Revision history:

	Sun Nov 18 2018
	- José Joaquín Atria contributed a change to base the module on
	  Moo instead of bless-this-hash. Thanks!
	- Stop using "use integer" and explicitly call int() where needed.
	- Bump up the required perl version (starting to base my modules
	  on version 5.16.1).
	- Turn README into With MetaCPAN replacing CPAN,
	  there is no reason to have a plain-text README anymore
	  (MetaCPAN doesn't treat the file as special).
	- Add a file, which is becoming a thing.

	Mon Sep 10 2012
	- Bareword -norequire as part of the "use parent"  line makes
	  perl sad. Put everthing in qw().
	- Added 'create_license => 1' to the Build.PL list of attributes.
	- Added 'dynamic_config => 0' to the Build.PL list of attributes.
	- Added LICENSE to the MANIFEST file.
	Fri Jun 07 2012
	- POD changes from Ben Bullock merged in.

	Fri May 25 2012
	- Huh. A tester didn't have in its BSD set-up. Weird.
	  Put 'parent' in the 'requires' block in Build.PL.

	Wed May 23 2012
	- Sigh. Typo-d the 'resources' keyword in Build.PL.

	Wed May 23 2012
	- Took a while to work out that 'provides', whatever its
	  purpose, does not tell Build to put the version number
	  of the child modules in the META files. You have to
	  put the $VERSION declaration in the child modules' code.

	Tue May 22 2012
	- Fixed a missing closing brace in the BUILD.PL 'meta_merge'
	  key. The github URL should show up now on CPAN.
	- Removed the PREREQUISITE section in the POD. Useful back
	  when CPAN was simpler, but it's unneccesary now.
	- Add the 'provides' key to BUILD.PL. Now PAUSE should
	  be able to figure out the versions of the child modules.

	Mon May 21 2012
	- Named the distribution Games-Mazes insteade of Games-Maze
	  in the Build.PL. Very stupid typo fixed.
	- Changed 'use base' in submodules to 'use parent'.
	- Missed changing the internal hash keys to bareword
	  format in the previous version. Caught them all now.

	Tue May 15 2012
	- The lc operator now complains about undef values in
	  modern (5.14 right now) perl. Made the undef checking
	  more robust.
	- Changed old style {'quoted'} hash keys to {barewords}.

	Sun May 6 2012
	- Some more documentation changes, including updating the
	  e-mail address and adding some Build.PL instructions in
	  the README file.

	Fri May 4 2012
	- Minimum perl requirement bumped up to 5.8.3, as module
	  support code begins to drop 5.6 and related constructs.
	- Updated the Build.PL file to meet current perl standards.

1.03	Wed Feb 21 2007
	- Some changes to the structure of the distribution directory
	  in order to keep up with the current state of perl.
	- Use Test::Simple for tests, Module::Build for buiding, and
	  add a META.yml file.

1.02	Mon Jun 03 19:29:53 2002
	- Documentation fixes as pointed out by David Hand.

1.01	Tue May 07 21:55:39 2002
	- Bug fixes for bugs found by Pete Stewart.  Having added the 'entry'
	  and 'exit' parameters, the case where those parameters aren't used
	  became broken.
	- Fix to two of the test files involving an unintialized variable.
	- Add a Northward entry to the cell with the entrance.
	  Hadn't done that before because the printing routine to_ascii()
	  didn't need it, but now that another module will be using this,
	  i need to knock down the wall on both the border and maze cells.
	- Added the unsolve() method.

1.0	Wed Apr 24 13:27:19 2002
	- Hexagonal Hexagon mazes
	- Big changes in the stringifying functions, mostly by deleting
	  code, and fixing up the border cells values.
	- The 'cols', 'rows', and 'lvls' parameters have been replaced by
	  a single 'dimensions' parameter.
	- New parameter 'start' tells the make() method where to start
	  the random walk from.
	- New parameter 'fn_choosedir' chooses which direction your random
	  walk takes while making the maze.
	- New parameters 'entry' and 'exit' let you place the beginning
	  and ending points.

0.9	Thu Oct 19 11:39:54 2000
	- The maze object is now created via a call to
	  Games::Maze->new() only.  The child classes ::Rect and
	  ::Hex are dealt with in new().  The user need not deal
	  with it at all.

0.8	Wed Apr 19 10:57:35 2000
	- Games::MazeFlat goes away.  There was no real reason
	  beyond sentimentality to keep it.
	- The direction values for all mazes are now the same.  South,
	  for example, is now 0x20 for all mazes, instead of being 0x08
	  for square mazes, and 0x10 for hexagon mazes.  This is an
	  obvious thing to do in retrospect, but way back when they were
	  separate programs with separate constants.
	- This lets me move the contants and some more functions
	  into the base class, which makes the reading of the hex
	  dump easier.

0.7	Tue Feb	8 16:51:03 2000
	- Many incremental changes, leading up to:
	  1) Hash-style parameters to new()
	  2) Inheritance in similarly-dimensioned classes.
	     We now have Games::Maze, and Games::MazeFlat.
	  3) Code re-organized as a result of inheritance -
	     there are a lot more hidden member functions
	     created from sniplets of code that cannot cross the
	     package barrier.
	  4) More efficient use of internal constants.  We
	     no longer have separate constants for the direction
	     and the wall in that direction.  Now one constant
	     handles both the wall and the direction we face.
	  5) "make test" has tests for the 48-bit rand() function.

0.01	Wed Apr 28 14:14:56 1999
	- original version; created by h2xs 1.18


The code started out as a fairly direct translation of the CDC Pascal
source, which in turn was translated from the original CDC Fortran
(well, MNF to be precise) program written back in 1978 or 1979. 

The Algorithm

The algorithm is simple. The maze is considered to be an array of cells,
each cell bounded by walls that are shared with its neighbor.  The
program starts with a random cell, and randomly 'walks' through a wall
(knocking it down in the process) into a neighboring enclosed cell. If
it reaches a dead end, it returns to a point where it can continue its
random walk. When it runs out of enclosed cells to break into, the
algorithm is done.


Since this process creates singly-connected mazes (that is, a maze where
there is only one path between any two points), the solving method is
straightforward.  Go forward until either you reach the exit, or until
you run into a wall.  If you have run into a wall, turn to the next
direction and walk forward.  Repeat.  Drop off or pick up path marks as
you exit one cube for another.

A 'hand-on-wall' method of solving (not included in the packages) that
you may have learned as a kid works, but only for the 2-dimensional
mazes.  Sadly, it will often wind up in an infinite loop when it is
extended to the three-dimensional case.  I suspect that for this solving
algorithm to work, it would have to chose its next move based not only
on the current wall, but the prior move's wall as well.  I haven't
bothered to test this, though.  Since the go-forward method works so well
with both 2- and 3-dimensional mazes, I haven't attempted to adapt the
hand-on-wall methd.

Representing Hexagonal Mazes

Hexagonal mazes, as you may notice if you study the code, are simply
rectangular mazes with a 'hexagonal drift'.  That is,

      hexagonal mesh                 square grid with 'hexagonal drift'

  1  2  3  4  5  6  7  8             1   2   3   4   5   6   7   8
  __    __    __    __              ___     ___     ___     ___    
 /  \__/  \__/  \__/  \__          |   |___|   |___|   |___|   |___
 \__/  \__/  \__/  \__/  \         |___|   |___|   |___|   |___|   |
 /  \__/  \__/  \__/  \__/         |   |___|   |___|   |___|   |___|
 \__/  \__/  \__/  \__/  \         |___|   |___|   |___|   |___|   |
 /  \__/  \__/  \__/  \__/         |   |___|   |___|   |___|   |___|
 \__/  \__/  \__/  \__/  \         |___|   |___|   |___|   |___|   |
 /  \__/  \__/  \__/  \__/         |   |___|   |___|   |___|   |___|
 \__/  \__/  \__/  \__/  \         |___|   |___|   |___|   |___|   |
    \__/  \__/  \__/  \__/             |___|   |___|   |___|   |___|

Moving the squares back in position gives us a basic 8x4 matrix, which
is easy to represent in perl code.  This leaves only the question of
which six of the eight cells we can move to after choosing one of the
six possible directions to go.  There are two possible choices:

      _\|/_  or   __|__
        |          /|\

Which one we use depends upon whether we are in an up-shifted column or

Incidently, these are also the moves allowed to the opposing Gold
General pieces in the game Shogi (a chess-like game from Japan).
Coincidence? Well, probably, but it is interesting to contemplate.