1. Getting started

1.1. About this tutorial

There is no better way to learn how to use a new tool than to see it applied in a real world situation. To that end, this tutorial explains how to use csvkit tools by analyzing a real dataset.

The data we will be using is a subset of the United States Defense Logistic Agency Law Enforcement Support Office’s (LESO) 1033 Program dataset, which describes how surplus military arms have been distributed to local police forces. This data was widely cited in the aftermath of the Ferguson, Missouri protests. The particular data we are using comes from an NPR report analyzing the data.

This tutorial assumes you have some basic familiarity with the command line. If you don’t have much experience, fear not! This has been written with beginners in mind. No prior experience with data processing or analysis is assumed.

1.2. Installing csvkit

Installing csvkit is easy:

sudo pip install csvkit

If you have problems installing, look for help in the Tips and Troubleshooting section of the documentation.


If you’re familiar with virtualenv, it is better to install csvkit in its own environment. If you are doing this, then you should leave off the sudo in the previous command.

1.3. Getting the data

Let’s start by creating a clean workspace:

mkdir csvkit_tutorial
cd csvkit_tutorial

Now let’s fetch the data:

curl -L -O https://raw.githubusercontent.com/wireservice/csvkit/master/examples/realdata/ne_1033_data.xlsx

1.4. in2csv: the Excel killer

For purposes of this tutorial, I’ve converted this data to Excel format. (NPR published it in CSV format.) If you have Excel you can open the file and take a look at it, but really, who wants to wait for Excel to load? Instead, let’s convert it to a CSV:

in2csv ne_1033_data.xlsx

You should see a CSV version of the data dumped into your terminal. All csvkit tools write to the terminal output, called “standard out”, by default. This isn’t very useful, so let’s write it to a file instead:

in2csv ne_1033_data.xlsx > data.csv

data.csv will now contain a CSV version of our original file. If you aren’t familiar with the > syntax, it means “redirect standard out to a file”. If that’s hard to remember it may be more convenient to think of it as “save to”.

We can verify the that the data is saved to the new file by using the cat command to print it:

cat data.csv

in2csv can convert a variety of common file formats to CSV, including both .xls and .xlsx Excel files, JSON files, and fixed-width formatted files.

1.5. csvlook: data periscope

Now that we have some data, we probably want to get some idea of what’s in it. We could open it in Excel or Google Docs, but wouldn’t it be nice if we could just take a look in the command line? To do that, we can use csvlook:

csvlook data.csv

At first the output of csvlook isn’t going to appear very promising. You’ll see a mess of data, pipe character and dashes. That’s because this dataset has many columns and they won’t all fit in the terminal at once. You have two options:

  1. Pipe the output to less -S to display the lines without wrapping and use the arrow keys to scroll left and right:
csvlook data.csv | less -S
  1. Reduce which columns of our dataset are displayed before we look at it. This is what will do in the next section.

1.6. csvcut: data scalpel

csvcut is the original csvkit tool. It inspired the rest. With it, we can select, delete and reorder the columns in our CSV. First, let’s just see what columns are in our data:

csvcut -n data.csv
 1: state
 2: county
 3: fips
 4: nsn
 5: item_name
 6: quantity
 7: ui
 8: acquisition_cost
 9: total_cost
10: ship_date
11: federal_supply_category
12: federal_supply_category_name
13: federal_supply_class
14: federal_supply_class_name

As you’ll can see, our dataset has fourteen columns. Let’s take a look at just columns 2, 5 and 6:

csvcut -c 2,5,6 data.csv

Now we’ve reduced our output CSV to only three columns.

We can also refer to columns by their names to make our lives easier:

csvcut -c county,item_name,quantity data.csv

1.7. Putting it together with pipes

Now that we understand in2csv, csvlook and csvcut we can demonstrate the power of csvkit’s when combined with the standard command-line “pipe”. Try this command:

csvcut -c county,item_name,quantity data.csv | csvlook | head

In addition to specifying filenames, all csvkit tools accept an input file via “standard in”. This means that, using the | (“pipe”) character we can use the output of one csvkit tool as the input of the next.

In the example above, the output of csvcut becomes the input to csvlook. This also allow us to pipe output to standard Unix commands such as head, which prints only the first ten lines of its input. Here, the output of csvlook becomes the input of head.

Piping is a core feature of csvkit. Of course, you can always write the output of each command to a file using >. However, it’s often faster and more convenient to use pipes to chain several commands together.

We can also pipe in2csv, allowing us to combine all our previous operations into one:

in2csv ne_1033_data.xlsx | csvcut -c county,item_name,quantity | csvlook | head

1.8. Summing up

All the csvkit tools work with standard input and output. Any tool can be piped into another and into another. The output of any tool can be redirected to a file. In this way they form a data processing “pipeline” of sorts, allowing you to do non-trivial, repeatable work without creating dozens of intermediary files.

Make sense? If you think you’ve got it figured out, you can move on to Examining the data.