Using bash to regex on output of function to evaluate another function using found regex groups

Question

I have a bash function, FUNCTION_A who's output looks like this:

    ID          HELD_SINCE  HOLD_REASON
    123456.0   7/1  20:40 Peak usage: 2370 megabytes.
    123457.0   7/1  20:40 Peak usage: 5008 megabytes.

I then need to grab the ID number and the number after Peak usage: to pass to FUNCTION_B.

If I were running it by hand it would look like:

FUNCTION_B 123456.0 2370
FUNCTION_B 123457.0 5008

I am trying to write a bash script using regular expressions to grab these two numbers and run in the shell for an arbitrary number of output lines of FUNCTION_A. I'm getting stuck at a few different places.

I am not sure how to loop through output line by line, and I'm not sure why the array isn't holding any output, but I want to do something like this

re='([0-9-]+\.[0-9-]+).*Peak usage: ([0-9-]+) megabytes\.'
for line in FUNCTION_A output
   if [[ line =~ $re ]]; then
      myArray=($(echo s | egrep -Eo re))
      eval FUNCTION_B ${myArray[0]} ${myArray[1]}
   fi  

Answer 1

You can use the capturing parentheses in bash: use the BASH_REMATCH variable

function_A_output=$(FUNCTION_A)

re='([0-9-]+\.[0-9-]+).*Peak usage: ([0-9-]+) megabytes\.'
while IFS= read -r line; do
   if [[ $line =~ $re ]]; then
      FUNCTION_B "${BASH_REMATCH[1]}" "${BASH_REMATCH[2]}"
   fi  
done <<<"$function_A_output"

No eval required.

Using a Process Substitution, we don't need the $function_A_output variable:

re='([0-9-]+\.[0-9-]+).*Peak usage: ([0-9-]+) megabytes\.'
while IFS= read -r line; do
   if [[ $line =~ $re ]]; then
      FUNCTION_B "${BASH_REMATCH[1]}" "${BASH_REMATCH[2]}"
   fi  
done < <$(FUNCTION_A)
# ....^ that space is intentional

Answer 2

I wouldn't write this in bash or any dialect of sh (for the reasons already mentioned in comments), but it seems like you're trying to extract numeric values from the output of one function and pass them two at a time to a second function?

If that's correct, you could do something like this:

re='([0-9-]+\.[0-9-]+).*Peak usage: ([0-9-]+) megabytes\.'
while read -r a b ; do
  FUNCTION_B "$a" "$b"
done < <(FUNCTION_A | sed -nE "s/$re/\1 \2/p")

I'm not entirely sure that I trust that, though - it's far too easy for whatever is in $re to break the sed script. It would be better to pass $re to awk via awk's -v option or to either awk or perl as an exported environment variable.

e.g.

export re='([0-9-]+\.[0-9-]+).*Peak usage: ([0-9-]+) megabytes\.'
while read -r a b ; do
  FUNCTION_B "$a" "$b"
done < <(FUNCTION_A | perl -n -e 'print "$1 $2\n" if m/$ENV{re}/')

However, I strongly recommend you put some time into learning awk and/or perl. sed and grep too.