Pydocx

- Headings now preserve italic, webHidden and vanish styles

- Decimal font sizes are now handled properly

- Paragraphs that have numbering definitions with a level number format of None
are no longer considered list items.

- Headings in lists no longer break numbering. By default, in the HTML
exporter, headings in lists are represented using the "strong" tag,
regardless of the heading level.

- Note: This release consists of significant changes to the internal API and is not
backwards compatible with prior versions
- Removed ``ConvertRootUpperRomanListToHeadingMixin``
- Fixed issue where the same image referenced multiple times would not
display correctly after the first instance
- Removed the preprocessor and re-implemented the functionality into the exporter
- Re-implemented the exporter into a top-down generator algorithm
- Implemented the necessary object classes for each element type (Paragraph,
Run, Text, etc)
- Implemented enumerated list detection and conversion to numbering lists

- Added support for python 3.4
- Added support for pypy
- No longer adding list-style-type attribute to ordered list tags.
We are now using a class to indicate these.
- Faked sub/super handling is no longer handled by default.
Instead,
that handling is implemented in a new mixin class.
See ``pydocx.export.mixins``
- ``pydocx.wordml`` and ``pydocx.openxml``
have been merged into ``pydocx.openxml.packaging``
to better mirror the MS implementation structure.
- ``pydocx.models.styles``
has been moved to
``pydocx.openxml.wordprocessing.*``
- ``pydocx.managers.styles``
has been merged into
``pydocx.openxml.wordprocessing.style_definition_part``
- Added
``XmlCollection``
field type,
now used by ``openxml.wordprocessing.styles.Style``
- Implemented several model classes for Numbering.
- Added numbering property to the numbering definitions part.
- XmlModels now define their own tags
- Simplified importing PyDocX
- Header processing now occurs in the exporter rather than the pre-processor
- PyDocXExporter.heading signature has changed from accepting
heading_level which was an HTML tag
to accepting
heading_style_name
which is the raw style name of the heading.
- The ``convert_root_level_upper_roman``
option has been replaced
with an optional mixin
``pydocx.export.mixins.ConvertRootUpperRomanListToHeadingMixin``.
- Preprocessor no longer manages table membership.
Instead, that is handled in the base iterative parser.
- ``ConvertRootUpperRomanListToHeadingMixin``
would fail for paragraphs that had no properties.