Dhparser

..................................

- refactoring (breaking changes!) "Node.name" instead
of "Node.tag_name" and type Context is now called "Trail"
- various bug fixes
- documentation extended

..................................

- Added (experimental) support for adding markup to a tree.
(See nodetree.markup())
- Added content_of() and strlen_of()-functions to nodetree.py
which allow to select only parts of the tree.
- nodetree-content-map can now also be restricted to parts
of a tree.

..................................

- refactoring of error reporting: it is now done centrally through the
(newly introduced) syntaxtree.RootNode object; makes it faster and easier
- "step by step" guide added to the documentation
- XML is now a first class citizen for serialization next to S-expressions:
xml serialization can now also be read with syntaxtree.parse_xml() plus
better serialization with syntaxtree.Node.as_xml()
- added example: XML-Parser
- added optional CST-reporting on a case by case basis to the testing
framework, using the "*"-marker to indicate which test should add CSTs to
the report
- moved compilation support to a separate module: compile.py
- source mapping added if preprocessor changes source code
(see module preprocess.py)
- new semantics for syntaxtree.Node.__str__: now includes error messages;
use Node.content to retrieve the content without any error messages
- LaTeX-example: better AST-transformations.

...................................

- Workaround: Cython-dependency set to version 0.29, because DHParser
1.3+ could otherwise not be installed with pip. DHParser can
run completely without Cython. Cython is only needed, if you want
to speed up DHParser a little (up to 2x). Starting from DHParser 1.3.0
this requires the 3.0 alpha Version of Cython, because older versions
stumble over "from __future__ import annotations" (which however is
needed by the sphinx-documentation builder to build the type information
in the documentation properly).

..................................

- Complete refactoring of left-recursion algorithm. No performance surprises
or running out of stack space any more (see parse.Forward.__call__())
- transform functions now try to keep error messages attached to nodes when
reshaping tree (see transform.update_attr())
- now using argparse instead of hand-written code for argument parsing

..................................

first public release