How can I join elements of an array in Bash?

Viewed 313591

If I have an array like this in Bash:

FOO=( a b c )

How do I join the elements with commas? For example, producing a,b,c.

32 Answers

Yet another solution:

#!/bin/bash
foo=('foo bar' 'foo baz' 'bar baz')
bar=$(printf ",%s" "${foo[@]}")
bar=${bar:1}

echo $bar

Edit: same but for multi-character variable length separator:

#!/bin/bash
separator=")|(" # e.g. constructing regex, pray it does not contain %s
foo=('foo bar' 'foo baz' 'bar baz')
regex="$( printf "${separator}%s" "${foo[@]}" )"
regex="${regex:${#separator}}" # remove leading separator
echo "${regex}"
# Prints: foo bar)|(foo baz)|(bar baz

This isn't all too different from existing solutions, but it avoids using a separate function, doesn't modify IFS in the parent shell and is all in a single line:

arr=(a b c)
printf '%s\n' "$(IFS=,; printf '%s' "${arr[*]}")"

resulting in

a,b,c

Limitation: the separator can't be longer than one character.


This could be simplified to just

(IFS=,; printf '%s' "${arr[*]}")

at which point it's basically the same as Pascal's answer, but using printf instead of echo, and printing the result to stdout instead of assigning it to a variable.

x=${arr[*]// /,}

This is the shortest way to do it.

Example,

# ZSH:
arr=(1 "2 3" 4 5)
x=${"${arr[*]}"// /,}
echo $x  # output: 1,2,3,4,5

# ZSH/BASH:
arr=(1 "2 3" 4 5)
a=${arr[*]}
x=${a// /,}
echo $x  # output: 1,2,3,4,5

Here's a single liner that is a bit weird but works well for multi-character delimiters and supports any value (including containing spaces or anything):

ar=(abc "foo bar" 456)
delim=" | "
printf "%s\n$delim\n" "${ar[@]}" | head -n-1 | paste -sd ''

This would show in the console as

abc | foo bar | 456

Note: Notice how some solutions use printf with ${ar[*]} and some with ${ar[@]}?

The ones with @ use the printf feature that supports multiple arguments by repeating the format template.

The ones with * should not be used. They do not actually need printfand rely on manipulating the field separator and bash's word expansion. These would work just as well with echo, cat, etc. - these solutions likely use printf because the author doesn't really understand what they are doing...

Combine best of all worlds so far with following idea.

# join with separator
join_ws()  { local IFS=; local s="${*/#/$1}"; echo "${s#"$1$1$1"}"; }

This little masterpiece is

  • 100% pure bash ( parameter expansion with IFS temporarily unset, no external calls, no printf ... )
  • compact, complete and flawless ( works with single- and multi-character limiters, works with limiters containing white space, line breaks and other shell special characters, works with empty delimiter )
  • efficient ( no subshell, no array copy )
  • simple and stupid and, to a certain degree, beautiful and instructive as well

Examples:

$ join_ws , a b c
a,b,c
$ join_ws '' a b c
abc
$ join_ws $'\n' a b c
a
b
c
$ join_ws ' \/ ' A B C
A \/ B \/ C

Many, if not most, of these solutions rely on arcane syntax, brain-busting regex tricks, or calls to external executables. I would like to propose a simple, bash-only solution that is very easy to understand, and only slightly sub-optimal, performance-wise.

join_by () {
    # Argument #1 is the separator. It can be multi-character.
    # Argument #2, 3, and so on, are the elements to be joined.
    # Usage: join_by ", " "${array[@]}"
    local SEPARATOR="$1"
    shift

    local F=0
    for x in "$@"
    do
        if [[ F -eq 1 ]]
        then
            echo -n "$SEPARATOR"
        else
            F=1
        fi
        echo -n "$x"
    done
    echo
}

Example:

$ a=( 1 "2 2" 3 )
$ join_by ", " "${a[@]}"
1, 2 2, 3
$ 

I'd like to point out that any solution that uses /usr/bin/[ or /usr/bin/printf is inherently slower than my solution, since I use 100% pure bash. As an example of its performance, Here's a demo where I create an array with 1,000,000 random integers, then join them all with a comma, and time it.

$ eval $(echo -n "a=("; x=0 ; while [[ x -lt 1000000 ]]; do echo -n " $RANDOM" ; x=$((x+1)); done; echo " )")
$ time join_by , ${a[@]} >/dev/null
real    0m8.590s
user    0m8.591s
sys     0m0.000s
$ 

Using variable indirection to refer directly to an array also works. Named references can also be used, but they only became available in 4.3.

The advantage of using this form of a function is that you can have the separator optional (defaults to the first character of default IFS, which is a space; perhaps make it an empty string if you like), and it avoids expanding values twice (first when passed as parameters, and second as "$@" inside the function).

This solution also doesn't require the user to call the function inside a command substitution - which summons a subshell, to get a joined version of a string assigned to another variable.

function join_by_ref {
    __=
    local __r=$1[@] __s=${2-' '}
    printf -v __ "${__s//\%/%%}%s" "${!__r}"
    __=${__:${#__s}}
}

array=(1 2 3 4)

join_by_ref array
echo "$__" # Prints '1 2 3 4'.

join_by_ref array '%s'
echo "$__" # Prints '1%s2%s3%s4'.

join_by_ref 'invalid*' '%s' # Bash 4.4 shows "invalid*[@]: bad substitution".
echo "$__" # Prints nothing but newline.

Feel free to use a more comfortable name for the function.

This works from 3.1 to 5.0-alpha. As observed, variable indirection doesn't only work with variables but with other parameters as well.

A parameter is an entity that stores values. It can be a name, a number, or one of the special characters listed below under Special Parameters. A variable is a parameter denoted by a name.

Arrays and array elements are also parameters (entities that store value), and references to arrays are technically references to parameters as well. And much like the special parameter @, array[@] also makes a valid reference.

Altered or selective forms of expansion (like substring expansion) that deviate reference from the parameter itself no longer work.

Update

In the release version of Bash 5.0, variable indirection is already called indirect expansion and its behavior is already explicitly documented in the manual:

If the first character of parameter is an exclamation point (!), and parameter is not a nameref, it introduces a level of indirection. Bash uses the value formed by expanding the rest of parameter as the new parameter; this is then expanded and that value is used in the rest of the expansion, rather than the expansion of the original parameter. This is known as indirect expansion.

Taking note that in the documentation of ${parameter}, parameter is referred to as "a shell parameter as described (in) PARAMETERS or an array reference". And in the documentation of arrays, it is mentioned that "Any element of an array may be referenced using ${name[subscript]}". This makes __r[@] an array reference.

Join by arguments version

See my comment in Riccardo Galli's answer.

Perhaps late for the party, but this works for me:

function joinArray() {
  local delimiter="${1}"
  local output="${2}"
  for param in ${@:3}; do
    output="${output}${delimiter}${param}"
  done

  echo "${output}"
}

Here's one that most POSIX compatible shells support:

join_by() {
    # Usage:  join_by "||" a b c d
    local arg arr=() sep="$1"
    shift
    for arg in "$@"; do
        if [ 0 -lt "${#arr[@]}" ]; then
            arr+=("${sep}")
        fi
        arr+=("${arg}") || break
    done
    printf "%s" "${arr[@]}"
}

I believe this is the shortest solution, as Benamin W. already mentioned:

(IFS=,; printf %s "${a[*]}")

Wanted to add that if you use zsh, you can drop the subshell:

IFS=, printf %s "${a[*]}"

Test:

a=(1 'a b' 3)
IFS=, printf %s "${a[*]}"
1,a b,3
Related