dlib 0.6.0

• dlib.core
  • Got rid of ManuallyAllocatable interface in manual memory management for classes. Added support for deleting via interface or parent class. Deleting can be abstractized with Freeable interface
• dlib.filesystem
  • Added GC-free implementations for FileSystem and file streams
• dlib.image
  • dlib.image.unmanaged provides generalized GC-free Image class with corresponding factory function
  • JPEG decoder had been greatly improved, added more subsampling modes support, COM and APPn markers detection. Decoder now understands virtually any imaginable baseline JPEGs, including those from digital cameras
• dlib.math
  • New module dlib.math.combinatorics with factorial, hyperfactorial, permutation, combinations, lucas number and other functions
  • dlib.math.sse brings x86 SSE-based optimizations for some commonly used vector and matrix operations, namely, 4-vector arythmetics, dot and cross product, 4x4 matrix multiplication.
• dlib.container
  • DynamicArray now supports indexing (as a syntactic sugar).

dlib 0.5.3

• Added Protobuf-style varint implementation (dlib.coding.varint)
• Streams are now ManuallyAllocatable
• Triangle struct in dlib.geometry.triangle now has tangent vectors
• Fixed unittest build failure (#59)

dlib 0.5.2

• Automated vector type conversion (#57), modulo operator for vectors (#58)
• Fixed warning in dlib.image.io.bmp (#56)

dlib 0.5.1

Small bugfix release:

• Fixed wrong module name in dlib.geometry.frustum
• Updated license information

dlib 0.5.0

• dlib.core
  • Added manual memory management support. dlib.core.memory provide memory allocators based on standard C malloc/free. They can allocate arrays, classes and structs
  • Added prototype-based OOP system for structs (dlib.core.oop) with support for multiple inheritance and parametric polymorphism
• dlib.image
  • Image loaders are now GC-free
  • dlib.image.io.zlib and dlib.image.io.huffman modules are moved to new package dlib.coding. dlib.image.io.bitio moved to dlib.core.
  • Image allocation is based on a factory interface that abstracts over GC or MMM
  • Improved support for indexed PNGs - added alpha channel support
• dlib.container
  • Added GC-free dynamic array implementation (dlib.container.array)
  • BST and AArray now use manual memory management
• dlib.math
  • Quaternion is now based on and interchangeable with Vector via incapsulation
  • Dual quaternion support (dlib.math.dualquaternion)
  • Fixed incorrect dual number pow implementation
• dlib.geometry
  • Breaking change: Frustum plane normals are now pointing outside frustum. Also Frustum-AABB intersection API is changed
  • Fixed bugs in AABB and Plane

dlib 0.4.1

• dlib.image
  • Baseline JPEG decoder (dlib.image.io.jpeg)
• dlib.math
  • New matrix printer with proper alignment (a la Matlab)

dlib 0.4.0

• dlib.filesystem
  • Platform-specific modules are now grouped by corresponding packages (dlib.filesystem.windows, dlib.filesystem.posix)
  • findFiles is now a free function and can be used with any ReadOnlyFileSystem
• dlib.math
  • Implemented LU decomposition for matrices (dlib.math.decomposition)
  • dlib.math.linear is now dlib.math.linsolve. Added solveLU, a new LU-based direct solver
  • Matrix inversion now uses LU decomposition by default (4x performance boost compared to old analytic method). 4x4 affine matrices use an optimized inversion, which is about 6 times faster
  • dlib.math.utils now uses clamp from latest Phobos if available
  • Removed deprecated functionality
• dlib.core:
  • Moved container modules (bst, linkedlist, etc) from dlib.core to separate package dlib.container. Removed useless dlib.core.method
• Overall: improved compatibility with DMD 2.067.

dlib 0.3.3

Mainly bugfix release. Changes:

• Fixed compilation with DMD 2.066
• Added dlib.geometry.frustum
• Improved dlib.math.quaternion.

dlib 0.3.2

Bugfix release. The main improvement is fixed compilation with some versions of LDC.

dlib 0.3.1

Bugfix release.

Changes:

• Improved dlib.image, added interpolated pixel reading
• Added matrix addition and subtraction, tensorProduct now works with any matrix sizes
• Addressed many bugs in dlib.image and dlib.math.