[MATH-1319] Refactor the package "o.a.c.m.random"

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 4.0
• Labels: api refactoring
• Assignee: Gilles Sadowski
• Reporter: Gilles Sadowski
• Created: 03/Feb/16 14:06
• Updated: 29/May/16 21:28
• Resolved: 29/May/16 21:28

Description

It is proposed to refactor the whole package in order to fix the various problems mentioned in recent issues and discussions
concerning the random number generators.

For background, please refer to the separate reports on the issue tracking system, and on the "dev" ML.

The main goals are:

1. Make it clear that implementations do not generate an arbitrary number of bits (at the request of the user); but rather as fixed size chunks (i.e. 32 bits, a.k.a. Java "int" type).
2. Allow implementations that generate 64 bits chunks (a.k.a. Java "long" type).
3. Separate the generation of sequences (for all Java primitive types) that are uniformly distributed (in the range of the corresponding type) from all further processing (e.g. required to match other distributions).

Issue Links

• relates to: MATH-1341 "RandomDataGenerator" is brittle
• supercedes: MATH-1309 RNG: public "setSeed" method should not be called from inside the constructor
• supercedes: MATH-1304 "nextBytes" methods in RNG implementations
• supercedes: MATH-1306 Add public void nextBytes(byte[] bytes, int position, int length) method into the RandomGenerator interface
• supercedes: MATH-734 Code cleanup: "ISAACRandom"
• supercedes: MATH-1307 Create a base class for all RNGs
• supercedes: MATH-1324 RNGs implement "Serializable" (package "o.a.cm.random")
• supercedes: MATH-1308 Deprecate and remove "AbstractRandomGenerator"

Sub-Tasks

• Factor out RNG implementations: Resolved
• "ValueServer" has various shortcomings: Resolved
• Replace usage (within CM) of old RNG API: Resolved
• Move "EmpiricalDistribution" to "o.a.c.m.distribution": Resolved