Question

I am looking for a way to split a string in bash over a delimiter string, and place the parts in an array.

Simple case:

#!/bin/bash
b="aaaaa/bbbbb/ddd/ffffff"
echo "simple string: $b"

IFS='/' b_split=($b)
echo ;
echo "split"
for i in ${b_split[@]}
do
    echo "------ new part ------"
    echo "$i"
done

Gives output

simple string: aaaaa/bbbbb/ddd/ffffff

split
------ new part ------
aaaaa
------ new part ------
bbbbb
------ new part ------
ddd
------ new part ------
ffffff

More complex case:

#!/bin/bash
c=$(echo "AA=A"; echo "B=BB"; echo "======="; echo "C==CC"; echo "DD=D"; echo "======="; echo "EEE"; echo "FF";)
echo "more complex string"
echo "$c";
echo ;
echo "split";

IFS='=======' c_split=($c) ;#    <----    LINE TO BE CHANGED 

for i in ${c_split[@]}
do
    echo "------ new part ------"
    echo "$i"
done

Gives output:

more complex string
AA=A
B=BB
=======
C==CC
DD=D
=======
EEE
FF

split
------ new part ------
AA
------ new part ------
A
B
------ new part ------
BB

------ new part ------

------ new part ------

------ new part ------

------ new part ------

------ new part ------

------ new part ------

------ new part ------

C
------ new part ------

------ new part ------
CC
DD
------ new part ------
D

------ new part ------

------ new part ------

------ new part ------

------ new part ------

------ new part ------

------ new part ------

------ new part ------

EEE
FF

I would like the second output to be like

------ new part ------
AA=A
B=BB
------ new part ------
C==CC
DD=D
------ new part ------
EEE
FF

I.e. to split the string on a sequence of characters, instead of one. How can I do this?

I am looking for an answer that would only modify this line in the second script:

IFS='=======' c_split=($c) ;#    <----    LINE TO BE CHANGED 

Answer 1

Introduction

At bottom of this, you will find a function to transform string to an array with following syntax:

ssplit "<string>" "<array name>" "<delimiter string>"

For this:

ssplit "$c" c_split $'\n=======\n'
declare -p c_split 
declare -a c_split=([0]=$'AA=A\nB=BB' [1]=$'C==CC\nDD=D' [2]=$'EEE\nFF')

IFS disambiguation

IFS mean Input Field Separators, as list of characters that could be used as separators.

By default, this is set to \t\n, meaning that any number (greater than zero) of space, tabulation and/or newline could be one separator.

So with the string: $' blah foo=bar \nbaz '

 read -a c_split <<<"    blah  foo=bar 
 baz  "
 declare -p c_split 
 declare -a c_split=([0]="blah" [1]="foo=bar")

Leading and trailing separators would be ignored and this string will contain only 3
parts: blah, foo=bar and baz.

But except for spaces, IFS consider each separator for itself:

IFS=Z read a b c d e f <<<ZaZZbZcZZdZeZf
declare -p a b c d e f
declare -- a=""
declare -- b="a"
declare -- c=""
declare -- d="b"
declare -- e="c"
declare -- f="ZdZeZf"

Splitting a string using IFS is possible if you know a valid field separator not used in your string, so you could replace your pattern by this character (by using ${var//<pattern>/<separator>} syntax):

OIFS="$IFS"
IFS='§'
c=$'AA=A\nB=BB\n=======\nC==CC\nDD=D\n=======\nEEE\nFF'
c_split=(${c//=======/§})
IFS="$OIFS"
printf -- "------ new part ------\n%s\n" "${c_split[@]}"

------ new part ------
AA=A
B=BB

------ new part ------

C==CC
DD=D

------ new part ------

EEE
FF

But this work only while string do not contain any §.

You could use another character, like IFS=$'\026';c_split=(${c//=======/$'\026'}) but anyway this may involve furter bugs.

You could browse character maps for finding one who’s not in your string:

myIfs=""
for i in {1..255};do
    printf -v char "$(printf "\\\%03o" $i)"
        [ "$c" == "${c#*$char}" ] && myIfs="$char" && break
  done
if ! [ "$myIFS" ] ;then
    echo no split char found, could not do the job, sorry.
    exit 1
  fi

but I find this solution a little overkill.

Splitting on spaces (or without modifying IFS)

Under bash, we could use this bashism:

b="aaaaa/bbbbb/ddd/ffffff"
b_split=(${b//// })

In fact, this syntaxe ${varname// will initiate a translation (delimited by /) replacing all occurences of / by a space , before assigning it to an array b_split.

Of course, this still use IFS and split array on spaces.

This is not the best way, but could work with specific cases.

You could even drop unwanted spaces before splitting:

b='12 34 / 1 3 5 7 / ab'
b1=${b// }
b_split=(${b1//// })
printf "<%s>, " "${b_split[@]}" ;echo
<12>, <34>, <1>, <3>, <5>, <7>, <ab>, 

or exchange thems…

b1=${b// /§}
b_split=(${b1//// })
printf "<%s>, " "${b_split[@]//§/ }" ;echo
<12 34 >, < 1 3 5 7 >, < ab>, 

Splitting line on delimiter strings:

So you have to not use IFS for your meaning, but bash do have nice features:

#!/bin/bash

c=$'AA=A\nB=BB\n=======\nC==CC\nDD=D\n=======\nEEE\nFF'
echo "more complex string"
echo "$c";
echo ;
echo "split";

mySep='======='
while [ "$c" != "${c#*$mySep}" ];do
    echo "------ new part ------"
    echo "${c%%$mySep*}"
    c="${c#*$mySep}"
  done
echo "------ last part ------"
echo "$c"

Let see:

more complex string
AA=A
B=BB
=======
C==CC
DD=D
=======
EEE
FF

split
------ new part ------
AA=A
B=BB

------ new part ------

C==CC
DD=D

------ last part ------

EEE
FF

About Leading newline

Leading and trailing newlines are not deleted in previous samples. For this, you could simply:

mySep=$'\n=======\n'

instead of =======.

Or you could rewrite split loop for keeping explicitely this out:

mySep=$'======='
while [ "$c" != "${c#*$mySep}" ];do
    echo "------ new part ------"
    part="${c%%$mySep*}"
    part="${part##$'\n'}"
    echo "${part%%$'\n'}"
    c="${c#*$mySep}"
  done
echo "------ last part ------"
c=${c##$'\n'}
echo "${c%%$'\n'}"

Any case, this match what SO question asked for (: and his sample 🙂

------ new part ------
AA=A
B=BB
------ new part ------
C==CC
DD=D
------ last part ------
EEE
FF

Finaly creating an array.

#!/bin/bash
c=$'AA=A\nB=BB\n=======\nC==CC\nDD=D\n=======\nEEE\nFF'
echo "more complex string"
echo "$c";
echo ;
echo "split";

mySep=$'======='
export -a c_split
while [ "$c" != "${c#*$mySep}" ];do
    part="${c%%$mySep*}"
    part="${part##$'\n'}"
    c_split+=("${part%%$'\n'}")
    c="${c#*$mySep}"
  done
c=${c##$'\n'}
c_split+=("${c%%$'\n'}")

for i in "${c_split[@]}"
do
    echo "------ new part ------"
    echo "$i"
done

Do this finely:

more complex string
AA=A
B=BB
=======
C==CC
DD=D
=======
EEE
FF

split
------ new part ------
AA=A
B=BB
------ new part ------
C==CC
DD=D
------ new part ------
EEE
FF

Some explanations:

• export -a var to define var as an array and share them in childs
• ${variablename%string*}, ${variablename%%string*} result in the left part of variablename, upto but without string. One % mean last occurence of string and %% for all occurences. Full variablename is returned is string not found.
• ${variablename#*string}, do same in reverse way: return last part of variablename from but without string. One # mean first occurence and two ## man all occurences.

Nota in replacement, character * is a joker mean any number of any character.

The command echo "${c%%$'\n'}" would echo variable c but without any number of newline at end of string.

So if variable contain Hello WorldZorGluBHello youZorGluBI'm happy,

variable="Hello WorldZorGluBHello youZorGluBI'm happy"

$ echo ${variable#*ZorGluB}
Hello youZorGlubI'm happy

$ echo ${variable##*ZorGluB}
I'm happy

$ echo ${variable%ZorGluB*}
Hello WorldZorGluBHello you

$ echo ${variable%%ZorGluB*}
Hello World

$ echo ${variable%%ZorGluB}
Hello WorldZorGluBHello youZorGluBI'm happy

$ echo ${variable%happy}
Hello WorldZorGluBHello youZorGluBI'm

$ echo ${variable##* }
happy

All this is explained in the manpage:

$ man -Len -Pless\ +/##word bash

$ man -Len -Pless\ +/%%word bash

$ man -Len -Pless\ +/^\\\ *export\\\ .*word bash

Step by step, the splitting loop:

The separator:

mySep=$'======='

Declaring c_split as an array (and could be shared with childs)

export -a c_split

While variable c do contain at least one occurence of mySep

while [ "$c" != "${c#*$mySep}" ];do

Trunc c from first mySep to end of string and assign to part.

    part="${c%%$mySep*}"

Remove leading newlines

    part="${part##$'\n'}"

Remove trailing newlines and add result as a new array element to c_split.

    c_split+=("${part%%$'\n'}")

Reassing c whith the rest of string when left upto mySep is removed

    c="${c#*$mySep}"

Done 😉

done

Remove leading newlines

c=${c##$'\n'}

Remove trailing newlines and add result as a new array element to c_split.

c_split+=("${c%%$'\n'}")

Into a function:

ssplit() {
    local string="$1" array=${2:-ssplited_array} delim="${3:- }" pos=0
    while [ "$string" != "${string#*$delim}" ];do
        printf -v $array[pos++] "%s" "${string%%$delim*}"
        string="${string#*$delim}"
      done
    printf -v $array[pos] "%s" "$string"
}

Usage:

ssplit "<quoted string>" [array name] [delimiter string]

where array name is $splitted_array by default and delimiter is one single space.

You could use:

c=$'AA=A\nB=BB\n=======\nC==CC\nDD=D\n=======\nEEE\nFF'
ssplit "$c" c_split $'\n=======\n'
printf -- "--- part ----\n%s\n" "${c_split[@]}"
--- part ----
AA=A
B=BB
--- part ----
C==CC
DD=D
--- part ----
EEE
FF