dlib 0.3

• dlib.core
  • Added simple yet robust I/O streams (dlib.core.stream), which are completely Phobos-independent
• dlib.filesystem
  • Abstract FS interface and it's implementations for Windows and POSIX filesystems
• dlib.image
  • Breaking change: all pixel I/O is now floating-point (via Color4f). This gives an opportunity to define image classes of arbitrary floating-point pixel formats and enables straightforward HDRI: sample 32-bit implementation provided in dlib.image.hdri
  • Pixel iteration now can be done with row and col ranges
  • Parallel filtering is now easy with dlib.image.parallel. You can add multithreading to your existing filter code with just a few changes
  • Added support for TGA and BMP formats (only loading for now)
  • All image format I/O is now stream-based
• dlib.math
  • Breaking change: matrices in dlib.math.matrix are now column-major
  • Imporved constness support in dlib.math.vector, as well as added unittest to the module. Using new string constructor, vectors now can be parsed from strings (e.g., "[0, 1, 2]")
• Overall improvements & bugfixes
  • Much saner DUB support, addressed some serious problems with building, added configuration for pre-compiling as a static library