v0.5.3

- Minor CI tweaks and README update

v0.5.2

- [FIXED] Add cmath include for M_PI
- [ADDED] Callbacks (#131)
- [FIXED] Add virtual destructor to GaussTransform
- [FIXED] Correspondence calculation (#137)
- [REMOVED] AppVeyor
- [CHANGED] From travis to github actions

v0.5.1

- [FIXED] Documentation builds on Travis.
- [FIXED] Setting the GaussTransform on a Transformation.
- [FIXED] Bug in the normalization test (#107).
- [ADDED] Test for scaling in rigid transformations.
- [ADDED] Better error message when allocation fails when creating an affinity matrix (#104).
- [ADDED] A new example, `random`, that can be used to test large allocations.
- [ADDED] An example to demonstrate setting the transform to fgt.
- [ADDED] "Unlinked" normalization, where the two datasets can be scaled differently (#112).
- [ADDED] A `matrix` method for both rigid and affine results. This matrix contains the entire transformation. Also added an example that calculates and applies a transformation matrix (#108).
- [ENHANCED] The `rigid` example.
- [ENHANCED] CMake configuration for jsoncpp builds (#105).
- [ENHANCED] Documentation.

v0.5.0

There were enough wires to rip up that I'm mashing all of these changes
into one big changeset for the v0.5 update.

The big changes include:

- Remove the `Runner`, move all the common iterative logic to a
  templated `Transform` virtual base class, from which `Rigid` and
  friends derive.
- Remove vendored upstream dependencies (other than gtest). Managing
  vendored dependencies was a big pain. Instead, I'm trying out
  "components", e.g. the jsoncpp functionality is included as a
  component that requires an external jsoncpp to be present. We'll see
  how that goes.
- Remove all logging, for now. I might add it back (via the component
  mechanism mentioned above) but I'm not sure. While it's useful for
  keeping tabs, it mucks up the business part of the code a lot. I have
  a vauge idea about adding callbacks into the main cpd loop and using
  those for logging, but nothing's there yet.
- Refactored tests and fixtures, to hopefully make it easier to
  modify/etc in the future. This rewrite mostly happened becuase I
  couldn't bring myself to rip apart the v0.4 test suite without just
  redoing it all.

Note that a lot of the business code in this rewrite was *not* touched —
this is more of an interface change than anything that affects the
actual math.

v0.4.3

- [FIXED] Bug in rigid result's points.
- [FIXED] Whitespace in .travis.yml.
- [ADDED] The affine transformation.
- [ADDED] Documentation.
- [ADDED] Seperate script to build/install doxygen on Travis.
- [CHANGED] Rename affinity_matrix files to affinity.
- [CHANGED] Refactor test fixtures.
- [CHANGED] Use googletests's targets, but don't install googletest files.
- [CHANGED] Fgt is now an optional dependency, that defaults to ON if fgt is present.
- [CHANGED] If a doxygen executable is present, default to `WITH_DOCS=ON`.
- [CHANGED] Move Doxygen.in to a docs/ directory.

v0.4.2

- [FIXED] Denormalization bug for rigid transformations
- [ADDED] Runtime (in microseconds) to results
- [ADDED] More information to logging
- [ADDED] Example of using PDAL and artificial translation

v0.4.1

- [FIXED] Normalization for non-unit-cube point sets.
- [FIXED] Documentation builds on Travis.
- [FIXED] Include syntax in normalize.hpp.
- [FIXED] MSVC warnings.
- [FIXED] Datatype of `Runner.m_max_iterations`.
- [FIXED] Bad AppVeyor link in README.
- [ADDED] Allow various Fgt strategies.
- [ADDED] CMake policy CMP0054.
- [ADDED] sigma2 test for identical points.
- [ADDED] Better sigma2 logging.
- [ADDED] Helheim Glacier test data file.
- [CHANGED] Refactored top-level registration functions to use a more general uplevel upstream function.
- [CHANGED] `ProbabilityComputer` to `Comparer`.
- [CHANGED] Explicitly build static libs in AppVeyor.
- [CHANGED] Default to static libs in MSVC.
- [CHANGED] Precision on sigma2 reporting.

v0.4.0

Complete rewrite, again.