How to construct a command from the output of another command?

This awk command will take the output from your srvctl command - or at least one that's formatted similarly to the one posted in the question - and generate a command you can use to repeat the configuation:

srvctl config database -d MYDB | awk -F': *' ' /^Database/ { dun=$2 } /^Oracle home/ { oh=$2 } END { printf "srvctl add database -d %s -o %s\n", dun, oh } ' 

Output:

srvctl add database -d MYDB -o /path/to/home 

Or, to match your requirement more closely:

srvctl config database -d MYDB >srvctl.txt awk -F': *' ' /^Database/ { dun=$2 } /^Oracle home/ { oh=$2 } END { printf "srvctl add database -d %s -o %s\n", dun, oh } ' srvctl.txt 

Output:

srvctl add database -d MYDB -o /path/to/home 

You can, of course, crash the multiple lines of each awk offering into a single line if you must, but it's far less readable:

 awk -F': *' '/^Database/{dun=$2} /^Oracle home/{oh=$2} END{printf "srvctl add database -d %s -o %s\n",dun,oh}' 

Using printf, awk (with : , a colon followed by a space, as the field separator), and readarray:

$ readarray -t a <(srvctl config database -d MYDB | awk -F': ' '/:/ {print $2}') $ printf "srvctl add database -d %q -o %q\n" "${a[@]}" srvctl add database -d MYDB -o /path/to/home 

The output from srvctl ... | awk ... is first read into array $a. The default delimiter for readarray is a newline, so each output line becomes one element of the array. The array is then printed using printf.

The printf format string uses %q rather than %s, so the arguments in the output will be quoted if necessary, just in case they contain spaces or other inconvenient characters.

From help printf in bash:

%q quote the argument in a way that can be reused as shell input

NOTE: The built-in printf in bash, ksh93, and zsh as well as /usr/bin/printf from GNU coreutils support %q. Other implementations of printf might not.

srvctl add database -d MYDB -o /path/to/home is not a Linux command, it's code in the language of most Unix shells that most Unix shells interpret as executing a file (something like /usr/bin/srvctl) whose path is looked up in directories of $PATH with a number of arguments (srvctl, add, database, -d, MYDB, -o, /path/to/home) in a child process and wait for its termination.

In the perl language, the equivalent would be:

system("srvctl", "add", "database", "-d", "MYDB", "-o", "/path/to/home") 

So what you want here is run the srvctl command with srvctl, config, database, -d, MYDB as arguments, collect its output and extract information for that to pass as the arguments following -d and -o in the command above.

perl backronymed as practical extration and report language is the perfect tool to extract information from text.

From a POSIX-like or rc-like shell (or fish), you could do for instance:

LC_ALL=C srvctl config database -d MYDB | perl -lwne ' if (/^\s*(.*?)\s*: (.*?)\s*$/) { $f{lc($1)} = $2; } END { exec( "srvctl", "add", "database", "-d", $f{"database unique name"}, "-o", $f{"oracle home"} ) || die "Can't run srvctl: $!\n" }' 

Where the shell runs the srvctl config and perl commands with a pipe connecting the standard output of the former to the standard input of the latter, and perl parses that output to build an associative arrays and in the END executes srvctl by itself (with exec(), so in the same process, not a child) with some of the arguments being some of the elements of the associative array.

Note the LC_ALL=C which should ensure those Database unique name, Oracle home are not translated to the language of the user, if not English.

Also note the -w option which enables a few warnings and in particular here will warn you if elements with database unique name or oracle home keys are not found in the %f associative array (but will still run the srvctl with empty arguments in place of the missing elements).

You could also do it the whole thing in perl which would make it a bit more complicated but would allow us to check the exit status of srvctl config before running srvctl add. Doing it with maximum error checking:

#! /usr/bin/perl -- use POSIX; $ENV{LC_ALL} = "C"; open(my $output, "-|", qw(srvctl config database -d MYDB)) or die "Can't start srvctl config: $!"; my %field; while (<$output>) { if (/^\s*(.*?)\s*: (.*?)\s*$/) { $field{lc($1)} = $2; } } unless (close($output)) { if ($!) { die "Error closing pipe to srvctl config: $!"; } elsif (WIFSIGNALED(${^CHILD_ERROR_NATIVE})) { die "srvctl config was killed by signal " . WTERMSIG(${^CHILD_ERROR_NATIVE}); } else { die "srvctl config failed with exit status " . WEXITSTATUS(${^CHILD_ERROR_NATIVE}); } } sub get_field { return $field{$_[0]} // die "Could not find $_[0] in srvctl config output" } exec( "srvctl", "add", "database", "-d", get_field("database unique name"), "-o", get_field("oracle home") ) || die "Can't run srvctl add: $!\n"; 

A few shells have associative array support and builtin pattern matching support.

For instance, in zsh (which also supports perl-like regexps), something similar could be done with:

#! /bin/zsh - lines=( ${(f)"$(LC_ALL=C srvctl config database -d MYDB)"} ) || exit typeset -A field set -o re_match_pcre for line ($lines) [[ line =~ '^\s*(.*?)\s*: (.*?)\s*$' ]] && field[$match[1]:l]=$match[2] exec srvctl add database \ -d "${field[database unique name]?}" \ -o "${field[oracle home]?}"