Picking selected items from stdin via choices typed into console

Question

Can anyone point me at a 'pick' filter script that works somewhat as described below?

I've spent about an hour hunting for a simple bash script/filter that will allow me to pipe in a list of values and will spit out a subset of them depending on choices I make at the console. I know there are examples written in C but I wanted a mostly-portable bash script I can use in Cygwin / Gitbash etc. (The context: I want to be able to run some command in some subdirectories, and I want to separate the choice of which directories to run the command, from the choice of command to run.)

As hypothetical example of usage:

$ echo "foo
> bar
> baz" | pick.sh
* Options:
* 1. foo
* 2. bar
* 3. baz
* Choices? 2 3 
bar
baz

The lines marked * are supposed to be where the script interactively lets me choose which elements to 'pick' and once I decided lines 2 and 3 it proceeds to send those out to STDOUT.

Choices ideally could be a combination of space-separated numbers eg 2 3 4, inclusive ranges eg 2-4 .. or maybe even fancy enough to use some kind of autocompletion allowing typing the first few letters of the items themselves.

Well, there it is, I think it would be a very useful bash pipeline filter in general!

(Thanks for reading this far..)

Answer 1

Self-demonstrating example using sed line specifiers (where N,M means lines N through M):

$ ./pick.sh < ./pick.sh
Lines:
     1  #!/usr/bin/env bash
     2  
     3  set -o errexit -o nounset -o noclobber
     4  
     5  trap 'rm --force --recursive "$working_directory"' EXIT
     6  working_directory="$(mktemp --directory)"
     7  input_file="${working_directory}/input.txt"
     8  
     9  cat > "$input_file"
    10  
    11  echo 'Lines:'
    12  cat --number "$input_file"
    13  
    14  IFS=' ' read -p 'Choices: ' input < /dev/tty
    15  lines=($input)
    16  
    17  sed --quiet "$(printf '%sp;' "${lines[@]}")" "$input_file"
Choices: 1 5,7 17
#!/usr/bin/env bash
trap 'rm --force --recursive "$working_directory"' EXIT
working_directory="$(mktemp --directory)"
input_file="${working_directory}/input.txt"
sed --quiet "$(printf '%sp;' "${lines[@]}")" "$input_file"

Basically, save standard input to a temporary file, print the file with line numbers, prompt for input ranges, and pass the input ranges to sed to print each of them.

One quirk of this method is that lines will be printed in the order they appear in the file, not the order you specify:

…
Choices: 3 1
#!/usr/bin/env bash
set -o errexit -o nounset -o noclobber

If you really need input order it would be simple to loop over lines, although this is of course less efficient.

The script assumes you have GNU cat, sed, etc. installed. If you're using BSD tools the command flags will be different.