Inspired by docopt (, docprompt allows developers to write configuration prompts in RST or docstring format.

Why? Because I need this shit for my research project l-o-l.


Suppose you want to write an interactive configurator for a blog system, you can express the entire configuration in RST or docstring.

In either a RST file or as a docstring in a Python file:

Database Configurations

XXX Blog Engine works with major SQL servers. You must provide
proper credentials to connect the engine with the database.

    :prompt: Choose database (mysql,postgres,oracle) [mysql]:
    :name: db_type
    :choices: ['mysql', 'postgres', 'oracle']
    :default: 'mysql

    :prompt: Database host [localhost]:
    :name: db_host
    :default: 'localhost'

    :prompt: Database user:
    :name: db_user

Site Configurations

Almost done! Just a few questions to configure your site!

    :prompt: Choose an admin account [admin]:
    :name: admin_acc_name
    :default: 'admin'

    :prompt: Allows search engines (y/n) [y]:
    :name: allow_search
    :choices: ['y', 'n']
    :default: 'n'


There are two parts:

  1. construct prompts in rst/docstring
  2. access the results from users

Construct prompts

docprompt provides two required arguments prompt and name, as well as two optional arguments called default and choices.

A prompt is an interactive ask/answer input. All inputs will be treated as raw strings. We don't implement casting because a configurator will either write to a config file or pass the information to some functions such as connecting to database. Thus, in nearly every single real usecase, casting a string to another type (e.g. int) should be handled by the developers.

name is the variable name used to reference the answer from the user. It has to be a legal python variable name. We encourage to adapt standard variable naming convention. An example of a variable name is db_host instead of db-host which is illegal as a variable name.

For a simple question without need for choices or default:

:prompt: <Start typing your question here>
:name: <legal-python-variable-name>

In addition, if prompts requires a default value, add default`:

:default: 'a string'

Finally, if a list of choices are required, add choices:

:choices: ['item1-in-string', 'item2-in-string', 'and-so-on']

Order doesn't matter, but for consistency we encourage to use the format shown in the examples above.

Access results from users

All the name of all prompts are collected and made into member attributes of the object docprompt.

For example, if we assume the text in the example above is a global docstring (the ver first one, actually), we can construct the example above:

import docprompt

if __name__ == '__main__':
    quick_start = docprompt(source=__doc__)
    assert quick_start.db_host in ['mysql', 'postgres', 'oracle']

You are welcome to write up multiple doc strings using multiple docprompt objects. This is common for an application provides a command-line interface.

Bonus: Build command-line interface with docopt

I actually find docopt very handy for building a command-line client. My project docprompt can migiate the need to write a bunch of raw_input.

Suppose you want to build a command-line interface, first construct the commands using docopt, and then use docprompt to construct the prompts and access the user inputs in a novel way!