name>dfs.permissions</name>
<value>false</value>
</property>
</configuration>
I want to remove some of the entries present in <name>dfs.data.dir</name><value> and </value></name> tags. Which entry to remove is decided by one parameter to shell script.
I am new to sed and I have written following sed command to find particular entry and delete it. This works as expected when sed is executed very first time but when same command is executed next time, all the contents of the file are wiped out and file becomes a blank file.
sed -ni '1h; 1!H; ${g; s#\(<name>dfs\.data\.dir<\/name>[^a-zA-Z0-9]*<value>.*\)'$data_dir_path'[^,<]\(.*<\/value>\)#\1\2# p}' hdfs-site.xml
In this command $data_dir_path variable decides which entry to be deleted.
For example, if value of data_dir_path is /mnt_test_volume/data/dfs/dn then I am expecting following output
<name>dfs.data.dir</name> <value>/data/dfs/dn,,/mnt_test_volume/data/dfs/dni,/mnt_test_v5olume/data/dfs/dn,/mnt_test_volume/d5ata/dfs/dgn</value>
which is working fine when command is executed once but if same command is executed next time, entire file becomes empty.
Can anyone please tell me what am I doing wrong here?

Answer 1

You can use a much simpler sed as
sed "/<name>dfs.data.dir<\/name>/ {n; s#$data_dir_path##}" hdfs-site.xml
What it does?
-i inplace editing of the file
'/<name>dfs.data.dir<\/name>/ checks if the line matches the pattern. If yes then the commands following are excecuted. Note that the commands following are grouped in {} as {n; s/'$data_dir_path'//}'
n; reads the next line from file into the pattern space
s/'$data_dir_path'// substiture the value in $data_dir_path with null
Test
$ sed "/<name>dfs.data.dir<\/name>/ {n; s#$data_dir_path##}" test
bash-3.2$ cat test
:
:
:
<name>dfs.data.dir</name>
<value>/data/dfs/dn,,i,/mnt_test_v5olume/data/dfs/dn,/mnt_test_volume/d5ata/dfs/dgn</value>
:
:
: