I have an array, and I would like to repeat each element except the first and last elements.

For example if the array has five elements 1 2 3 4 5, then after repeating, its elements should be 1 2 2 3 3 4 4 5.

My bash commands are

$ newarr=("${myarr[0]}")
$ for i in $(seq 1 $((${#myarr[@]}-2))) ; do newarr+=( "${myarr[i]}" "${myarr[i]}"); done
$ newarr+=("${myarr[-1]}")

Is there a more clear way than mine?

I am also wondering about how to wrap that into a function which takes myarr as argument, and returns newarr. (After creating such a function, I will read an array from a file, so that each element stores a line in the file, and then call the function on the array. If creating such a function is not a good approach, let me know.)



You can get the number of operation down ever so slightly, and skip the call to seq:

for (( i = 1; i < ${#myarr[@]} - 1; ++i )); do
    newarr+=( "${myarr[i]}" "${myarr[i]}" )
newarr=( "${myarr[0]}" "${newarr[@]}" "${myarr[-1]}" )

This assumes that newarr is empty to start with. Do unset newarry first if it's not.

As a function (this modifies the array that is passed):

dup_interal_items () {
    typeset -n arr=$1
    local tmparr

    for (( i = 1; i < ${#arr[@]} - 1; ++i )); do
        tmparr+=( "${arr[i]}" "${arr[i]}" )
    arr=( "${arr[0]}" "${tmparr[@]}" "${arr[-1]}" )

The name of the array is passed into the function and the name-reference variable arr is used to access the elements in the array. At the end, the original array is updated to contain the result.

I did it this way rather than returning an array as you can only return an exit status from a function. The other approach would have been to pass both the names of an input array and an output array and use two name-reference variables in the function.

Or, you could possibly echo or printf the values in the function (in which case you don't have to construct a new array at all) and then parse that data in the main code. The function would then be required to be called inside a command substitution.

Note that you can't call this function with an array called arr due to the particular name scoping rules used by bash. You may want to rename the arr variable in the function if this is an issue.


$ myarr=( 1 2 3 "here we go" )
$ dup_interal_items myarr
$ printf 'Element: %s\n' "${myarr[@]}"
Element: 1
Element: 2
Element: 2
Element: 3
Element: 3
Element: here we go

To duplicate all lines of a file except for the first and last line:

sed -e '1b' -e '$b' -e 'p' <file

The sed script branches to the end (where there is an implicit print statement) if it's on the first or last line, but prints all other lines (all other lines are therefore both explicitly printed by that last p and implicitly printed).

Or, as contributed by don_crissti,

sed 'p;1d;$d' <file

which explicitly prints each line, then ends the cycle for the first and last line, but prints all other lines implicitly (a second time).

An equivalent awk program that does not store more than a single line in memory would be non-trivial to write.

  • Thanks. (1) "you can't call this function with an array called arr due to the particular name scoping rules used by bash." Do you mean arr declared as typeset -n arr=$1 inside the function body has global (script-wide) scope? (2) Among the three approaches when creating a bash function, which one do you suggest more or none? – Tim Nov 17 '18 at 21:08
  • @Tim (1) A name reference variable can't reference another variable that has the same name as itself (see also Circular name references in bash shell function, but not in ksh). (2) It depends on what you'd like to do and how you want/need to get there. If the original data was coming from a file, and wasn't used for anything else, I would probably go with reading the array from sed instead of reading it and manipulating it. If the resulting data would go to a file, I would not read it into an array at all. – Kusalananda Nov 17 '18 at 21:13
  • Thanks. "If the resulting data would go to a file, I would not read it into an array at all." Could you be more specific? – Tim Nov 17 '18 at 22:39
  • @Tim Well, if all you want to do is to duplicate the internal lines of a text file and write that to a new file, then there is that sed command that could do that for you. Why read it into an array in bash? – Kusalananda Nov 17 '18 at 22:41
  • 1
    @Cbhihe No, you are misreading. It says "Array variables cannot be given the nameref attribute. However, nameref variables can reference array variables and subscripted array variables." In my code, arr (in the function) is not an array variable, it's an ordinary variable that I give the nameref attribute to. It then references an array variable. – Kusalananda Nov 18 '18 at 17:55

Try also - no function needed -

$ ARR=(1 2 3 4 5)
$ IFS=$'\n\t '
$ readarray NEWARR < <(echo "${ARR[*]}" | sed '1!p; $d')
$ echo ${NEWARR[@]}
1 2 2 3 3 4 4 5

or even

NEWARR=($(printf "%s\n" ${ARR[@]} | sed '1!p; $d'))

A solution that accepts newlines:

arr=( 1 2 "3 3" $'41\n42' 5 )

readarray -t -d $'\0' newarray < <(printf '%s\0' "${arr[@]}" | sed -z '1!p; $d')

printf '<%s>\n' "${newarr[@]}"

Runs as:

$ ./script
<3 3>
<3 3>

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.