FAQ contains error regarding import of bash arrays

[chepner]

The FAQ suggests using the following to encode a bash array as a JSON array:

jq -n --argjson args "$(printf '\"%s\"\n' "${x[@]}" | jq -s)" '$args'

This assumes, however, that the elements of the array are themselves valid JSON values. A counterexample is x=(1 $'a\nb' 2), where the second element contains a literal newline.

I believe the only safe way to do this is with multiple calls to jq, building the array one element at a time:

arr='[]'
for element in "${x[@]}"; do
  arr=$(jq -n --arg elt "$element" --argjson "$arr" '$arr + [$x]')
done

to ensure that each element of x is properly encoded before adding it to the array.

[pkoppstein]

The FAQ explicitly says:

This approach is applicable so long as none of the values contains a newline character.

It then goes on to provide a generic approach that uses NUL to avoid the iteration that your proposed solution entails.

[chepner]

$ x=(1 $'a\nb' 2)
$ jq -n --argjson args "$(printf '%s\0' "${x[@]}" | jq -Rc 'split("\u0000")')" '$args'
jq: invalid JSON text passed to --argjson
Use jq --help for help with command-line options,
or see the jq manpage, or online docs  at https://stedolan.github.io/jq

The problem is not the delimiter, but the assumption that each element of the array is already a valid JSON string.

[pkoppstein]

The problem (now fixed) was the example forgot to include the -s option:

$ x=(1 $'a\nb' 2)
$  jq -n --argjson args "$(printf '%s\0' "${x[@]}" | jq -Rsc 'split("\u0000")')" '$args'
[
  "1",
  "a\nb",
  "2"
]

Thanks for keeping us honest!

[trailstrider]

Another candidate to close per #2305 ?