This forks some of the pydocstyle AST parser into out mapper. Eventually, some
of the other operations such as parsing arguments and performing full name
lookup can be moved in as well. For now, this is not doing any extra assignment
tracking/etc, several of these operations will just throw this information out.
Refs #99
* Moves relative path parsing away from the base mapper implementation
* Change argument parsing from splitting first line of source with ',' to use
AST traversal instead. This is not complete, but mostly PoC for now. Full
traversal into argument type nodes will allow us to get nested dict() etc.
We should open a ticket to track this work
* Cleans up some of the templates to reduce duplicate titles
* Adds a directive for nesting rST from constructs that might have headings.
Remove the first heading in this case to address the case where a module has a
docstring with a heading up front
* Adds tests
* Replaces example module with module that has more failing cases of parsing
Closes#78Fixes#80Fixes#81Fixes#82Fixes#83Fixes#84Fixes#85
This updates some small outdated pieces with docfx integration:
* Support docfx.json first, if no patterns were explicitly specified
* Refactor output path, use new _api path
* Add missing operator type to .net parsing and template output
* Fix indent issue with code samples
* Add docs on how to actually use docfx + autoapi
Fixes#45Fixes#46Fixes#48
This will resolve more lookups as proper references, not anchor links. The :any:
lookup is strict though, the domain should implement it's own :dn:obj: generic
lookup.
This resolves some issues with special characters in paths on Windows based
systems. It replaces special characters and unicode characters to ensure paths
are predictable.
Relies on regex replace and multiple replacement substitions, rather than
iterating over the pattern and applying logic to the found patterns.
Handles:
* see & seealso - creates a reference link
* paramref & typepararef - we can't reference params in sphinx, so this just
replaces with bold text
This removes dependence on real tooling.
The main downside is that it’s fragile to output changes in the tooling.
Not a lot we can do though,
other than have real integration tests, also.
