You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Collect docnames of analyzed source in advance (#2)
* Remove container_of and support "of {key: value}" syntax
This syntax is used by Pandas [1]. I'm not entirely sold yet, but let's
add it for now. To avoid confusion, literals should probably be made to
only be accepted on the top-level. Otherwise something like
`dict of {{"a", "b"}: int}` becomes possible.
Co-authored-by: Oriol Abril-Pla <oriol.abril.pla@gmail.com>
* Only allow literals on top-level
which avoids potentially confusing constructs like
`dict of {{"a", "b"}: int}`
Co-authored-by: Oriol Abril-Pla <oriol.abril.pla@gmail.com>
* Make "= | :" optional in default syntax
since this is what NumPyDoc recommends as well [1].
[1] https://numpydoc.readthedocs.io/en/latest/format.html#parameters
Co-authored-by: Oriol Abril-Pla <oriol.abril.pla@gmail.com>
* Add ipython as dev dependency
* Move matching to imports to qualname level
Doing this on the NAME level meant that stuff like "np.int16" would be
turned into "np.np.int16" because both "np" and "int16" where matched.
* Split docname into KnownImport and replace
Reduce responsibility of the former DocName class. Replacing docstring
specific type description should be handled separately.
* Test correctness when nesting classes in classes
* Rework analysis and other major changes
While not perfect or anywhere near to finished the refactored code
uses clearer separation of responsiblities and is one step closer to an
architecture that feels right. :)
---------
Co-authored-by: Oriol Abril-Pla <oriol.abril.pla@gmail.com>
0 commit comments