Printing command by reading CSV file with AWK into a bash array

Question

I have the following CSV file (mycsv.csv):

love,hate,--command yes this is bad --other-command
fox,duck,--command yes this is bad --other-command
turtle,rabbit

This means the command would be --command "command_argument" --other-command

Then I'm creating an array:

my_csv="/home/sometimesihatebash/mycsv.csv"
first_column=( $(awk -F "\"*,\"*" '{print $1}' $my_csv) )
second_column=( $(awk -F "\"*,\"*" '{print $2}' $my_csv) )
IFS=$'\n' command=( $(awk -F "\"*,\"*" '{print $3}' $my_csv) )

Then printing all columns into a command

for i in "${!first_column[@]}"; do
    echo "${first_column[i]}"
    echo "${second_column[i]}"
    echo "${command[i]}"
    
    execute_program_name --first "${first_column[i]}" --second "${second_column[i]}" "${command[i]}"

done

But then, for the third row (which has an empty third column), I get:

error: 1 unrecognized argument:
'

And for the first two rows, I get:

no such option: --command yes this is bad --other-command

When echoing each array element, I don't see any problems. Pasting the exact output of echo also runs the program normally.

Running the for loop with just the first_column and second_column arrays works fine

Probably this has something to do with special characters in the last array but I`m lost

Any ideas?

Answer 1

Assumptions:

• the input string --command yes this is bad --other-command needs to be passed to execute_program_name in the following format: --command 'yes this is bad' --other-command
• the 3rd column consists of long form options (ie, options always start with --)
• if the 3rd column is populated the first string will always be a long form option (ie, string will always have the format --option)
• the contents of mycsv.csv only need to be parsed for this one for loop; net result: we'll only need a single array (args[]); (see 2nd half of answer for an approach that allows for a single parse and multiple re-use)

We'll make some changes to OP's inputs to demonstrate handling options with and without arguments:

$ cat mycsv.csv
love,hate,--command yes this is bad --other-command
fox,duck,--command yes this is bad --other-command single_string
turtle,rabbit 
sweet,sour,--new-command --newer-command

A simple script to list input parameters:

$ cat execute_program_name
#!/bin/bash

echo "### $0 $@"
cnt=0

for arg in "$@"
do
    ((cnt++))
    echo "arg #${cnt} : ${arg}"
done

One bash approach

$ cat parse_n_run
#!/bin/bash

while IFS=, read -r c1 c2 c3
do
    printf "\n########### %s : %s : %s\n\n" "${c1}" "${c2}" "${c3}"

    args=( "--first" "${c1}" "--second" "${c2}" )       # store first 2 sets of options in array

    while read -r option string
    do
        [[ -z "${option}" ]] && continue                # skip first line (blank)
        args+=( "${option}" )                           # add "--option" to array
        [[ -n "${string}" ]] && args+=( "${string}" )   # if there's an argument then add to the array

    done <<< "${c3//--/$'\n'&}"                         # split c3 into lines by placing a "\n" before each "--"

    typeset -p args                                     # display contents of array
    echo ""
    execute_program_name "${args[@]}"
done < mycsv.csv

Taking for a test drive:

$ parse_n_run

########### love : hate : --command yes this is bad --other-command

declare -a args=([0]="--first" [1]="love" [2]="--second" [3]="hate" [4]="--command" [5]="yes this is bad" [6]="--other-command")

### ./execute_program_name --first love --second hate --command yes this is bad --other-command
arg #1 : --first
arg #2 : love
arg #3 : --second
arg #4 : hate
arg #5 : --command
arg #6 : yes this is bad
arg #7 : --other-command

########### fox : duck : --command yes this is bad --other-command single_string

declare -a args=([0]="--first" [1]="fox" [2]="--second" [3]="duck" [4]="--command" [5]="yes this is bad" [6]="--other-command" [7]="single_string")

### ./execute_program_name --first fox --second duck --command yes this is bad --other-command single_string
arg #1 : --first
arg #2 : fox
arg #3 : --second
arg #4 : duck
arg #5 : --command
arg #6 : yes this is bad
arg #7 : --other-command
arg #8 : single_string

########### turtle : rabbit  :

declare -a args=([0]="--first" [1]="turtle" [2]="--second" [3]="rabbit ")

### ./execute_program_name --first turtle --second rabbit
arg #1 : --first
arg #2 : turtle
arg #3 : --second
arg #4 : rabbit

########### sweet : sour : --new-command --newer-command

declare -a args=([0]="--first" [1]="sweet" [2]="--second" [3]="sour" [4]="--new-command" [5]="--newer-command")

### ./execute_program_name --first sweet --second sour --new-command --newer-command
arg #1 : --first
arg #2 : sweet
arg #3 : --second
arg #4 : sour
arg #5 : --new-command
arg #6 : --newer-command

---

If OP needs to parse mycsv.csv and (re)use multiple times, and OP has access to bash 4.3+, I'd probably opt to use a nameref to manage arrays args_1[], args_2[], ..., args_n[].

Making a few tweaks to the current script:

$ cat parse_n_run
#!/bin/bash

cnt=0

##########
# parse and store

echo "Parsing ..."

while IFS=, read -r c1 c2 c3
do
    printf "\n########### %s : %s : %s\n\n" "${c1}" "${c2}" "${c3}"

    ((cnt++))
    declare -n _args="args_${cnt}"                          # name ref

    _args=( "--first" "${c1}" "--second" "${c2}" )

    while read -r option string
    do
        [[ -z "${option}" ]] && continue
        _args+=( "${option}" )
        [[ -n "${string}" ]] && _args+=( "${string}" )
    done <<< "${c3//--/$'\n'&}"

    typeset -p "${!_args}"
done < mycsv.csv

### at this point we have 4 arrays: args_1[], args_2[], args_3[] and args_4[]

##########
# make use of arrays

printf "\n###########\n"
echo "Using ..."

for ((i=1; i<=cnt; i++))
do
    declare -n _args="args_$i"

    printf "\n### %s\n\n" "${!_args}"

    execute_program_name "${_args[@]}"
done

Taking for a test drive:

NOTE: notice the 4 declare -a lines reference our 4 arrays args_1[], args_2[], args_3[] and args_4[]

$ parse_n_run
Parsing ...

########### love : hate : --command yes this is bad --other-command

declare -a args_1=([0]="--first" [1]="love" [2]="--second" [3]="hate" [4]="--command" [5]="yes this is bad" [6]="--other-command")

########### fox : duck : --command yes this is bad --other-command single_string

declare -a args_2=([0]="--first" [1]="fox" [2]="--second" [3]="duck" [4]="--command" [5]="yes this is bad" [6]="--other-command" [7]="single_string")

########### turtle : rabbit  :

declare -a args_3=([0]="--first" [1]="turtle" [2]="--second" [3]="rabbit ")

########### sweet : sour : --new-command --newer-command

declare -a args_4=([0]="--first" [1]="sweet" [2]="--second" [3]="sour" [4]="--new-command" [5]="--newer-command")

###########
Using ...

### args_1

### execute_program_name --first love --second hate --command yes this is bad --other-command
arg #1 : --first
arg #2 : love
arg #3 : --second
arg #4 : hate
arg #5 : --command
arg #6 : yes this is bad
arg #7 : --other-command

### args_2

### execute_program_name --first fox --second duck --command yes this is bad --other-command single_string
arg #1 : --first
arg #2 : fox
arg #3 : --second
arg #4 : duck
arg #5 : --command
arg #6 : yes this is bad
arg #7 : --other-command
arg #8 : single_string

### args_3

### execute_program_name --first turtle --second rabbit
arg #1 : --first
arg #2 : turtle
arg #3 : --second
arg #4 : rabbit

### args_4

### execute_program_name --first sweet --second sour --new-command --newer-command
arg #1 : --first
arg #2 : sweet
arg #3 : --second
arg #4 : sour
arg #5 : --new-command
arg #6 : --newer-command