Skip to content

Why Columbo?

columbo started as an internal library used at Wayfair by the Python Platforms team. It is now an interdependently maintained open source project with the hope that it will provide benefit from the greater Python community.

What problem did Columbo initially solve?

One of the primary resources the Python Platforms team provided to Wayfair engineers was a set of project templates. This allowed an engineering team to hit the ground running when starting a new HTTP application or Python package. The templates contained best-practice patterns and allow developers to initialize and deploy a new application to production in minutes.

The project templates used Cookiecutter in order to generate the files that will be used in the new project. While Cookiecutter makes many things easy for maintainers of a template, the terminal user interface it provides has a few issues:

  • The text Cookiecutter presents to the user is also the name used to access the value within the template. This can make it hard for a user to know exactly what information the template is asking for.
  • Cookiecutter always prompts the user for a value for every item defined by the template. Our templates have optional features that need additional information when used. When a user chooses not to use a given feature, they are still prompted to provide a value for the items that will not be used when generating files.
  • Cookiecutter does not support validating given values when the user prompting the user to provide input. If a template requires the value for the first item to be a number, the user will be prompted for values for the remaining items. When the validation error occurs, the user always needs to start again from scratch.

Cookiecutter can be executed programmatically by passing in a dictionary of values. This allows columbo to handle the user prompts resolving the issues we experienced, while still using Cookiecutter under the covers. With this pattern, all project templates managed by language platform teams at Wayfair benefited from the user experience of working with Cookiecutter templates.

Alternatives

columbo is not the only library that exists which provides a way to codify prompting a user for answers to a set of questions. This section compares columbo with some libraries which were created to achieve this task.

PyInquirer

PyInquirer was a major inspiration in the development of columbo. PyInquirer supports optional questions and validation in similar way to how columbo supports those features.

PyInquirer has some features that columbo does not. It has some additional interaction types like selecting multiple options from a list and using an external editor to provide a long form response. It also exposes more rendering options of prompt-toolkit, which is used internally by both libraries.

columbo has some features that PyInquirer does not. PyInquirer does not support dynamic values as arguments used to construct a question. Additionally, PyInquirer does not provide an option to parse command line arguments as answers to a set of questions.

The primary reason why we chose to create columbo instead of using PyInquirer was that the project was still using version 1 of prompt-toolkit while the maintainers of that library were preparing to release version 31. Additionally, the library uses an un-typed dictionary to describe each question, which can make it easier to make configuration mistakes.

python-inquirer

python-inquirer is similar to columbo in that they both support dynamic values, validation and optional questions. They also both use classes to represent interactions.

python-inquirer has some features that columbo does not. It has some additional interaction types like selecting multiple options from a list and using an external editor to provide a long form response. It also exposes more rendering options of blessed.

columbo has some features that python-inquirer does not. python-inquirer does not provide an option to parse command line arguments as answers to a set of questions or have interactions that do not store an answer. Additionally, colombo's API is fully type annotated, while python-inquirer's API is not.

questionary

questionary is similar to columbo in that they both support validation, optional questions, and use classes to represent interactions2. The API's for both libraries are fully type annotated.

questionary has some features that columbo does not. It has some additional interaction types like supporting tab completion for some answers. It also exposes more rendering options of prompt-toolkit, which is used internally by both libraries. questionary can also ask questions asynchronously.

columbo has some features that questionary does not. questionary does not support dynamic values as arguments used to construct a question or provide an option to parse command line arguments as answers to a set of questions.

  1. At the time of writing this document (February 2021), the PyInquirer repo has been updated to use version 3 of prompt-toolkit, but a new release has not been published to PyPi. ↩

  2. Dynamic optional questions requires questions to be specified using a dictionary instead of the class representation. ↩