ExDoc

Documentation extractor.

Extracts pieces of documentation from your code to build a document which can be fed to template processors.

Output can be in JSON, YAML, whatever. Use any command-line templating engine, like j2cli, to render templates from it.

It does not do any automatic background magic: it just offers helpers which allows you to extract the necessary pieces.

Currently supports parsing the following documentation formats:

• Sphinx (reST)
• Google

Collectors

ExDoc is just a set of helper functions that collects information into dictionaries.

Python

Helpers for Python objects

doc(obj, of_class=None)

Get parsed documentation for an object as a dict.

This includes arguments spec, as well as the parsed data from the docstring.

from exdoc import doc

The doc() function simply fetches documentation for an object, which can be

• Module
• Class
• Function or method
• Property

The resulting dictionary includes argument specification, as well as parsed docstring:

def f(a, b=1, *args):
    ''' Simple function

    : param a: First
    : type a: int
    : param b: Second
    : type b: int
    : param args: More numbers
    : returns: nothing interesting
    : rtype: bool
    : raises ValueError: hopeless condition
    '''

from exdoc import doc

doc(f)  # ->
{
  'module': '__main__',
  'name': 'f',
  'qualname': 'f',  # qualified name: e.g. <class>.<method>
  'signature': 'f(a, b=1, *args)',
  'qsignature': 'f(a, b=1, *args)',  # qualified signature
  'doc': 'Simple function',
  'clsdoc': '',  # doc from the class (used for constructors)
  # Exceptions
  'exc': [
    {'doc': 'hopeless condition', 'name': 'ValueError'}
  ],
  # Return value
  'ret': {'doc': 'nothing interesting', 'type': 'bool'},
  # Arguments
  'args': [
    {'doc': 'First', 'name': 'a', 'type': 'int'},
    {'default': 1, 'doc': 'Second', 'name': 'b', 'type': 'int'},
    {'doc': 'More numbers', 'name': '*args', 'type': None}
  ],
}

Note: in Python 3, when documenting a method of a class, pass the class to the doc() function as the second argument:

doc(cls.method, cls)

This is necessary because in Python3 methods are not bound like they used to. Now, they are just functions.

getmembers(obj, *predicates)

Return all the members of an object as a list of (key, value) tuples, sorted by name.

The optional list of predicates can be used to filter the members.

The default predicate drops members whose name starts with '_'. To disable it, pass None as the first predicate.

subclasses(cls, leaves=False)

List all subclasses of the given class, including itself.

If leaves=True, only returns classes which have no subclasses themselves.

SqlAlchemy

Documenting SqlAlchemy models.

from exdoc.sa import doc

doc(User)  # ->
{
  'name': 'User',
  # List of tables the model uses
  'table': ('users',),
  'doc': 'User account',
  # PK: tuple[str]
  'primary': ('uid',),
  # Unique keys
  'unique': (
    # tuple[str]
    ('login',),
  ),
  # Foreign keys
  'foreign': (
    {'key': 'uid', 'target': 'users.uid', 'onupdate': None, 'ondelete': 'CASCADE'},
  ),
  # Columns
  'columns': [
    {'key': 'uid', 'type': 'INTEGER NOT NULL', 'doc': ''},
    {'key': 'login', 'type': 'VARCHAR NULL', 'doc': 'Login'},
    {'key': 'creator_uid', 'type': 'INTEGER NULL', 'doc': 'Creator'},
    {'key': 'meta', 'type': 'JSON NULL', 'doc': ''},
  ],
  # Relationships
  'relations': [
    {'key': 'creator', 'model': 'User',
     'target': 'User(creator_uid=uid)', 'doc': ''},
    {'key': 'devices[]', 'model': 'Device',
     'target': 'Device(uid)', 'doc': ''},
    {'key': 'created[]', 'model': 'User',
     'target': 'User(uid=creator_uid)', 'doc': ''},
  ]
}

Building

Create a python file that collects the necessary information and prints json:

#! /usr/bin/env python
from exdoc import doc
import json

from project import User

print json.dumps({
  'user': doc(User),
})

And then use its output:

./collect.py | j2 --format=json README.md.j2