Creating an array from a text file in Bash

Question

A script takes a URL, parses it for the required fields, and redirects its output to be saved in a file, file.txt. The output is saved on a new line each time a field has been found.

file.txt

A Cat
A Dog
A Mouse 
etc... 

I want to take file.txt and create an array from it in a new script, where every line gets to be its own string variable in the array. So far I have tried:

#!/bin/bash

filename=file.txt
declare -a myArray
myArray=(`cat "$filename"`)

for (( i = 0 ; i < 9 ; i++))
do
  echo "Element [$i]: ${myArray[$i]}"
done

When I run this script, whitespace results in words getting split and instead of getting

Desired output

Element [0]: A Cat 
Element [1]: A Dog 
etc... 

I end up getting this:

Actual output

Element [0]: A 
Element [1]: Cat 
Element [2]: A
Element [3]: Dog 
etc... 

How can I adjust the loop below such that the entire string on each line will correspond one-to-one with each variable in the array?

Solution 1

Use the mapfile command:

mapfile -t myArray < file.txt

The error is using for -- the idiomatic way to loop over lines of a file is:

while IFS= read -r line; do echo ">>$line<<"; done < file.txt

See BashFAQ/005 for more details.

Solution 2

mapfile and readarray (which are synonymous) are available in Bash version 4 and above. If you have an older version of Bash, you can use a loop to read the file into an array:

arr=()
while IFS= read -r line; do
  arr+=("$line")
done < file

In case the file has an incomplete (missing newline) last line, you could use this alternative:

arr=()
while IFS= read -r line || [[ "$line" ]]; do
  arr+=("$line")
done < file

---

Related:

• Need alternative to readarray/mapfile for script on older version of Bash

Solution 3

You can do this too:

oldIFS="$IFS"
IFS=$'\n' arr=($(<file))
IFS="$oldIFS"
echo "${arr[1]}" # It will print `A Dog`.

Note:

Filename expansion still occurs. For example, if there's a line with a literal * it will expand to all the files in current folder. So use it only if your file is free of this kind of scenario.

Solution 4

Use mapfile or read -a

Always check your code using shellcheck. It will often give you the correct answer. In this case SC2207 covers reading a file that either has space separated or newline separated values into an array.

Don't do this

array=( $(mycommand) )

Files with values separated by newlines

mapfile -t array < <(mycommand)

Files with values separated by spaces

IFS=" " read -r -a array <<< "$(mycommand)"

The shellcheck page will give you the rationale why this is considered best practice.

Solution 5

You can simply read each line from the file and assign it to an array.

#!/bin/bash
i=0
while read line 
do
        arr[$i]="$line"
        i=$((i+1))
done < file.txt

View more solutions

Comments

[Hugues]

Is there any way to set IFS only temporarily (so that it recovers its original value after this command), while still persisting the assignment to arr?

[Hugues]

Note that filename expansion still occurs; e.g. IFS=$'\n' arr=($(echo 'a 1'; echo '*'; echo 'b 2')); printf "%s\n" "${arr[@]}"

[Jahid]

@Hugues : yap, filename expansion still occurs. I will add that bit of info..thnks..

[Hugues]

Sorry, I disagree. IFS=... command does not change IFS in the current shell. However, IFS=... other_variable=... (without any command) does change both IFS and other_variable in the current shell.

[Jahid]

@Hugues : You are right again, sorry about that... Fixed it with save-and-reset way.

[Hugues]

Thanks! This works; it's unfortunate that there is no simpler way as I like the arr= notation (compared to mapfile/readarray).

[fedorqui]

Since this is being promoted as the canonical q&a, you could also include what is mentioned in the link: while IFS= read -r; do lines+=("$REPLY"); done <file.

[ericslaw]

mapfile does not exist in bash versions prior to 4.x

[glenn jackman]

Bash 4 is about 5 years old now. Upgrade.

[hola]

How do you access the array?

[De Novo]

Despite bash 4 being released in 2009, @ericslaw's comment remains relevant because many machines still ship with bash 3.x (and will not upgrade, so long as bash is released under GPLv3). If you're interested in portability, it's an important thing to note

[glenn jackman]

Sure, an OS may ship with older bash, but individuals can upgrade their own installations, or install bash 4 separately (with homebrew or the like).

[ericslaw]

macos mojave bash reports as 3.2.57 (wow really apple?). I suspect work wont let me touch too many things on the laptop though :(

[glenn jackman]

You should be able to use whatever tools you need to do your job.

[De Novo]

the issue isn't that a developer can't install an upgraded version, it's that a developer should be aware that a script using mapfile will not run as expected on many machines without additional steps. @ericslaw macs will continue to ship with bash 3.2.57 for the foreseeable future. More recent versions use a license that would require apple to share or allow things they don't want to share or allow.

[Tatiana Racheva]

I find it that I have to put parentheses around IFS= read -r line || [[ "$line" ]] for it to work. Otherwise, it works great!

[codeforester]

@TatianaRacheva: isn't it that the semicolon that was missing before do?