Well… all the tests except the Peter Ellis Jones tests.
Pass around whole EnPassant types instead of pulling out just the e.p. square.
Make sure that Castling moves have their target and origin squares populated.
Add a color field to the Castle move style to make this possible.
Replace Position's en_passant_target_square() and en_passant_capture_square()
with a single en_passant() method that returns a new EnPassant struct that has
the target and capture squares for the en passant move.
Build Position::sight_of_piece() and ::is_king_in_check() for cfg(test) only.
Expand the doc comment for ::king_danger() slightly.
Remove an unused Rank import.
# Conflicts:
# position/src/position/position.rs
Use u64::MIN and u64::MAX to define the empty and full bitboards
Write From<Square> as `1u64 << sq as u32`
Write a doc test for BitBoard::is_single_square()
Make library::RANKS and library::FILES pub(crate) instead of pub(super)
The pawn move generator only generated pushes for white pawns.
The is_king_in_check returned an inverted flag.
MoveSet needed a couple more validation methods: can_move_to_square and can_castle.
The MakeMoveBuilder also needed a little more move validation using the above methods.
Use CheckingPieces to determine which and how many pieces check the king.
- If there are no checks, proceed with move generation as normal.
- If there is one checking piece, calculate push and capture masks and use those
to generate legal moves.
- If there are more than one checking pieces, the only legal moves are king
moves. Indicate this by setting the push and capture masks to EMPTY.
Add an origin square argument to its one method ::ray_to_square(). Pushing this
implementation down a layer means I don't have to care about the color of the
piece.
Position::default specifies the defaults for all field. Then, ::new() and
::starting() can use ..Default::default() in their implementations to avoid
having to specify empty values for all the internal structures.
