About linux : Extract-a-word-between-brackets-and-replace-it-in-a-sentence

so let’s say i have this sentence

It's {raining|snowing|cold} outside

What I want is to randomly extract a word between the brackets, which i did with awk -vRS="}" -vFS="|" '{print $2}' (still working to extract them randomly). The output usually is the second word, in our case snowing.

Thing is that the output is only snowing, and the actual output i want is something like It's snowing outside, so how do I extract any word from the brackets and replace with only one word.

Using any awk in any shell on every Unix box:

$ cat tst.awk
BEGIN { srand() }
match($0,/\{[^}]+}/) {
    wordList = substr($0,RSTART+1,RLENGTH-2)
    numWords = split(wordList,words,/[|]/)
    wordNr = int(rand() * numWords + 1)
    $0 = substr($0,1,RSTART-1) words[wordNr] substr($0,RSTART+RLENGTH)
}
{ print }

$ awk -f tst.awk file
It's snowing outside

$ awk -f tst.awk file
It's cold outside

See rand() (and srand()) at https://www.gnu.org/software/gawk/manual/gawk.html#Numeric-Functions for details on it’s randomness.

An interesting solution is to (ab)use Perl’s glob function to generate a list of possible combinations. The caveat is that you cannot use glob meta characters in the string. For example:

use strict;
use warnings;
use Data::Dumper;
use feature 'say';

my $str = "It's {freaking|fudging} {raining|snowing|cold} outside";

$str =~ s#({[^}]+})#$1 =~ s/\|/,/gr#ge;   # replace | with , inside curly braces
my @list = glob $str;
print Dumper \@list;

say "Our random string: " . $list[ int rand(@list) ];

Output:

$VAR1 = [
          'It\'s freaking raining outside',
          'It\'s freaking snowing outside',
          'It\'s freaking cold outside',
          'It\'s fudging raining outside',
          'It\'s fudging snowing outside',
          'It\'s fudging cold outside'
        ];
Our random string: It's fudging cold outside

Since you need to rewrite a string by finding and replacing a part of it, regex is suitable

use warnings;
use strict;
use feature 'say';

sub pick_one { 
    my ($pattern) = @_; 
    my @choices = split /\|/, $pattern; 
    return $choices[int rand @choices]; 
}

my $sentence = q(It's {raining|snowing|cold} outside); 

$sentence =~ s/\{ ( [^}]+ ) \}/pick_one($1)/ex; 

say $sentence; 

That /e modifier makes it evaluate the replacement side as code, and the produced value is then used as the replacement. So there we run a sub in which the choosing happens. Having this in a sub is a good way for later refinements/changes, implemented in the sub.

An element of the array @choices is selected using rand. An expression for its upper bound is evaluated in the scalar context so we can directly use the @choices array since then its length ends up being used.

echo "It's {raining|snowing|cold} outside" |\
  perl -pe 's/\{(.*?)\}/(split("[|]",$1)[rand(3)]/e'

or for arbitrary number of weather conditions:

echo "It's {raining|snowing|cold} outside" |\
  perl -pe 's/\{(.*?)\}/@a=split("[|]",$1); $a[rand(@a)]/e'

With jq it is a one-liner:

echo "It's {raining|snowing|cold} outside" | \
jq -rR --argjson rand $RANDOM 'gsub("{(?<words>[^}]+)}"; .words | split("|") | .[$rand % length])'

echo "It's {raining|snowing|cold} outside"|awk -F"[{|}]" -v w="$(shuf -i 2-4 -n 1)" '{print $1$w$NF}'

In plain bash, without using any external command:

#!/bin/bash

while read -r line; do
    if [[ $line =~ (.*)\{([^}]*)}(.*) ]]; then
        IFS='|' read -ra words <<< "${BASH_REMATCH[2]}"
        line=${BASH_REMATCH[1]}${words[RANDOM%${#words[*]}]}${BASH_REMATCH[3]}
    fi
    printf '%s\n' "$line"
done < file