Regression Testing for epydoc.apidoc

This file serves to provide both documentation and regression tests for the epydoc.apidoc module. The main purpose of this module is to define the APIDoc class hierarchy, which is used to encode API documentation about Python programs. The API documentation for a Python program is encoded using a graph of APIDoc objects, each of which encodes information about a single Python variable or value.

>>> import epydoc; epydoc.DEBUG = True
>>> from epydoc.apidoc import *
>>> from epydoc.test.util import print_warnings
>>> print_warnings()

Unknown Value

Epydoc defines a special object, epydoc.apidoc.UNKNOWN, which is used as the value of attributes when their real value is not (yet) known.


This object only compares equal to itself:

>>> UNKNOWN == False
>>> UNKNOWN == True
>>> UNKNOWN == 0
>>> UNKNOWN == []
>>> UNKNOWN == object()

If UNKNOWN is used in a context where it is cast to bool, then it will raise an exception. This helps prevent acidentally interpreting an UNKNOWN value as true or false:

>>> if UNKNOWN:
...     print 'ok'
Traceback (most recent call last):
ValueError: Sentinel value <UNKNOWN> can not be used as a boolean

To test an attribute whose value might be UNKNOWN, you should explicitly compare that value to True or False. E.g.:

>>> x = UNKNOWN
>>> if x is True:
...     print 'we know x is true, and not unknown'
>>> if x is not False:
...     print 'x might be true or unknown.'
x might be true or unknown.
>>> if x in (True, UNKNOWN):
...     print 'x might be true or unknown.'
x might be true or unknown.

Dotted Names

The DottedName class is used to encode dotted names, such as 'epydoc.apidoc.DottedName', and make them easier to work with. Conceptually, a dotted name consists of a sequence of identifiers, separated by periods.

Dotted names can be constructed from strings:

>>> name1 = DottedName('')
>>> name1
DottedName('foo', 'bar')

Note that the given name is split on periods. You may also pass multiple strings to the constructor; they will be combined together into a single sequence:

>>> name2 = DottedName('x.y', 'z')
>>> name2
DottedName('x', 'y', 'z')

Each string can be a single identifier or a sequence of identifiers joined py periods. You may also pass DottedName objects to the constructor; their sequence of identifiers will be used:

>>> name3 = DottedName(name1, name2)
>>> name3
DottedName('foo', 'bar', 'x', 'y', 'z')

The string representation of a dotted name is formed by joining the identifiers with periods:

>>> str(name1)
>>> str(name2)
>>> str(name3)

The individual identifiers of a dotted name can be accessed via indexing; and the number of identifiers is returned by the len operator:

>>> name1[0], name1[1]
('foo', 'bar')
>>> name3[-1]
>>> name3[1:3]
DottedName('bar', 'x')
>>> len(name2)

As a result, you can iterate over the identifiers in a dotted name:

>>> for ident in name1:
...     print ident

Two dotted names compare equal if they have the same number of identifies and they are pairwise equal:

>>> DottedName('') == DottedName('foo', 'bar')
>>> DottedName('') == DottedName('foo.baz')
>>> DottedName('') == DottedName('')

Dotted names may be combined with the addition operator:

>>> name1 + name2
DottedName('foo', 'bar', 'x', 'y', 'z')
>>> name1 + name2 == name3
>>> name2 + name1 == name3

The container method may be used to construct a new dotted name with the last identifier stripped off:

>>> name1.container()
>>> name3.container()
DottedName('foo', 'bar', 'x', 'y')

If a dotted name has only one identifier, then its container is None:

>>> print DottedName('baz').container()
>>> print name1.container().container()

It is an error to create an empty dotted name; or a dotted name that contains a string that's not a valid python identifier:

>>> DottedName()
Traceback (most recent call last):
InvalidDottedName: Empty DottedName
>>> DottedName('1+2', strict=True)
Traceback (most recent call last):
InvalidDottedName: Bad identifier '1+2'
>>> DottedName({})
Traceback (most recent call last):
TypeError: Bad identifier {}: expected DottedName or str
>>> DottedName('1+2', strict=False)
Identifier '1+2' looks suspicious; using it anyway.

The one exception is that '??' is treated as if it were a valid python identifier:

>>> DottedName('??', 'foo')
DottedName('??', 'foo')

This is used when we can't find any name for an object (e.g., if there's a class that was used as the base class, but is not contained in any module or class).

A dotted name can be queried into a context to obtain a reduced version:

>>> DottedName('').contextualize(DottedName('foo'))
>>> DottedName('').contextualize(DottedName(''))
DottedName('baz', 'qux')
>>> DottedName('').contextualize(DottedName('baz'))
DottedName('foo', 'bar')
>>> DottedName('').contextualize(DottedName('foo').container())
DottedName('foo', 'bar')
>>> DottedName('').contextualize(UNKNOWN)
DottedName('foo', 'bar')

But a contextualization can't reduce to an empty DottedName:

>>> DottedName('foo').contextualize(DottedName('foo'))

APIDoc Objects

API documentation about Python programs is broken into small pieces, each of which is encoded using a single APIDoc object. Each APIDoc object describes a single value, variable, or function argument.

The APIDoc base class has 2 direct subclasses, for the 2 basic types of entity that it can record information about: ValueDoc and VariableDoc. ValueDoc is further subclassed to specify the different pieces of information that should be recorded about each value type.

APIDoc objects record information about each entity using attributes. Attribute values may be specified in the constructor. Any attributes that are not specified will be given a default value (usually UNKNOWN). The APIDoc base class defines the attributes shared by all APIDoc objects: docstring, docstring_lineno, descr, summary, metadata, and extra_docstring_fields.

>>> api_doc = APIDoc(docstring='foo')
>>> api_doc.docstring
>>> api_doc.summary

The constructor does not accept positional arguments; and any keyword argument that does not correspond to a valid attribute will generate a TypeError (but only if epydoc.DEBUG is true):

>>> APIDoc('foo')
Traceback (most recent call last):
TypeError: __init__() takes exactly 1 argument (2 given)
>>> APIDoc(foo='foo')
Traceback (most recent call last):
TypeError: APIDoc got unexpected arg 'foo'

Any assignment to an attribute that's not valid will also generate a TypeError (but only if epydoc.DEBUG is true):

>>> api_doc = APIDoc(docstring='ds')
>>> = 0
Traceback (most recent call last):
AttributeError: APIDoc does not define attribute 'foo'

APIDoc defines a pretty-print method, pp(), which can be used to display the information that an APIDoc contains:

>>> val_doc = ValueDoc(pyval=3)
>>> var_doc = VariableDoc(name='x', value=val_doc)
>>> class_doc = ClassDoc(bases=(), variables={'x':var_doc})
>>> print class_doc.pp()
ClassDoc [0]
 +- bases = ()
 +- variables
    +- x => VariableDoc for x [1]
       +- is_public = True
       +- name = 'x'
       +- value
          +- ValueDoc [2]
             +- pyval = 3

This is mainly intended to be used as a debugging and testing tool. The attributes that will be pretty-printed for an APIDoc object are determined by its class's _STR_FIELDS variable. (But any attribute whose value is UNKNOWN will not be displayed.) Attributes are listed in alphabetical order.