sed remove section between curly braces - sed

Question

I'm trying to delete a section elasticsearch { * } from a configuration file due to lack other options.
output {
if [env] {
if [containerimage] and [containerimage] != "apache" {
file {
path => "/var/log/logstash/logs/%{[env]}/%{file}-%{+YYYY-MM-dd}"
}
elasticsearch {
hosts => ["{{elasticsearch/host}}:{{elasticsearch/port}}"]
template_overwrite => true
}
}
}
}
I tried to reverse the sed command found on stackoverflow but it deletes ending } from configuration files too.
[root#indexer conf.d]# sed '/{/{:1; /}/!{N; b1}; /elasticsearch/!p}; d' example
output {
if [env] {
if [containerimage] and [containerimage] != "apache" {
file {
path => "/var/log/logstash/logs/%{[env]}/%{file}-%{+YYYY-MM-dd}"
It may be due to }} in to cfg that must be there. Any ideas to overcome this problem? Thanks in advance
Edit:
some other experiment:
sed '/elasticsearch {/{:1; /}/!{N; b1} }; /elasticsearch/!p; d' example
output {
if [env] {
if [containerimage] and [containerimage] != "apache" {
file {
path => "/var/log/logstash/logs/%{[env]}/%{file}-%{+YYYY-MM-dd}"
}
template_overwrite => true
}
}
}
}
But leaves template_overwrite and some closing braces.

Answer 1

This might work for you (GNU sed):
sed '/^\s*elasticsearch {$/{:a;N;/^\s*}$/M!ba;d}' file
Gather up the lines beginning elasticsearch { and ending } and delete them.
N.B. This uses the M flag for a multiline match for the end condition which is GNU sed specific. An alternative:
sed -n '/^\s*elasticsearch {$/{:a;n;/^\s*}$/!ba;d};p' file

Answer 2

Here's one in awk:
awk '
BEGIN {
RS="^$" # read in the whole file
c=1 # brace counter
}
{
match($0,/\n? *elasticsearch \{/) # find the start of string to remove
print substr($0,1,RSTART-1) # RUNNING OUT OF BATTERY
$0=substr($0,RSTART+RLENGTH)
while(c>0) {
match($0,/(\{|\} *\n*)/)
if(substr($0,RSTART,1)=="{")
c++
else
c--
$0=substr($0,RSTART+RLENGTH)
}
print
}' file
Output:
output {
if [env] {
if [containerimage] and [containerimage] != "apache" {
file {
path => "/var/log/logstash/logs/%{[env]}/%{file}-%{+YYYY-MM-dd}"
}
}
}
}
Only tested with provided data.

Answer 3

Is this all you're trying to do?
$ sed '/elasticsearch[[:space:]]*{/,/^[[:space:]]*}/d' file
output {
if [env] {
if [containerimage] and [containerimage] != "apache" {
file {
path => "/var/log/logstash/logs/%{[env]}/%{file}-%{+YYYY-MM-dd}"
}
}
}
}
The above will work with any POSIX sed. If that's not all you need then please edit your question to show a more truly representative example so we can help you. FWIW I wouldn't really do the above as it's not extensible if/when your requirements vary slightly, I'd do this instead:
$ awk '/elasticsearch[[:space:]]*{/{f=1} !f; /^[[:space:]]*}/{f=0}' file
output {
if [env] {
if [containerimage] and [containerimage] != "apache" {
file {
path => "/var/log/logstash/logs/%{[env]}/%{file}-%{+YYYY-MM-dd}"
}
}
}
}