Bash; creating arrays which names are computed from other variables

Question

My goal is to create a list of partitions for each block device listed in /sys/block;

#!/bin/bash
block_devices_list=($(ls /sys/block))
partition_list=($(cat /proc/partitions | awk '{print $4}'))
unset partition_list[0]

for block_device in ${block_devices_list[@]}; do
    for partition in ${partition_list[@]}; do
        partitions+=($(echo $partition | grep $block_device))
    done
    # Right here?
    unset partitions
done

Every time the outside 'for loop' completes it's cycle it ends up with an array of partitions for a particular block device. At that point I would like to transfer that data to a separate array, dynamically named after the device it belongs to (like 'partitions_sda' for example).

I have read a few questions/answers about 'dynamic' variable names, 'associative' arrays and whatnot but don't seem to be able to figure this out. Any help much appreciated.

Answer 1

Not sure I understood what you are trying to do but this is an example that you can, maybe, start with:

#!/bin/bash
list1=(a b c)
list2=(j k l)
for x in ${list1[@]}; do
    for y in ${list2[@]}; do
        tmplist+=(${x}${y})
    done
    cmd="declare -a list_${x}=(${tmplist[@]})"
    eval $cmd
    unset tmplist
done
echo "list_a: ${list_a[@]}"
echo "list_b: ${list_b[@]}"
echo "list_c: ${list_c[@]}"

It shows how to create arrays which names are computed from other variables, how to unset a temporary array...

Answer 2

I would create two types of arrays. One for containing disk names, then different arrays for each disk for storing partition specific information.

#!/usr/bin/env bash                                                                                                                                   
while read -r l; do
    t="${l#* }"
    if [[ $t == "disk" ]]; then
        # disks list contains all the disk names
        disks+="${l% *} "
    elif [[ $t == "part" ]]; then
        # for each disk 'XXX', a separate partitions array 'parts_XXX' is created
        [[ $l =~ [^a-z0-9]*([a-z]*)([0-9]*)\  ]] && d="${BASH_REMATCH[1]}" && p="${BASH_REMATCH[2]}"
        eval parts_$d+=\"my-value=\$d\$p \" # append to partitians array
    else
        echo "unknown type $t in $l"
        exit 1
    fi  
done <<< $(lsblk -n --output=NAME,TYPE | tr -s ' ')

# arrays are created. now iterate them
for i in ${disks[@]}; do
    echo "iterating partitions of disk: $i"
    # following is the name of the current disks partitions array
    var=parts_$i[@]
    # iterate partitians array of the current disk
    for j in ${!var}; do
        echo ">> $j"
    done
done

Answer 3

As a best-practices example (for bash 4.3 or newer):

#!/bin/bash

for blockdev in /sys/block/*; do
  devname=${blockdev##*/}               # remove leading path elements
  devname=${devname//-/_}               # rename dashes in device name to underscores
  declare -a "partitions_${devname}=()" # define an empty array
  declare -n _current_partitions="partitions_$devname"        # define a nameref
  for part in "$blockdev"/*/dev; do   # iterate over partitions
    [[ -e $part ]] || continue                                # skip if no matches
    part=${part%/dev}                                         # strip trailing /dev
    _current_partitions+=( "${part##*/}" )                    # add match via nameref
  done
  unset -n _current_partitions          # clear the nameref
  declare -p "partitions_$devname"      # print our resulting array
done

For my local test VM, this emits:

declare -a partitions_dm_0=()
declare -a partitions_dm_1=()
declare -a partitions_sda=([0]="sda1" [1]="sda2")
declare -a partitions_sdb=()
declare -a partitions_sr0=()

...which is correct, as sda is the only partitioned device.

---

The basic mechanism here is the nameref: declare -n name1=name2 will allow one to refer to the variable name2 under the name name1, including updates or assignments, until unset -n name1 is performed.