If you want to be able to allow people to call some methods using None
you have to do use a sentinel object when you define the method.
_sentinel = object()
def foo(param1=_sentinel):
...
This would allow you to call foo(param1=None)
and be able to make the difference between a call like foo()
.
The problem is that when Sphinx does document the method it will write something like
mymodule.foo(param1=<object object at 0x108c1a520>)
How can I convince Sphinx to have a user friendly output for these functions?
Note, Imagine how the documentations look if you have 3-4 parameters using the sentinel approach.
I don't think it is possible to persuade Sphinx to be more "friendly" as long as you have a sentinel that creates an object outside the function. Sphinx' autodoc extension imports the module, which means that module-level code is executed.
Are you sure you can't use something like this?
def foo(param1=None):
if param1 == None:
param1 = whatever you want...
else:
...
This can be handled by manually specifying function signature in autodoc directive, e.g.:
.. automodule:: pymorphy.contrib.tokenizers
.. autofunction:: extract_tokens(foo, bar)
.. autofunction:: extract_words
The <object object at 0x108c1a520>
part of generated method signature can be changed by overriding the __repr__
method of the sentinel object.
_sentinel = type('_sentinel', (object,),
{'__repr__': lambda self: '_sentinel'})()
It will be rendered by Sphinx as something like this:
mymodule.foo(param1=_sentinel)