Multiline pattern match and delete lines - sed/awk/grep

Question

Input.txt

-------
Database alias = ABC
Node name = node01
Hostname = hostnode01
Service name = 12345
-------
Hostname    = hostnode01
Service name    = 12345
-------
Database alias  = PQR
Node name   = node01
Hostname    = hostnode01
Service name    = 12345
-------
Hostname    = hostnode01
Service name    = 12345
-------
Database alias  = XYZ ......

Expected Output

-------
Database alias  = ABC
Node name   = node01
Hostname    = hostnode01
Service name    = 12345
-------
Database alias  = PQR
Node name   = node01
Hostname    = hostnode01
Service name    = 12345
-------
Database alias  = XYZ ......

Idea is to remove the Duplicate Values of Hostname & Service Name if repetative or you can say Remove lines between '-------', if there are 2 lines in between.

Tried using sed with Multiple Match, but not getting desired o/p

sed '/-------/{$!N;/\n.*Hostname/d;}' Input.txt

Answer 1

This isn't a job for sed (you shouldn't depend upon the duplicate entries being consecutive). Use awk:

1. Set the record separator to the regex ^-+$.

2. Use an associative array, where the index is a concatenation of HostName and ServiceName. For any record read, place the record in the array only if it does not already exist.

3. At the end of processing the file, print out all elements of the array, formatted into separate lines, with your ---- delimiter lines.

Answer 2

GNU awk approach:

awk 'BEGIN{ 
         RS = ORS = "-------\n";
         pat = "^Hostname += ([^ ]+)\nService name += ([^ ]+)\n$";  
     }
     NR == FNR{ 
         if (match($0, pat, a)) { items[a[1], a[2]]++ } 
         next 
     }
     match($0, pat, a) && (a[1], a[2]) in items \
     && items[a[1], a[2]] > 1{ next }1' inout.txt input.txt

The output:

-------
Database alias = ABC
Node name = node01
Hostname = hostnode01
Service name = 12345
-------
Database alias  = PQR
Node name   = node01
Hostname    = hostnode01
Service name    = 12345
-------
Database alias  = XYZ ......
-------

Answer 3

Perhaps the simplicity of your sample dataset is not typical of the actual use case, but given the triviality of the example you post, grep alone would suffice:

$ grep --no-group-separator -A4 '^Database alias' input.txt 
Database alias = ABC
Node name = node01
Hostname = hostnode01
Service name = 12345
-------
Database alias  = PQR
Node name   = node01
Hostname    = hostnode01
Service name    = 12345
-------
Database alias  = XYZ ......