      random
      Copyright (c)  2011-2013 John Abbott, Anna M. Bigatti; code by John Abbott
      GNU Free Documentation License, Version 1.2
%!includeconf: ../aux-txt2tags/config.t2t
      TeXTitle{RandomSource}{code: John Abbott; doc: John Abbott, Anna M. Bigatti}


== Examples ==[examples]
%----------------------------------------------------------------------
- [ex-RandomSource1.C ../../examples/index.html#ex-RandomSource1.C]
- [ex-RandomSource2.C ../../examples/index.html#ex-RandomSource2.C]
- [ex-RandomLong1.C ../../examples/index.html#ex-RandomLong1.C]
- [ex-RandomBool1.C ../../examples/index.html#ex-RandomBool1.C]
- [ex-RandomBigInt1.C ../../examples/index.html#ex-RandomBigInt1.C]
-

Here is a typical example of how to use a ``RandomSeqLong``; note that
we create the generator **before** entering the loop, then **inside** the
loop we use the function ``NextValue`` to get 100 successive random values
(between -9 and 9) from the generator:
```
  RandomSeqLong rnd(-9,9);
  for (int i=0; i < 100; ++i)
    cout << NextValue(rnd) << endl;
```

== User documentation ==
%======================================================================

Below, in [[random RandomSourceOperations]] we list these handy
functions for generating random values:
 | ``RandomBool()``, ``RandomLong(lo, hi)``, ``RandomBigInt(lo, hi)``
they are probably just what you want in a simple program,
**but using them will make your code thread-unsafe**
(which is quite acceptable in a small program for personal use).

For a **thread-safe** solution you should create your own 
//random generator//.
If you just want to generate many random values of the
same type, you should consider using one of the three specialized
random sequence generators (which are faster than the more general ``RandomSource``):

- The class ``RandomSeqLong`` is for representing generators of
  (independent) uniformly distributed random integers (``long``)
  in a given range; the range is specified when creating the generator
  (and cannot later be changed).

- The class ``RandomSeqBigInt`` is for representing generators of
  (independent) uniformly distributed random integers ([[BigInt]]) in
  a given range; the range is specified when creating the generator
  (and cannot later be changed).

- The class ``RandomSeqBool`` models a binary random variable (with
  independent ``bool`` samples, each having 50% probability of being
  //true// and 50% of being //false//).

- The class ``RandomSource`` is for representing general sources of
  (pseudo-)randomness: you can use it to produce random ``bool``, 
  ``long``, and ``BigInt`` values.

-

=== Constructors ===[constructors]
%----------------------------------------------------------------------

All constructors have an optional argument which is the initial
//seed// -- it determines the initial state of the generator.  If you
do not give a seed, the default is ``0``.

If you create several random sequence generators of the same kind and with the same
seed, they will each produce **exactly the same sequence of values**.
If you want to obtain different results each time a program is run,
you can seed the generator with the system time (//e.g.// by supplying
as argument ``time(0)``); this is likely desirable unless you're
trying to debug a randomized algorithm.

For ``RandomSource`` there is also the ``reseed`` function documented
below (see [[random reseed]]).

==== RandomSource ====

   | ``RandomSource()``    | seeded with ``0`` |
   | ``RandomSource(n)``   | seeded with ``n`` |

For convenience, there is a global ``RandomSource`` object (belonging to
[[GlobalManager]]); you can get a reference to it by calling
``GlobalRandomSource()``, **but using it will make your code thread-unsafe**.

==== RandomSeqXXXX ====

There are two families of constructors.

Constructors with default seed (0):

 | ``RandomSeqBigInt(lo,hi)``    | seeded with ``0``        |
 | ``RandomSeqLong(lo,hi)``      | seeded with ``0``        |
 | ``RandomSeqBool()``           | seeded with ``0``        |


Constructors with given (small integer) seed:

 | ``RandomSeqBigInt(lo,hi, n)`` | seeded with ``abs(n)``   |
 | ``RandomSeqLong(lo,hi, n)``   | seeded with ``abs(n)``   |
 | ``RandomSeqBool(n)``          | seeded with ``abs(n)``   |

Each ``RandomSeqBigInt`` (or ``RandomSeqLong``) will produce (pseudo)
random values uniformly distributed in the range from ``lo`` to ``hi``
(with both extremes included).  An ``ERR::BadArg`` exception is thrown
if ``lo > hi``; the case ``lo == hi`` is allowed.

=== RandomSource Operations ===[RandomSourceOperations]
%----------------------------------------------------------------------
These are the most convenient functions for generating random values;
but they use ``GlobalRandomSource``, so they are **thread-unsafe**:
- ``RandomBool()`` -- returns ``true`` with probability 50%
- ``RandomBiasedBool(P)`` -- returns ``true`` with probability ``P`` (a ``double`` between 0 and 1)
- ``RandomLong(lo, hi)`` -- in range ``lo``..``hi`` (both ends included)
- ``RandomBigInt(lo, hi)`` -- in range ``lo``..``hi`` (both ends included)
- ``RandomSmallPrime(N)`` -- generate a random prime up to ``N``; error if ``N < 2`` or ``N >= 2^31``.
-

A cleaner way is to pass as an argument the specific ``RandomSource``
object to be used (in these examples we call it ``RndSrc``):
- ``RandomBool(RndSrc)``
- ``RandomLong(RndSrc, lo, hi)`` -- in range ``lo``..``hi`` (both ends included)
- ``RandomBigInt(RndSrc, lo, hi)`` -- in range ``lo``..``hi`` (both ends included)
-

==== Reseed ====[reseed]

A ``RandomSource`` object may be reseeded at any time; immediately after reseeding
it will generate the same random sequence as a newly created ``RandomSource``
initialized with that same seed.  The seed must be an integer value.
- ``reseed(RndSrc, seed)``
-

**Note about thread-safety:** the various operations on a fixed
``RandomSource`` object are not thread-safe; to achieve thread safety, you
should use //different objects in different threads//.  In particular, it is
best not to use ``GlobalRandomSource()`` in a multi-threaded environment.


=== RandomSeqXXXX Operations ===[RandomSeqXXXXOperations]

Once you have created a ``RandomSeqXXXX`` you may perform the following
operations on it:
- ``*rnd``             -- get the current value of rnd (as a boolean).
- ``++rnd``            -- advance to next value of rnd.
- ``rnd++``            -- advance to next value of rnd **INEFFICIENTLY**.
- ``NextValue(rnd)``   -- advance and then return new value; same as *++rnd
- ``out << rnd``       -- print some information about rnd.
- ``rnd.myIndex()``    -- number of times rnd has been advanced,
                          (same as the number of random bools generated).

Note that a ``RandomSeqXXXX`` supports input iterator syntax.

Moreover, for ``RandomSeqBool`` there is a special function
- ``NextBiasedBool(RndBool, P)`` --  use several samples from ``RndBool`` to produce a boolean with probability ``P`` of being
                       true; may consume many values from ``RndBool`` but on average consumes less than 2 values per call.
-

You may assign or create copies of ``RandomSeqXXXX`` objects; the copies
acquire the complete state of the original, so will go on to produce exactly
the same sequence of values as the original will produce.


== Maintainer documentation ==
%======================================================================

==== RandomSource ====

The impl is mostly quite straightforward since almost all the work is
done by GMP.

``RandomLong(RndSrc, lo, hi)`` is a bit messy for two reasons:
+ CoCoALib uses signed longs while GMP uses unsigned longs;
+ the case when (``lo,hi``) specify the whole range of representable
      longs requires special handling.
+

==== RandomSeqLong  and  RandomSeqBigInt ====

The idea is very simple: use the pseudo-random number generator of GMP
to generate a random machine integer in the range 0 to ``myRange-1``
(where ``myRange`` was set in the ctor to be ``1+myUpb-myLwb``) and
then add that to ``myLwb``.  The result is stored in the data member
``myValue`` so that input iterator syntax can be supported.

There are two //non essential// data members: ``mySeed`` and ``myCounter``.
I put these in to help any poor blighter who has to debug a randomized
algorithm, and who may want to //fast forward// the random sequence to
the right place.

The data member ``myState`` holds all the state information used by the GMP
generator.  Its presence makes the ctors, dtor and assignment messier than
they would have been otherwise.

The advancing and reading member functions (//i.e.// ``operator++`` and ``operator*``)
are inline for efficiency, as is the ``NextValue`` function.

``myGetValue`` is a little messy because the value generated by the GMP function
``gmp_urandomm_ui`` cannot generate the full range of ``unsigned long`` values.
Instead I have to call ``gmp_urandomb_ui`` if the full range is needed.

The data members ``myLwb``, ``myUpb`` and ``myRange`` are morally constant,
but I cannot make them ``const`` because I wanted to allow assignment of
``RandomSeqLong`` objects.



==== RandomSeqBool ====

The idea is very simple: use the pseudo-random number generator of GMP
to generate a random integer, and then give out the bits of this
integer one at at time; when the last bit has been given out, get a
new random integer from the GMP generator.  The random integer is kept
in the data member ``myBuffer``, and ``myBitIndex`` is used to read
the bits one at a time.

The condition for refilling ``myBuffer`` is when the index goes beyond
the number of bits held in ``myBuffer``.

``myFillBuffer`` also sets the data member ``myBitIndex`` to zero; it
seemed most sensible to do this here.

The function ``prob`` is nifty; if you think about it for a moment, it
is obviously right (and economical on random bits).  It would be
niftier still if the probability were specified as an ``unsigned long``
-- on a 64-bit machine this ought to be sufficient for almost
all purposes.  If the requested probability is of the form N/2^k for
some odd integer N, then the average number of bits consumed by
``prob`` is 2-2^(1-k), which always lies between 1 and 2.  If the
requested probability is 0 or 1, no bits are consumed.


== Bugs, shortcomings and other ideas ==
%======================================================================

The printing function gives only partial information; //e.g.// two ``RandomSource`` objects
with different internal states might be printed out identically.

The implementation simply calls the GMP pseudo-random generator; this generator
is deterministic (so always produces the same sequence), but if you change
versions of GMP, the sequence of generated values may change.  You will have
to read the GMP documentation to know more.

**Discarded idea:** have a ctor which takes a ref to a [[RandomSource]],
and which uses that to obtain randomness.  I discarded the idea because of
the risks of an **invisible external reference** (//e.g.// a dangling reference,
or problems in multithreaded code).  Instead of passing a reference to a
[[RandomSource]] to the ctor, you can use the [[RandomSource]] to create an initial
seed which is handed to the ctor -- this gives better separation.

Why can ``RandomSource`` be seeded with a ``BigInt`` but the others not?
Does anyone really care?

=== Doubts common to RandomSeqBigInt, RandomSeqBool, RandomSeqLong  ===

It might be neater to put ``++myCounter`` inside ``myGenValue``, though
this would mean that ``myCounter`` gets incremented inside the ctor.

Should ``NextValue`` advance **before** or **after** getting the value?

Is the information printed by ``myOutputSelf`` adequate?  Time will tell.


Is there a better way of writing the four ctors (for ``RandomSeqBigInt``)
without repeating many lines of essentially identical source code?

Are there too many inline fns?  Is run-time speed so important here?  How
many algorithms really consume millions of random bits/numbers?  Surely the
computations on the random values should exceed the cost of generating
them, shouldn't they?


== Main changes ==
%======================================================================

- **December 2012 (v0.9953)**
 - major rewriting: now all classes are defined in one single file
   ``random.[HC]``
 - some classes and functions have been renamed:
   ``RandomXXXXSource`` to ``RandomSeqXXXX``, and ``sample`` to ``NextValue``
 - documentation is unified in ``random.txt``
- **October 2012 (v0.9952)**
 - clarified doc; added comments about thread-safety.
- **February 2011 (v0.9949)**
 - removed ``RandomLong(src)`` (//i.e.// with no range)
 - added ``RandomBool()``, ``RandomLong(lo,hi)``, ``RandomBigInt(lo,hi)``
          (//i.e.// with no ``RandomSource``)
- **April 2011 (v0.9943)** first release
-

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
