9

I want to write a script to reference multiple arrays from another array which holds these array's variable names.

Here's my code so far:

#!/bin/bash
array1=('array1string1' 'array1string2')
array2=('array2string1' 'array2string2')

array_names=('array1' 'array2')

for a in ${array_names[@]}
do
        for b in ${a[@]}
        do
                echo $b
        done
done

I'd like the output to scan through both arrays (from the outer for loop) and print the respective strings in the inner for loop which calls echo. My current output is just showing me: 

array1
array2

I'd be grateful for any pointers regarding this. Thank you!

chnppp
  • 93
  • Is there any reason why you can't just do for b in "${array1[@]}" "${array2[@]}"; do ...; done? – Kusalananda Sep 06 '17 at 17:14
  • I'd like the number of arrays to be flexible. Therefore, if I add an array later, I would just add it to array_names and let the loop take care of it. – chnppp Sep 06 '17 at 17:22
  • I think this is a case for indirect expansion. See https://stackoverflow.com/questions/8515411/what-is-indirect-expansion-what-does-var-mean - but, basically, changing ${a[@]} to ${!a} does what you want (I think). – parkamark Sep 06 '17 at 17:24
  • 1
    @parkamark No, that just gives him the first element of each array. And ${!a[@]} gives a length of the array a. – Kusalananda Sep 06 '17 at 17:27
  • Yes, changing to ${!a} is just giving me the first elements. – chnppp Sep 06 '17 at 17:29

3 Answers3

9

Bash 4.3 and later supports "name references", or namerefs (a similar concept exists in ksh93, but the scoping is annoyingly different):

#!/bin/bash

array1=('array1string1' 'array1string2')
array2=('array2string1' 'array2string2')

array_names=('array1' 'array2')

for a in "${array_names[@]}"; do
    declare -n arr="$a"

    for b in "${arr[@]}"; do
        echo "$b"
    done
done

The variable arr is a nameref that acts like an alias for the named variable (the variable with name $a in this example).

Without namerefs, in earlier Bash versions, one solution would be to create a new array that contains all the elements from the other arrays:

all=( "${array1[@]}" "${array2[@]}" )

... a bit like the array_names array in the question but with the contents of all arrays, and then iterate over "${all[@]}".

It's also possible to use eval, but the resulting code looks astoundingly awful.

See glenn jackman's answer for a variation with variable indirection (introduced in its current form with Bash version 2).

Kusalananda
  • 333,661
2

@Kusalananda has the best answer for recent versions of bash. For earlier versions, you can use an indirect variable:

for a in ${array_names[@]}; do 
    tmp="${a}[@]"
    for b in "${!tmp}"; do echo "$b"; done     # or: printf "%s\n" "${!tmp}"
done

See the 4th paragraph of https://www.gnu.org/software/bash/manual/bashref.html#Shell-Parameter-Expansion

glenn jackman
  • 85,964
0

As a variation on what's been said:

#!/bin/bash

array1=('array1 string1' 'array1 string2')
array2=('array2 string1' 'array2 string2')
array_names=('array1' 'array2')

for (( i=0; i<${#array_names[@]}; i++ )); do

    declare -n arr="${array_names[i]}"

    for (( j=0; j<${#arr[@]}; j++ )); do
        echo "${arr[j]}"
    done

done

Accessing the elements by index instead

chris
  • 21