Bash capture groups

gistfile1.md

data:

locator: urn:asset:a1e2d1df-a741-6547-93e2-e84692c8e981
locator: urn:bbc:cps:asset:27510321

script:

#!/bin/bash

while read l
    do
        if [[ $l =~ [(\w*:\s\w*:\w*:)(\w*-.*)] ]]; then
            for i in "${BASH_REMATCH[@]}"
            do
                echo "Index: $i"
            done
            
            echo "${BASH_REMATCH[2]}"
        fi
done < data.txt

#!/bin/bash

while read -r l
    do
        if [[ $l =~ .+:(.*)$ ]]; then
            for i in ${BASH_REMATCH[@]}
            do
                echo "Index: $i"
            done

            echo ${BASH_REMATCH[1]}
        fi
done < data.txt

Not sure why my regex works though! ;-)

@pgchamberlin I think yours works because it's matching the whole string, whereas Crystal's is matching only the individual letters.

@pgchamberlin chars! I knew it could be vastly simplified.

I was using rubular.com to verify the matching. The result was misleading:

Match groups:
1.  locator: urn:asset:
2.  f3e8dae6-d5f5-b34e-999a-e256595c65b5

@edwellbrook that definitely makes sense as doing:

for i in ${BASH_REMATCH[@]}
            do
                echo "Index: $i"
            done

Was echoing what looked like one char at a time. We couldn't work out why. Ah well.

I think Crystal's matches because surrounding the whole regex in [ ] makes Bash treat the whole regex as a single character class, and because that class contains characters that could match anything it does match everything - so you're right, it is matching all the individual letters, but is capturing none of them.

Looking at mine again, of course I had forgotten that .+ matches spaces, so it matches all characters up until the final :, then captures everything up to the end.

Rubular uses PCRE and Bash uses ERE. There's subtle differences, but they're there.