The first piece of advice that we can give is to be patient. Working with data on the command line is different from using a programming language, and therefore it requires a different mindset.

Moreover, the command-line tools themselves are not without their quirks and inconsistencies. This is partly because they have been developed by many different people, over the course of multiple decades. If you ever find yourself at a loss regarding their mind-dazzling options, don’t forget to use –help, man, or your favorite search engine to learn more.

The second, related piece of advice is to be creative. The command line is very flexible. By combining the command-line tools, you can accomplish more than you might think.

We encourage you to not immediately fall back onto your programming language. And when you do have to use a programming language, think about whether the code can be generalized or reused in some way. If so, consider creating your own command-line tool with that code using the steps we discussed in . If you believe your command-line tool may be beneficial for others, you could even go one step further by making it open source.

We’ve discussed four steps for doing data science at the command line. In practice, the applicability of the command line is higher for step 1 than it is for step 4. You should use whatever approach works best for the task at hand. And it’s perfectly fine to mix and match approaches at any point in your workflow. The command line is wonderful at being integrated with other approaches, programming languages, and statistical environments. There’s a certain trade-off with each approach, and part of becoming proficient at the command line is to learn when to use which.

In conclusion, when you’re patient, creative, and practical, the command line will make you a more efficient and productive data scientist.