printf "%.3f" ${variable with newlines} - error with \n

Question

Like HERE I have a file.csv with numbers in quotes:

"0.2"
"0.3339"
"0.111111"

To round the number (3 decimals) this solutions works great:

printf "%.03f\n" $(sed 's/\"//g' file.csv)

But now I want to store sed 's/\"//g' file.csv as a variable

var_sed=$(sed 's/\"//g' file.csv);
printf "%.03f\n" ${var_sed}

Doesn't work. The output is

zsh: bad math expression: operator expected at `0.3339\n0.1...'
0.000

So the problem is that var_sed passes \n to printf "%.03f\n".

The only solution I know is:

var_sed=$(sed 's/\"//g' file.csv);
printf "%.03f\n" $(echo ${var_sed})

1. Maybe there is a cleaner way.
2. Also I want to put printf "%.03f\n" in a function like this:

printf_function () { printf "%.03f\n" $1;}

But this is not working:

printf_function () { printf "%.03f\n" $1;};
var_sed=$(sed 's/\"//g' file.csv);
printf_function ${var_sed}

also printf_function $(echo ${var_sed}) is not working.

____________________________________________________________

Why I try to save the file in a variable? ____________________________________________________________

The truth is that I actually want to put the sed command in a function. I'm sorry I put this as a variable to (try to) simplify the problem.

My script is

sed_01 () { sed 's/\"$// ; s/^\"// ; s/something_else//g' $1;};
printf_03 () { printf "%.03f\n" $1;};
printf_03 "$(sed_01 file.csv)"

as mentioned in the comments below, it works in bash.

output:

"0.200"
"0.334"
"0.111"

but in zsh the output: is:

printf_03: bad math expression: operator expected at `0.3339\n0.1...'
0.000

Answer 1

printf "%.03f\n" $(sed 's/\"//g' file.csv)

var_sed=$(sed 's/\"//g' file.csv);
printf "%.03f\n" ${var_sed}

This relies on word splitting to give the numbers to printf as separate arguments. The thing is, that zsh has dropped automatic word splitting as a silly remnant of the past, and doesn't do it for variable expansions. But it does do it for command substitutions, for whatever reason.

There's also no splitting when assigning from the command substitution to a regular scalar variable, so the newlines in the output of sed get saved in the variable.

That's why your code works with the command substitution, but not with the intermediate variable.

If you do need to load the values in a variable first (and you don't for just printing), consider storing them in an array (zsh):

lines=("${(f)$(sed 's/\"//g' file.csv)}")
printf "%.03f\n" $lines

in Bash, you'd use readarray/mapfile:

mapfile -t lines < <(sed 's/\"//g' file.csv)
printf "%.03f\n" "${lines[@]}"

Or if you want something that mostly works in either, you'll have to rely on word splitting again:

lines=( $(sed 's/\"//g' file.csv) )
printf "%.03f\n" "${lines[@]}"

Note the parentheses, they make that an array assignment, so word splitting again happens. _{("mostly" including the usual caveats of a modified IFS affecting the behaviour, and filename globbing also happening.)}

---

Then again, something like this might be easier done with just AWK:

$ awk '{gsub(/"/, ""); printf "%.03f\n", $1}' file.csv
0.200
0.334
0.111