Learning Haskell with Chess
This page is about learning Haskell using the board game Chess as a running example. The complete code can be found at http://www.steffen-mazanek.de/dateien/projekte/hsChess.zip.
@German Haskellers: You may also have a look at the tex-files used for our student exercises, http://www.steffen-mazanek.de/dateien/projekte/hsChess-teaching-german.zip.
Exercise 1 - data types
- recapitulate Haskell types (keywords
data, product and sum types)
- Helium: equality and show functions (pattern matching)
- Haskell: type classes (
- list handling (boards will be represented by lists of lists)
- special character
- Define data types that represent boards (
Board), squares (
Square), positions (
Pos), pieces (
Piece, supported by
PieceType) and game states (
- Helium: Implement suited eq and show functions.
- Haskell: Define/derive instances of
- Implement a function
prettyBoard::Board->String, that transforms a board into a clearly arranged string representation (human readable :-)). Support this function with auxiliary functions that pretty print pieces, squares, ...
- Define the initial board (
- Implement a simple evaluation function
evalBoard::Board->Intas the difference of material on board, for this purpose define a function
valuePiecethat maps pieces to their values (pawn->1, knight and bishop->3, queen->9, rook->5, king->"infinity"=1000).
Exercise 2 - move generator
- list comprehension
- stepwise refinement
- Define a function
movePos p1 p2 bmoves the piece at p1 to p2 (overwrite the square at p2 with the square at p1 and delete p1). Do not check whether the move is valid at this stage. Hint: Define a function
Exercise 3 - gametree generation and minimax algorithm
- break code in modules
- recursive data structures -> recursive algorithms
- Define a data type that represents a game tree (
- Roughly estimate the number of nodes of the gametree with depth 4.
- Define a function
play::Gametree->Int, that computes the value of a given game tree using the minimax Algorithm.
- Implement the function
doMove::State->State, that choses the (best) next state.