Bash CSV Parser

This repo provides a bash library called csv.bash which parses lines of comma-delimited-values into an array.

Here's example/parse-stdin.bash:

#!/bin/bash
source csv.bash                   # load the library
while IFS= read -r CSV_ROW;do     # read stdin line by line into CSV_ROW
  csv_row_to_cols                 # parse CSV_ROW into CSV_COLS
  printf "%s\n" "${CSV_COLS[@]}"  # do something with CSV_COLS!
done

It reads a stream of csv input and prints each parsed column on a new line. csv_row_to_cols parses the global variable CSV_ROW and populates the results in CSV_COLS (returning values via globals is necessary in bash).

echo 'foo,"bar,baz",,"boo\"",' | example/parse-stdin.bash
foo
"bar,baz"
""
"boo\""
""

This example shows the default behavior, but that can be changed by setting these global variables:

CSV_TER=""    # terminator for rows
CSV_DEL=","   # delimiter for columns
CSV_QUO='"'   # quote char - quoted delimiters are ignored
CSV_ESC="\\"  # escape char for quote char
CSV_CON=1     # consecutive delimiters declare empty columns
CSV_EMP='""'  # empty column replacement value

For example you can set CSV_DEL to tab to parse tab-delimited data. Less obviously, you could use csv.bash to read command arguments from input and split them into words, instead of using xargs, which is slow.

#/bin/bash
source csv.bash
CSV_DEL=" "
CSV_ROW='foo "bar baz"' # bash will wordsplit like this (foo,"bar,baz")
csv_row_to_cols         # CSV_COLS=(foo,"bar baz")
...

All global variables are prepended with CSV_ to avoid name clashes with other code. And that's not all the global variables either! CSV_COLC is set to the number of columns parsed, which is handy to use as the length of $CSV_COLS which bash may not report correctly if it contains empty strings. CSV_ROWC is an incrementing row counter, which you may want to reset to 0 if your code is parsing different inputs in the same process.