Bash script: Matching extended glob pattern to a string array

Question

Hello Amazing People,

I have been trying to achieve a goal of matching an extended glob pattern to a string array (file paths) and getting the list of file which matches to the pattern. Below is the code which works for one ext glob pattern but not for the other similar ext glob pattern.

This is the pattern which doesn't work. pattern="**cars-+(!(bad-cats|dogs))/src/bin/out/**"

Here is the shell code:

#!/bin/bash

# Enable extended globbing
shopt -s extglob


# pattern="**cars-!(bad-cats)/src/bin/out/**" # this extended glob pattern works 
pattern="**cars-+(!(bad-cats|dogs))/src/bin/out/**" # this extended glob pattern doesn't work and also goes in a very long loop cos of cars-ok-kk/src/main/out/start.txt

files="cars-white-pens/src/bin/out/file.txt,cars-ok-kk/src/main/out/start.txt,cars-grey-dogs/src/bin/out/bottle.txt,cars-bad-cats/src/bin/out/computer.txt,cars-whales/src/bin/mouse.txt,cars-dogs/src/bin/out/mouse.txt"
IFS=',' read -r -a files_array <<< "$files"

matching_files=""

for file in "${files_array[@]}"; do
  if [[ $file == $pattern ]]; then
    if [ -z "$matching_files" ]; then
      matching_files="$file"
    else
      matching_files="$matching_files,$file"
    fi
  fi
done

echo "Match: $matching_files"

PS: We can change the ext glob pattern too if needed for the code to work but I want the pattern to be ext glob pattern only with support of exclusion of one or more directories as shown in the pattern.

Thanks in advance.

The commented ext glob pattern pattern="**cars-!(bad-cats)/src/bin/**" works fine with the shell script but the other one doesn't.

When I run it, it takes some time (not sure why) and then prints below:

Match: cars-white-pens/src/bin/out/file.txt,cars-grey-dogs/src/bin/out/bottle.txt,cars-bad-cats/src/bin/out/computer.txt,cars-dogs/src/bin/out/mouse.txt

It removed src/main/out and src/bin/mouse.txt but failed to remove cars-bad-cats and cars-dogs file strings.

I expect output as Match: cars-white-pens/src/bin/out/file.txt,cars-grey-dogs/src/bin/out/bottle.txt since src/main/out and src/bin/mouse.txt doesn't match and cars-bad-cats and cars-dogs is excluded.

Answer 1

+(!(bad-cats|dogs)) matches bad-cats because it can (for instance) match it as strings bad followed by -cats, neither of which matches bad-cats or dogs (so the !(dogs|bad-cats) pattern matches both of them). Similarly, dogs can be matched as strings do followed by gs.

To fix the code change the pattern assignment to:

pattern='*cars-!(dogs|bad-cats)/src/bin/out/*'

• The pattern !(bad-cats|dogs) does not match eitherbad-cats or dogs because it doesn't allow them to be broken into parts that can be matched separately.
• Removing the +( and corresponding ) in the pattern also massively speeds up the matching because it eliminates huge amounts of pointless backtracking.
• I replaced each of two occurrences of ** in the original pattern with * because the second * is redundant, and might make pattern matching slower. (** is only special when doing filename expansion and the globstar shell option is enabled (see the globstar section in glob - Greg's Wiki).)

Answer 2

Not an answer but a workaround:

case $file in
*cars-bad-cats/* | *cars-dogs/* ) ;;
*cars-*/src/bin/out/* )
    matching_files=${matching_files:+$matching_files,}$file
esac