The en passant square is only important if it is
target of an enemy pawn.
If that is not the case it should not be set
and shouldn't appear in the fen string either.
Now it is checked if the en passant square is
the target of an enemy pawn
Repetitions can only occur if the pieces and castling
rights are exactly the same.
For memory sake the sequence of last boards
is now emptied if a capture occurs,
a pawn moves, or if the castle
rights change.
That directly makes the saving of the castling
rights obsolete.
Now there is a method to turn a chess
object into a fen-string.
Additionally the enPassant handling was
changed.
There can be only one enPassant field
at a time.
So a new attribute was added to the chess
object that stores the field for the
enPassant capture.
If there is no enPassant field it will be set to -1.
If the threefold repitition check is called before enough moves
are saved in the previous board state it throws a nullpointer.
A simple null check solves the problem
I created an engine, that uses a basic version of the minimax algorithm
to evaluate a position.
It then evaluates all the possible next moves in a given position and picks
the one that has the best evaluation.
Basic test cases were also added.
Moved was a 10x12 1-d array, that stored bools.
It was only used for checking if a pawn moved (especially important for
the double move of the pawns) and if the other pieces moved (important
for checking if the player can castle or not).
The pawn moves are now checked by the starting position (pawns on the
second rank cant be already moved).
The castle rights of a game are expressed as CastleRights, a tuple with 4 bools.
This saves basically 116 boolean values.
