Say I have the file foo.txt
"The" "quick brown" "fox" "jumps over" "the" "lazy dog."
I would like to read these "fields" from the file into an array. However my attempt is failing if the field has a space
$ read -a bar < foo.txt
$ echo ${bar[0]}
"The"
$ echo ${bar[1]}
"quick
Use declare instead of eval:
declare -a "bar=( $( < foo.txt ) )"
This forces everything in the file to be treated as the right-hand side of the assignment. Using eval, the contents of the file can be interpreted as code. For example, if the file contains
Some text ); echo "you just erased all your files"; ( true
then the following is executed by eval:
bar=( Some text ); echo "you just erased all your files"; ( true )
The parentheses in the file balance the parentheses used outside the command substitution, resulting in the text between them in the file being executed as code rather than being used to populate the array.
echo ${bar[0]} : ( "The" "quick brown" "fox" "jumps over" "the" "lazy dog." ). check my solution
This works
$ read < foo.txt
$ eval bar=($REPLY)
$ echo ${bar[0]}
The
$ echo ${bar[1]}
quick brown
declare "bar=( $(< tmp.txt))" is a little safer; try the eval if the contents of the file are something like "The" ); echo malicious code; ( true.Your solution fails because bash normally uses whitespaces for input field separation IFS (and that's also why you get the quotes).
You can use plain read, if you prepare your text for processing:
IFS='"'
read -a bar < <(sed 's/"\([^"]*\)" */\1"/g' foo.txt)
echo ${bar[1]}
quick brown
Note: the sed command transforms the file into this format: The"quick brown"fox"jumps over"the"lazy dog.". I use " as delimiter as that's definitely not used in the words.
Here's one way using GNU awk and the FPAT variable:
IFS=$'\n'
array=($(awk 'BEGIN { FPAT = "([^ ]+)|(\"[^\"]+\")" } { for (i=1; i<=NF; i++) print $i }' file.txt))
echo "${array[1]}"
Result:
"quick brown"
IFS=$'\n' array=( `whatever_command_that_emits_lines_of_text` ) many times.