Sqlacodegen

- Refactored code for better reuse
- Dropped support for Python 2.6, 3.2 and 3.3
- Dropped support for SQLAlchemy < 0.8
- Worked around a bug regarding Enum on SQLAlchemy 1.2+ (``name`` was missing)
- Added support for Geoalchemy2
- Fixed invalid class names being generated (fixes 60; PR by Dan O'Huiginn)
- Fixed array item types not being adapted or imported
(fixes 46; thanks to Martin Glauer and Shawn Koschik for help)
- Fixed attribute name of columns named ``metadata`` in mapped classes (fixes 62)
- Fixed rendered column types being changed from the original (fixes 11)
- Fixed server defaults which contain double quotes (fixes 7, 17, 28, 33, 36)
- Fixed ``secondary=`` not taking into account the association table's schema name
(fixes 30)
- Sort models by foreign key dependencies instead of schema and name (fixes 15, 16)

- Fixed compatibility with SQLAlchemy 1.0
- Added an option to only generate tables

- Fixed potential assignment of columns or relationships into invalid attribute names
(fixes 10)
- Fixed unique=True missing from unique Index declarations
- Fixed several issues with server defaults
- Fixed potential assignment of columns or relationships into invalid attribute names
- Allowed pascal case for tables already using it
- Switched from Mercurial to Git

- Fixed compatibility with SQLAlchemy 0.9.0

- Fixed compatibility with SQLAlchemy 0.8.3+
- Migrated tests from nose to pytest

- Fixed non-default schema name not being present in __table_args__ (fixes 2)
- Fixed self referential foreign key causing column type to not be rendered
- Fixed missing "deferrable" and "initially" keyword arguments in ForeignKey constructs
- Fixed foreign key and check constraint handling with alternate schemas (fixes 3)