VTK/improved unicode support
Improved Unicode read/write support
A series of e-mails on the TitanDevelopers list about unicode support occurred this week leading to several proposals and a consensus was reached on October 7th 2010. Since the actual implementation of the basic functionality is in VTK I am posting here to get feedback on the method and classes proposed.
Proposed method
Current readers and writers (mainly the delimited text reader/writer) use either code specifically written to handle UTF16 and ASCII or use a 3 party library for UTF8 support. Knowledge of the specifics is written into each reader and adding a new method means directly editing each class. Since we expect that there will be more reader/writers requiring Unicode support and since we also expect that new encodings will continue to be added to the unicode standard we propose adding a factory method for finding a codec for a given format and modifying readers to accept that codec for use in translation.
Factory Pattern
We plan on a factory pattern similar to the one used to discover SQL Database capabilities in the vtkSQLDatabase class. Each class will register a callback at link time using a static object constructor.
Abstract Base class for codecs
A virtual base class for each codec to conform to will be needed so that a given reader/writer can deal with all of them as one.
Sample use of a codec for read and write
(In Progress...)
--prwolfe 14:49, 7 October 2010 (EDT)
