Question

I’m experimenting with bash scripts, need help solving this:
i have the following data in a text file: (test.txt)

have.a.nice.day.a=42and55  
have.a.nice.day.b=0  
have.a.nice.day.c=55and67  
go.to.sleep.a=111  
go.to.sleep.b=2and122and33  
go.to.sleep.c=64  

i want to separate the strings from their matching scores and the scores from their delimiters (in this case: “and”) and pick the string with highest score from each group.
in this case it would be “have.a.nice.day.c” for group “have.a.nice.day” and “go.to.sleep.b” for group “go.to.sleep”
so i thought best thing to do would be to separate the elements and assign them variables recursively. like so:

#!/bin/bash
names=$(cat test.txt | grep -o -P '.+(?==\d+)')
 for name in $names
 do
  echo -n "$name"" "
  scores=$(cat test.txt | grep -o -P '(?<='$name'=).+')
   for scores_group in $scores
   do
    single_score=$(echo $scores_group | grep -o -P '\d+') 
     for score in $single_score
     do
      echo -n "$score"" "
     done
     echo
   done
 done  

the output would be:

have.a.nice.day.a 42 55 
have.a.nice.day.b 0 
have.a.nice.day.c 55 67 
go.to.sleep.a 111 
go.to.sleep.b 2 122 33 
go.to.sleep.c 64  

but now i have no idea how to find the best score for each group.
thanks

Answer 1

So your actual question, I think, is how do you take “input text” that you’ve labelled as “output”, and find the line with the highest number?

Assuming your output is the input, I’d do this with awk:

$ awk '{name=$1; item=$1; sub(/\.[^.]+$/,"",name); sub(/.*\./,"",item); for (i=2; i<=NF; i++) {if($i>highest[name]){highest[name]=$i;which[name]=item}}} END{for(name in highest){printf("%s = %s = %s\n",name,which[name],highest[name])}}' input.txt
go.to.sleep = b = 122
have.a.nice.day = c = 67

Or, broken out in order to explain:

{

  # Get the parts of the first field...
  name=$1; sub(/\.[^.]+$/,"",name);
  item=$1; sub(/.*\./,"",item);

  # Walk through the scores, record the highest in an array
  for (i=2; i<=NF; i++) {
    if ($i>highest[name]) {
      highest[name]=$i;
      which[name]=item;
    }
  }
}

# Now, step through the resultant array
END {
  for (name in highest) {
    printf("%s = %s = %s\n",name,which[name],highest[name]);
  }
}

Will this do? Or do you really want to achieve this in pure bash? If so, the awk above can be expressed with the bash below:

#!/bin/bash

declare -A highest
declare -A which

while read word scores; do
    name=${word%.*}
    item=${word##*.}
    set -- $scores
    while [[ -n "$1" ]]; do
        if [[ $1 -gt highest[$name] ]]; then
            highest[$name]=$1
            which[$name]=$item
        fi
        shift
    done
done < input.txt

for name in "${!highest[@]}"; do
    printf "%s = %s = %s\n" "$name" "${which[$name]}" "${highest[$name]}"
done