Previous
Next

Using scripts

Licenced under CC-BY 4.0 and OSI-approved licenses, see licensing.

Overview

Teaching: 10 min
Exercises: 10 min
Questions

  • How can we document the data-cleaning steps we’ve applied to our data?

  • How can we apply these steps to additional data sets?

Objectives

  • Describe how OpenRefine generates JSON code representing the work done in an analysis session.

  • Demonstrate ability to export this JSON code from OpenRefine.

  • Demonstrate ability to import a JSON code file to apply the analysis to another dataset.

Lesson

How OpenRefine records what you have done

As you conduct your data cleaning and preliminary analysis, OpenRefine saves every change you make to the dataset. These changes are saved in a format known as JSON (JavaScript Object Notation). You can export this JSON script and apply it to other data files. If you had 20 files to clean, and they all had the same type of errors (e.g. misspellings, leading white spaces), and all files had the same column names, you could save the JSON script, open a new file to clean in OpenRefine, paste in the script and run it. This gives you a quick way to clean all of your related data.

Saving your work as a script

  1. In the Undo / Redo section, click Extract..., and select the steps you want to apply to other datasets by clicking the checkboxes.

History

  1. Click the Export button next to the window displaying the JSON script.
  2. A JSON file will be downloaded to your computer.

Importing a script to use against another dataset

Let’s practice running these steps on a new dataset. We’ll test this on an uncleaned version of the dataset we’ve been working with.

  1. Start a new project in OpenRefine using the messy dataset you downloaded before (samples_openrefine_lesson.csv). Give the project a new name.
  2. Click the Undo / Redo tab > Apply and then select ‘Choose file’ and upload the JSON file you downloaded earlier.
  3. Click Run operations. The dataset should now match your other cleaned dataset.

For convenience, we used the same dataset. In reality, you can use this process to clean related datasets. For example, data that you had collected over different fieldwork periods or data that was collected by other researchers (provided everyone uses the same column headings).