Add variables to a bash array in a loop-CodePudding

I have a json-file with keys and values that I would like to add to a bash array, but I cannot make it work.

1. Creating JSON-file with data

# JSON file
echo '{"persons": [{"name":"John", "age":30, "car":"big"}, {"name":"Alice", "age":29, "car":"big"}]}' > my.json

2. Iterating over Name and Age

# Looping over JSON
jq -c '.persons[]' my.json | while read p;
do
    name=$(echo $p | jq -r '.name')
    age=$(echo $p | jq -r '.age')
    echo "Name: $name"
    echo "Age: $age"
done

Names and ages are printed out as expected.

3. Manually adding Name and Age to Array

# Creating array and adding elements
declare -A mylookup
key="James"
value="31"
mylookup[$key]=$value

4. Checking if values were added

for key in "${!mylookup[@]}"
do
    echo "Key: $key"
    echo "Value: ${lookup[$value]}"
done

I oddly only shows the Key correctly

5. Adding Name and Age to Array in loop

# Creating array and adding elements
declare -A mylookup
jq -c '.persons[]' my.json | while read p;
do
    name=$(echo $p | jq -r '.name')
    age=$(echo $p | jq -r '.age')
    mylookup[$name]=$age
done

6. Printing out keys and values in Array

Now I check what was added to the array.

for key in "${!mylookup[@]}"
do
    echo "Key: $key"
    echo "Value: ${lookup[$value]}"
done

Nothing is printed out..... where am I failing? I suspect something is going wrong when I am adding the keys and values to the array, but I cannot find out what.

CodePudding user response：

Array values are getting lost because you are using a pipeline here. Since while loop is running in a forked sub-shell therefore all the variables are being set in the forked subshell only and you are seeing no changes in the parent shell.

You may use this to avoid this:

declare -A mylookup

while IFS=$'\t' read n a; do
   mylookup[$n]="$a"
done < <(jq -r '.persons[] | "\(.name)\t\(.age)"' my.json)

# check array content
declare -p mylookup
declare -A mylookup=([Alice]="29" [John]="30" )

< <(...) is process substitution

CodePudding user response：

The usual way would be to make jq output a format that the shell can read accurately. Here's an example with the lossless TSV format:

#!/bin/bash

declare -A mylookup

while IFS=$'\t' read -r name age
do
    printf -v name '%b' "$name"
    printf -v age  '%b' "$age"
    mylookup["$name"]=$age
done < <(
    echo '{"persons": [{"name":"John", "age":30, "car":"big"}, {"name":"Alice", "age":29, "car":"big"}]}' |
    jq -r '.persons[] | [ .name, .age ] | @tsv'
)

declare -p mylookup
#=> mylookup='([Alice]="29" [John]="30" )'

^{remark: The printf -v ... '%b' ... are meant to unescape the TSV values; they're probably unnecessary in your case}

CodePudding user response：

Alternatively, you can use jq to output a valid bash syntax for Bash's declare -A statement.

With proper @sh filtering in jq, it ensures no invalid or unintended execution can occurs from the JSON data.

This method saves the use of a shell iteration which is very inefficient.

#!/usr/bin/env bash

json_file=my.json

if dyn_declare=$(
  jq --raw-output '"(",(.persons[]|"[" (.name|@sh) "]=" (.age|@sh)),")"' \
    "$json_file"
) && declare -A mylookup=$dyn_declare; then
  # Do something useful with the populated associative array
  # Here we just dump it for debug/demo purpose
  declare -p mylookup
else
  printf 'Could not parse %s into an associative array\n' \
    "$json_file" >&2
  exit 1
fi