csvmod

A stream-oriented CSV modification tool. Like a stripped-down "sed" command, but for tabular data.

TL;DR

Install:

$ pip install csvsed

Use:

# given a sample CSV
$ cat sample.csv

Employee ID,Age,Wage,Status
8783,47,"104,343,873.83","All good, but nowhere to go."
2003,32,"98,878,784.00",A-OK

# modify that data with a series of `csvsed` pipes
$ cat sample.csv \
  | csvsed -c Wage s/,//g \                              # remove commas from the Wage column
  | csvsed -c Status 'y/A-Z/a-z/' \                      # convert Status to all lowercase
  | csvsed -c Status 's/.*(ok|good).*/\1/' \             # restrict to keywords 'ok' & 'good'
  | csvsed -c Age 'e/xargs -I {} echo "{}*2" | bc/'      # double the Age column

Employee ID,Age,Wage,Status
8783,94,104343873.83,good
2003,64,98878784.00,ok

Installation

$ pip install csvsed

Usage and Examples

Installation of the csvsed python package also installs the csvsed command-line tool. Use csvsed --help for all command line options, but here are some examples to get you going. Given the input file sample.csv:

Employee ID,Age,Wage,Status
8783,47,"104,343,873.83","All good, but nowhere to go."
2003,32,"98,878,784.00",A-OK

Removing thousands-separators from the "Wage" column using the "s" (substitute) modifier:

$ cat sample.csv | csvsed -c Wage s/,//g
Employee ID,Age,Wage,Status
8783,47,104343873.83,"All good, but nowhere to go."
2003,32,98878784.00,A-OK

Convert/extract some text using the "s" (substitute) and "y" (transliterate) modifiers:

$ cat sample.csv | csvsed -c Status 's/^All (.*),.*/\1/' \
  | csvsed -c Status 's/^A-(.*)/\1/' \
  | csvsed -c Status 'y/a-z/A-Z/'
Employee ID,Age,Wage,Status
8783,47,"104,343,873.83",GOOD
2003,32,"98,878,784.00",OK

Square the "Age" column using the "e" (execute) modifier:

$ cat sample.csv | csvsed -c Age 'e/xargs -I {} echo "{}^2" | bc/'
Employee ID,Age,Wage,Status
8783,2209,"104,343,873.83","All good, but nowhere to go."
2003,1024,"98,878,784.00",A-OK

That, however, called the external program for each column (quite inefficient with large data sets)... so let's do that more efficiently, with a "continuous" mode program. Given the following id2name.py program which takes a CSV on STDIN with a single column (an employee ID) and writes a CSV to STDOUT with the IDs converted to names:

#!/usr/bin/env python
import sys, csvkit
table = {'8783': 'ElfenKyng', '2003': 'Stradivarius'}
# NOTE: *not* using csvkit's reader because it reads-ahead
# causing problems since this must be stream-oriented...
writer = csvkit.CSVKitWriter(sys.stdout)
while True:
  item = sys.stdin.readline()
  if not item: break
  item = item.strip()
  writer.writerow([table[item] if item in table else item])
  sys.stdout.flush()

Then the following will efficiently convert the 'Employee ID' column to names:

$ cat sample.csv | csvsed -c 'Employee ID' 'e|./id2name.py|c'
Employee ID,Age,Wage,Status
ElfenKyng,47,"104,343,873.83","All good, but nowhere to go."
Stradivarius,32,"98,878,784.00",A-OK