Skip to content

Latest commit

 

History

History
469 lines (358 loc) · 22.3 KB

README_zhcn.md

File metadata and controls

469 lines (358 loc) · 22.3 KB

C++ 软件实践——扫雷

鲍一心 3014216002 计算机科学与技术 2014 级 1 班

问题描述

​ 实现扫雷游戏,要求界面上的格子数从 20x20 到 50x50 可选。

问题分析

对问题的理解:

​ 在很多人眼里,扫雷「MineSweeper」是 Microsoft 为 Windows 写作的一个经典小游戏。其实扫雷的历史远比这个悠久,它始于 1960 年代,经过多年的演变,如今的 Windows 10 App Store 里面的 MineSweeper 已经进化到一个全新的形态。不过究其根源,MineSweeper 的规则永远如下:

  • 游戏界面由一系列排列号的方块组成,方块分为两种:地雷与不是地雷。不是地雷的方块的显示内容为其四面八方紧邻的方块中地雷的个数

  • 用户左键点击砖块,方块被翻面显示其内容

  • 用户右键点击砖块,方块被旗帜标记或者被问号标记

  • 如果内容是地雷的砖块被显示,则游戏失败

  • 如果用户在不踩到地雷的情况下将所有安全的砖块翻面,则用户胜利

    那么创作这个游戏的过程就可以分为 GUI 绘制逻辑实现 两部分。GUI 绘制 有很多方案,不一一赘述。逻辑实现上,就要考虑到埋雷、计算每个砖块的显示内容、给砖块翻面递归翻面等。

解决问题的方案:

​ 对于 GUI 绘制,虽然个人偏爱 Apple 的 Cocoa 与 Cocoa Touch,但考虑到跨平台问题,我选择了在跨平台 C++ GUI 领域中较为优秀和广为人知的 Qt。

​ 对于 逻辑实现,我有以下几点:

  • 程序由 C++11 实现。

  • 在大多数的 MineSweeper 实现里,用户第一次点击是绝不会点击到地雷的。Microsoft 对这条规则的解决方案是:如果用户第一次点击到了地雷,则将地雷的位置移动,相当于简化的二次埋雷。而我的处理方案是,用户点击第一次以后再进行埋雷,此时程序的 AI 进行埋雷时会故意避开用户第一次点击的这块砖块。

  • 在大多数的 MineSweeper 实现里,当用户点击到一个空白的砖块(周围没有地雷)时,其周围的砖块也会自动地被翻面。我将我的解决方案称为展开与递归展开。C++ 的指针类型对我的实现有很大帮助。具体实现方法请看后面的算法描述。

  • 判断游戏失败或者成功。我的解决方案是:成功就是用户将所有不是地雷的砖块都翻面。失败就是用户踩到了地雷。

数据模型

Pure Header Files:

HEADERS DESCRIPTION
HeaderWrapper.h include header files to make including in other files simpler
resourses.h macro defines of file paths of game resourses like sounds and pics

Model Classes:

CLASS DESCRIPTION
AI AI controls the logic of the game
Cell Cell is the model of the cells in the game
Preferences Preferences is the user's preferences

Controller Classes:

CLASS DESCRIPTION
UserInteractionHandler Handles and filters users' interactons with the interface and send them to the model AI.
SoundHandler Plays sound

View Classes:

CLASS DESCRIPTION
Cell Displays the UI of each cell in the game according to their content
GameField Displays a Grid of cells that allows users to interact with
AboutPage Displays the About Page. A simple one.
ADOBEOCRLabel Custom Label Class with special font of Adobe OCR A STD
DefeatedDialog Displays the Defeated Page
WinningDialog Displays the Winning Page
Preferences Displays the Settings page
StopWatch A stopwatch
MainWindow The whole window of the game

For the Cell Class

PROPERTIES DESCRIPTION
Coordinate coordinate A cell's coordinate in the grid
int content 0 - 8 means the numbers of mines around the cell. -1 means it's a mine itself and -2 means it hasn't been initialized by the AI yet.
bool isMine Indicates if the cell is a mine
Status status Status is an enum type which has virgin, flagged, questioned, revealed
Cell *North, *NorthWest, *NorthEast, *East, *SouthEast, *South, *SouthWest, *West Pointers pointed to the 8 cells around a cell
INSTANCE METHODS DESCRIPTION
void setStatus(Status) sets its status
CLASS METHODS DESCRIPTION
Cell(int atRow, int atCol) The Constructor
SIGNALS DESCRIPTION
void clicked(Cell *, QMouseEvent *) Emitted when the user left_clicks a cell
void clicked_double(Cell *, QMouseEvent *) Emitted when the user double_clicks a cell
void statusChanged() Emitted when a cell's status has chan
SLOTS DESCRIPTION
void refreshUI() Refreshes the cell's UI on receiving the statusChanged()

For the AI Class:

PROPERTIES DESCRIPTION
CellMatrix *cells The 2-D matrix of cells in the game
FlatList *mines The 1-D Array of mines in the game
QString time The time the user has spent on a game
Board board The game board indicating rowCount and colCount of the matrix of cells in the game
UserInteractionHandler *_interactionHandler The User Interaction Handler
SoundHandler *_soundHandler The Sound Handler
INSTANCE METHODS DESCRIPTION
CellMatrix *initCells(int rowCount=12, int colCount=12) Generates a matrix of cells, rowCount and colCount being 12 by default
void layMines(Cell *clickedCell) Lay mines on cells except for the cell that the user clicked for the first time
void countNeighbourMines(CellMatrix *cells) Counts neighboring mines and set the cells' content
void revealCell(Cell *clickedCell) Recursively reveals the clickedCell
void judge() Judges if the user has won the game
void bindCellsToInteractionHandler(Cell *, UserInteractionHandler *) Binding signals and slots
void bindInteractionHandlerToAI(UserInteractionHandler *, AI *) Binding signals and slots
void bindInteractionhandlerToSoundHandler(UserInteractionHandler *, SoundHandler *) Binding signals and slots
void bindAIToSoundHandler(AI *, SoundHandler *) Binding signals and slots
CLASS METHODS DESCRIPTION
AI() The Constructor
static AI &sharedInstance() Returns the reference of the AI Singleton
SIGNALS DESCRIPTION
void steppedOnAMine(Cell *) Emitted when the user has stepped on a mine
void gameInitialized() Emitted when the game has been initialized
void waitingForTheTime() Emitted when the AI needs the stop watch to tell it the time
void succeeded() Emitted when user has won t
void reloadGame() Emitted when the user has changed the preferences or clicked restart / new game
void newWindowPopped() Emitted when a new Window pops up above the main window
void topViewDismissed() Emitted when the main windows became the top window in the view hierachy
SLOTS DESCRIPTION
void leftClickACell(Cell *) Left clicks a cell on receiving left click interaction from the _userInteractionHandler
void rightClickACell(Cell *) Right clicks a cell on receiving right click interaction from the _userInteractionHandler
void pause() Pauses the game (the stopwatch) on receiving the SIGNAL newWindowPopped
void resume() Resumes the game (the stopwatch) on receiving the SIGNAL topViewDismissed
void receivedNewPreferences(Preferences *) Changes the preferences on receiving a Preference_Changed signal
void userClickedRestart() Restarts the game when the user clicks restart
void receivedTime(QString) Sets the time when received time from the stopwatch

算法描述

存取方块

如何存取我当前游戏局面里的所有 cells 呢?我试图构建一个 C++ 的二维数组,像 Swift 这样的语言里面二维数组直接这样写:[[Cell]],数组的数组。C++ 没有这样的语法,我写了一个这样的指针的指针:

//返回一个指针的指针的指针 :-(
//很明显这样的存储非常的难看,非常的粗鲁,非常的不优雅
Cell *** AI::initCellsFoo(int rowCount, int columnCount) {
    Cell ***cells = new Cell **[rowCount];
    for (int i = 0; i < rowCount; ++i) {
        cells[i] = new Cell *[columnCount];
        for (int j = 0; j < columnCount; ++j) {
            Cell *foo = new Cell(i, j);
            cells[i][j] = foo;
        }
    }
    return cells;
}

以上不优雅的实现不仅很难看,还存在很多问题。于是我使用了 Qt 自带类型 QList 来构建二维数组(矩阵)

#define CellMatrix QList<QList<Cell *> *>
CellMatrix *AI::initCells(int rowCount, int columnCount) {
  CellMatrix *cells = new CellMatrix;
  for (int i=0; i < rowCount+2; ++i) {
    cells->append((new FlatList));
    for (int j=0; j<columnCount+2; ++j) {
      Cell *foo = new Cell(i, j);
      cells->at(i)->append(foo);
    }
  }
  return cells;
}

好了这下优雅多了。

表示方块之间的关系

接下来,表示砖块与砖块之间的联系是至关重要的一步,这涉及到递归展开的实现方法。最开始我想到单纯通过坐标来检索 Cell 以及其周围的 Cells,比如这个被展开的 Cell 的坐标为 (3, 5),那么我们就要同时考察它周围的八个 Cells 的状态。通过坐标计算的话,可以得出周围八个 Cells 的坐标分别是:(2, 4)(2, 5)(2, 6)(3, 4)(3, 6)(4, 4)(4, 5)(4, 6)。我们可以看到这样的计算是非常的复杂的(容易弄错)。更别提比如左上角的 Cell 坐标为 (0, 0),那么它的周围并没有八个 Cells,而程序会试图访问 (-1, -1) 这种坐标上的 Cell,很明显 (-1, -1) 上并没有 Cell

于是我想到了用指针来存取一个 Cell 周围的八个 Cells。在生成 AI::sharedInstance().cells 这个矩阵时,我通过计算坐标一次性地将指针附上。当然仍然会遇到处在边角上的 Cell 该如何赋指针——毕竟它们的周围并没有八个 Cells。最初的做法是,无论如何直接把当前坐标的值赋给那个指针,如cell->North = cells->at(-1)->at(-1),因为我想到的是,如果这个值不存在,那么程序应该会直接给我返回 NULL,如果是 NULL 的话就非常好处理。结果这样做了之后发现程序并不能正确地返回 NULL,而是会直接崩溃。

那么给不存在的地方的指针赋上 NULL 行不通之后,我想到了另一个方法,在生成矩阵的时候,在用户自定义大小矩阵外面多加一圈。在赋指针的时候,只考虑圈内的这些 Cells,这样就可以无忧无虑地赋指针了,因为每个 Cell 的周围都有八个 Cells 了。如图:

(其中青色的方块即为了解决问题加入的「无用」砖块,可以看到这时就不用担心有些 Cells 周围没有八个 Cells 了。)

就是说 AI::sharedInstance().cells 中有了一圈「无用」的不能被显示出来的 Cells,那么我就遇到下一个问题:该如何遍历这个矩阵呢?

本来的做法是,面对一个 m * n 的矩阵,我们的遍历做法应该是一个如下的嵌套 for-loop

for (int i=0; i<m; ++i) {
  for (int j=0; j<n; ++j) {
    //cells.at(i).at(j) 就是位于 (i, j) 上的一个 cell
  }
}

可是显然加入了外面一圈「无用」的Cells 之后,我们不能这样做了,此时用户本来想生成一个 m * n 的矩阵,而我们实则生成了 (m+2) * (n+2) 的矩阵。为了在遍历时不遍历外面一圈的「无用」Cells,我的做法是这样的:

for (int i=1; i<m+1; ++i) {
  for (int j=1; j<n+1; ++j) {
    //cells.at(i).at(j) 就是位于 (i, j) 上的一个 cell
  }
}

这样的遍历就能保证我们访问到的是「有用」的 cell

展开与递归展开

这个问题处理完之后,那么就面临递归展开的问题了。我的递归展开算法如下:

//大致的伪代码
void revealACell(Cell *cell) {
  if cell is not initialized (cell->content == -2)
    layMines()
  if cell is already revealed and its content != 0
    if the number of flagged cells = cell->content
      for cell in cell->cellsAround() {
        revealACell(cell)
      }
}

可以看到,如果一个 cell 已经被展开,此时计算它周围被 flag 的个数,如果这个个数等于该 cell 的 content,则将其周围所有没被标记的 cell 自动展开(因为用户自认为把这个 cell 周围的地雷都标记完了,可以自动展开其它)。那么此时还是会遇到边角上的 cell 周围会遇到之前所生成的「无用」cell,我们很显然无法对这种「无用」cell 进行展开,这时候怎么做呢?

在上面的伪代码里面可以看见对于 Cell 类我有一个实例方法叫做 FlatList *cellsAround(),这个方法返回了一个 cell 周围的 cells 的一维数组。在这个方法里面我对「无用」cell 进行了过滤。如下:

//Iterate Pointers Around a Cell and save'em into a list
FlatList *Cell::cellsAround() {
  FlatList *foo = new FlatList;
  if (-2 == content) {
    return NULL;
  }
  
  //如果它的周围有没有被初始化即 content 仍然等于 -2 的 cell,那么这样的 cell 是无用的,不能加入 list
  if (North->content != -2) foo->append(North);
  if (NorthWest->content != -2) foo->append(NorthWest);
  if (NorthEast->content != -2) foo->append(NorthEast);
  if (South->content != -2) foo->append(South);
  if (SouthWest->content != -2) foo->append(SouthWest);
  if (SouthEast->content != -2) foo->append(SouthEast);
  if (West->content != -2) foo->append(West);
  if (East->content != -2) foo->append(East);
  return foo;
}

至于埋雷的算法就不用赘述了,随机数解决。

主要程序代码:

//cells 矩阵的生成算法
CellMatrix *AI::initCells(int rowCount, int columnCount) {
    CellMatrix *cells = new CellMatrix;
    for (int i=0; i < rowCount+2; ++i) {
        cells->append((new FlatList));
        for (int j=0; j<columnCount+2; ++j) {
            Cell *foo = new Cell(i, j);
            cells->at(i)->append(foo);
        }
    }

    for(int k = 1; k < rowCount + 1; ++k) {
        for (int l = 1; l < columnCount + 1; ++l) {
            auto foo = cells->at(k)->at(l);

            //Binds Signals and Slots
            bindCellsToInteractionHandler(foo, _interactionHandler);
            foo->NorthWest = cells->at(k-1)->at(l-1);
            foo->North = cells->at(k-1)->at(l);
            foo->NorthEast = cells->at(k-1)->at(l+1);
            foo->East = cells->at(k)->at(l+1);
            foo->SouthEast = cells->at(k+1)->at(l+1);
            foo->South = cells->at(k+1)->at(l);
            foo->SouthWest = cells->at(k+1)->at(l-1);
            foo->West = cells->at(k)->at(l-1);
        }
    }

    return cells;
}

//给 cells 矩阵埋雷的算法
void AI::layMines(Cell *cell) {
    int maxRow = cells->count() - 2;
    int maxCol = cells->at(0)->count() - 2;
    int foo = board.mineCount;
    while (foo != 0) {
        int randomRow = (qrand() % maxRow) + 1;
        int randomCol = (qrand() % maxCol) + 1;
        if ((randomRow != cell->coordinate.atRow) && (randomCol != cell->coordinate.atCol) && (cells->at(randomRow)->at(randomCol)->isMine != true)) {
            cells->at(randomRow)->at(randomCol)->isMine = true;
            mines->append(cells->at(randomRow)->at(randomCol));
            foo -= 1;
        }
    }
    countNeighbourMines(cells);
}

//给每个 cell 计算其 content 的算法
void AI::countNeighbourMines(CellMatrix *cells) {

    for (int i = 1; i < cells->count()-1; ++i) {
        for (int j = 1; j < cells->at(i)->count()-1; ++j) {
            auto foo = cells->at(i)->at(j);
            if (foo->isMine) {
                foo->content = -1;
                continue;
            }
            foo->content = 0;
            if (foo->cellsAround() != NULL) {
            	FlatList *fooList = foo->cellsAround();
                for (int i=0; i<fooList->count(); ++i) {
                	if (fooList->at(i)->isMine) {
                    	foo->content += 1;
            		}
            	}
            }
        }
    }
}

//revealCell 算法
void AI::revealCell(Cell *cell) {
    //foo here represents the number of flagged cells around the clicked cell
    int foo = 0;

    if (-2 == cell->content) {
        layMines(cell);
    }

    if (cell->status == revealed && cell->content != 0) {
        FlatList *fooListToReveal = new FlatList;
        if (cell->cellsAround() != NULL) {
            FlatList *fooList = cell->cellsAround();
            for (int i=0; i<fooList->count(); ++i) {
                if (fooList->at(i)->status == flagged) {
                    foo += 1;
                } else {
                    fooListToReveal->append(fooList->at(i));
                }
            }
        }

        if (foo == cell->content) {
            for (int i=0; i<fooListToReveal->count(); ++i) {
                if (fooListToReveal->at(i)->status != revealed && fooListToReveal->at(i)->content != -2) {
                    revealCell(fooListToReveal->at(i));
                }
            }
        }

        //judge the game to see if the user has won
        this->judge();

    } else {
        cell->setStatus(revealed);
        board.numberOfCellsRevealed += 1;
        switch (cell->content) {
        case -1: {
            for (int i=0; i<mines->count(); ++i) {
                if (cell != mines->at(i)) {
                    revealCell(mines->at(i));
                }
            }
            emit steppedOnAMine(cell);
            break;
        }

        default: {
            FlatList *fooListToReveal = new FlatList;
            if (cell->cellsAround() != NULL) {
                FlatList *fooList = cell->cellsAround();
                for (int i=0; i<fooList->count(); ++i) {
                    if (fooList->at(i)->status == flagged) {
                        foo += 1;
                    } else {
                        fooListToReveal->append(fooList->at(i));
                    }
                }
            }

            if (foo == cell->content) {
                for (int i=0; i<fooListToReveal->count(); ++i) {
                    if (fooListToReveal->at(i)->status != revealed && fooListToReveal->at(i)->content != -2) {
                        revealCell(fooListToReveal->at(i));
                    }
                }
            }
            break;
        }
        }

        this->judge();
    }
}

//judge 算法
void AI::judge() {
    if (board.numberOfCellsRevealed == board.normalCellCount) {
      	//ready to receive the time from the stopwatch to show results
        emit waitingForTheTime();
    }
}

程序流程图:

用户交互处理:

AI 递归展开的逻辑:

实习心得

这是我第一次使用 C++ 来写 GUI 程序。已有的经验是 Swift 和 Objective-C 这样的拥有十分优雅的 GUI 库的语言。C++ 因为相对底层,让我在书写过程中感到一些原始野蛮的感觉。

当然 C++ 是非常棒的语言,能实现各种各样的编程范式,在写这个扫雷的时候,我非常想念 Swift 和 Haskell 里面的函数式编程特性,于是我尝试实现了 C++ 里面的 Lambda。虽然没有实际地去用(因为实在没时间去调……)。

这一次的写作过程,我写到一半意识到自己事先构思的 M-V-C 结构我没有完全实现,图一时简单将一些东西混在了一起,比如 Cell 这个类是 Model 和 View 的混合。但是后来想一想,其实这种游戏类别的程序并不一定要遵循这样的结构。

写每个类的时候,都要考虑到日后的可扩展性。各个模块之间的耦合度都应极力讲到最低甚至完全解耦,方便模块的热拔插。

对于递归算法中的每次遍历,可能会很粗鲁,虽然就算局面很大,也就几十万的量,但是应该找寻一个剪枝策略来优化过程。