Revision history for Statistics-Sampler-Multinomial

1.02 20220924
  - Use an internal clone method for the data.  This is much faster than Clone::clone.

1.01 20201130
  - Add missing dependencies to Makefile.PL

1.00 20201128
  - The draw method now uses the faster approach.
    The old draw method is now called "draw_slow".
    "draw1" is no longer available as a method
    (please raise an issue if you believe it should be).
  - Add a subclass, Statistics::Sampler::Multinomial::Indexed,
    that uses a hierarchical index to speed up the draw
    calls, at the cost of building and updating the index.

0.87 20201127
  - Add a draw1 method as a faster means of drawing a single class.
    It has not replaced the draw method (yet) because the
    random draw sequence differs between the two so repeatability
    is affected.  It might become the default in version 1, though. 

0.86 20201023
  - A few micro-optimisations

0.85 20201020
  - Statistics::Sampler::Multinomial: Add update_values, get_data, and get_sum methods
    AliasMethod is unchanged.  

0.84 20200930
  - Add draw_with_mask and draw_n_samples_with_mask to Statistics::Sampler::Multinomial

0.83 20200909
  - Modify Statistics::Sampler::Multinomial->draw method
    to match documentation (and Statistics::Sampler::Multinomial::Alias->draw)

0.82 20180915
  - add license meta, thanks MANWAR

0.81 20180915
  - use same version across pm files

0.8 20180915
  - micro-optimisation in S::S::M::draw_n_samples

0.7 20170219
  - See if we can avoid more test failures on some systems,
    this time due to planning twice errors.

0.6 20170218
  - Test changes only
  - Check $Config{ivsize} to determine if we should skip.
    The archname tests missed too many cases.

0.5 20170217
  - Skip most tests on x86 architectures, as Math::Random::MT::Auto
    generates different PRNG streams for a given seed compared with x64
    (unless it is an x86 long int build).

0.4 20170216
  - No functional changes from the 0.3_00x dev releases
  - Document that Alias method results will differ between
    double and long double builds

0.3_004 20170214
  - Skip some tests on long double builds

0.3_003 20170214
  - Shift to a tolerance based test (avoid sprintf)

0.3_002 20170214
  - Set minimum perl version to 5.014

0.3_001 20170214
  - See if we can get more insight into cpan testers fails
  - Add explicit return to cuckoo packages
  - Add checks for PRNG sequences

0.2  20170214
  - Fix typo in abstract
  - No functional changes

0.1  20170214
  - Initial release.