How to replace all but last matching in a file using bash? - bash

Question

Assuming using bash, having a configuration file like:
param-a=aaaaaa
param-b=bbbbbb
param-foo=first occurence <-- Replace
param-c=cccccc
# param-foo=first commented foo <-- Commented: don't replace
param-d=dddddd
param-e=eeeeee
param-foo=second occurence <-- Rreplace
param-foo=third occurence <-- Last active: don't replace
param-x=xxxxxx1
param-f=ffffff
# param-foo=second commented foo <-- Commented: don't replace
param-x=xxxxxx2
In which you can find multiple commented or uncommented lines of the param-foo,
how can you comment all the uncommented param-foos except the very last active one,
resulting in:
param-a=aaaaaa
param-b=bbbbbb
# param-foo=first occurence <-- Replaced
param-c=cccccc
# param-foo=commented foo <-- Left
param-d=dddddd
param-e=eeeeee
# param-foo=second occurence <-- Replaced
param-foo=third occurence <-- Left
param-x=xxxxxx1
param-f=ffffff
# param-foo=second commented foo <-- Left
param-x=xxxxxx2
Two parts of the question:
1. How to do it with only one known repeating param?
(only param-foo in the example above)
2. How to do it with all multiple active params at once?
(param-foo + param-x in the example above)
Attention: In this case I don't know previously the name of the repeating params!
Thanks

Answer 1

If awk is acceptable, this will do it for param-foo and param-x:
awk -F= -v p='param-foo param-x' 'BEGIN {
ARGV[ARGC++] = ARGV[ARGC - 1]
n = split(p, t, OFS)
for (i = 0; ++i <= n;) _p[t[i]]
}
NR == FNR {
$1 in _p && nr[$1] = NR
next
}
$1 in nr && FNR != nr[$1] {
$0 = "# " $0
}1' infile
You may use a single parameter: p=param-x or add more parameters separated by spaces: p='param-1 param-2 ... param-n'.
Edit: I'm assuming the real input file looks like this:
param-a=aaaaaa
param-b=bbbbbb
param-foo=first occurence
param-c=cccccc
# param-foo=commented foo
param-d=dddddd
param-e=eeeeee
param-foo=second occurence
param-foo=third occurence
param-x=xxxxxx1
param-f=ffffff
param-x=xxxxxx2
Let me know if it's different.
Second edit: providing a solution for mawk users:
awk -F= -v p='param-foo param-x' 'BEGIN {
n = split(p, t, OFS)
for (i = 0; ++i <= n;) _p[t[i]]
}
NR == FNR {
$1 in _p && nr[$1] = NR
next
}
$1 in nr && FNR != nr[$1] {
$0 = "# " $0
}1' infile infile
Adding solution for the latest requirement:
awk -F= 'NR == FNR {
if (NF && !/^#/)
_p[$1]++ && nr[$1] = NR
next
}
$1 in nr && FNR != nr[$1] {
FNR != nr[$1] && $0 = "# " $0
}1' infile infile

Answer 2

I have not tested fully the script, but it worked on the first example:
#!/bin/bash
input_file=/path/to/your/input/file
last_occurence=`nl $input_file | grep 'param-foo' | grep -v '#' | tail -1 | awk -F" " '{print $1}'`
sed -i '/#/!s/param-foo/# param-foo/g' $input_file
sed -i "${last_occurence}s/# param-foo/param-foo/" $input_file
It's very straight forward logic. First we get the last occurrence of param-foo, which is not commented.
The first sed goes and comments all param-foo, which are not commented.
The second sed uses the line_number of last occurence of param-foo and removes the # character. You can easily wrap that in a function and use it inside a loop, providing a list of parameters, instead of only one.

Answer 3

A bit slow for long files, but should work for all the parameters:
grep -v ^# $file |
cut -f1 -d= |
sort -u |
sed 's/^/grep -n . '$file' |
tac |
grep -m1 :/;s/$/= /' |
bash |
sed -r 's%([0-9]+):(.*)=(.*)%\1!s/^\2=/# \2=/%' |
sed -f- $file

Answer 4

This might work:
param="param-foo"
tac input_file |sed '/#/!{/'"$param"'/{x;/./{x;s/'"$param"'/# &/;t};x;h;}}'|tac >output_file
For multiple params:
cp input_file{,.backup}
params=(param-{foo,bar,baz})
tac input_file >backwards_file
for param in "${params[#]}"; do
sed -i '/#/!{/'"$param"'/{x;/./{x;s/'"$param"'/# &/;t};x;h;}}' backwards_file
done
tac backwards_file >output_file
Turn input_file backwards, preprend all but the first occurrence of $param with a comment #,then revert the file.
EDIT:
To extract the params from the file use this piece of code:
params=($(sed -rn '/^#/d;/^$/!s/^\s*([^=]*).*/\1/gp' input_file | sort | uniq))