{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 扫雷机器人\n",
    "\n",
    "项目原作者 [ArtrixTech](https://github.com/ArtrixTech), 项目地址 [BoomMine](https://github.com/ArtrixTech/BoomMine), 原文档发表在知乎 [@Artrix](https://zhuanlan.zhihu.com/p/35755039), 助教进行了文档的改进，和代码形式上的交互式重构。\n",
    "\n",
    "这个笔记本主要是为了让大家了解一个小型的 Python + OpenCV 项目是如何完成的，以便于以后进行更复杂的软件工程项目。\n",
    "\n",
    "通过对比 notebook 和 GitHub 上面的项目，大家也可以知道如何从交互式编程练习过渡到实际的 Python 脚本开发。\n",
    "\n",
    "## 热身环节\n",
    "\n",
    "一直对着命令行编程也累了，我们下载扫雷游戏 [Minesweeper Arbiter](http://saolei.net/Download/Arbiter_0.52.3.zip) 来玩一玩吧。\n",
    "\n",
    "如果你从前没有玩过扫雷，可以看看国内最大的扫雷爱好者论坛[扫雷网](http://saolei.net/Main/Index.asp)的基础入门教程：\n",
    "\n",
    "- [扫雷新手上路](http://saolei.net/BBS/Title.asp?Id=177)　作者: 门世运\n",
    "- [扫雷术语介绍](http://saolei.net/BBS/Title.asp?Id=227)　作者: 张砷镓\n",
    "- [扫雷定式及其变化](http://saolei.net/BBS/Title.asp?Id=362)　作者: 张砷镓\n",
    "- [关于数字1-7周围8格中雷的分布的各种形状](http://saolei.net/BBS/Title.asp?Id=8243)　作者: 杨萧杨\n",
    "- [扫雷游戏的起源](http://saolei.net/BBS/Title.asp?Id=1005)　作者: 张砷镓\n",
    "\n",
    "简单玩几局后，就可以考虑如何设计一个扫雷机器人，来帮助我们完成这个游戏啦~\n",
    "\n",
    "## 实现思路 \n",
    "\n",
    "在去做一件事情之前最重要的是什么？是将要做的这件事情在心中搭建一个步骤框架。只有这样，才能保证在去做这件事的过程中，尽可能的做到深思熟虑，使得最终有个好的结果。我们写程序也要尽可能做到在正式开始开发之前，在心中有个大致的思路。\n",
    "\n",
    "对于本项目而言，大致的开发过程是这样的：\n",
    "\n",
    "- 完成窗体内容截取部分\n",
    "- 完成雷块分割部分\n",
    "- 完成雷块类型识别部分\n",
    "- 完成扫雷算法\n",
    "\n",
    "好啦，既然我们有了个思路，那就撸起袖子大力干！\n",
    "\n",
    "### 窗体截取\n",
    "\n",
    "其实对于本项目而言，窗体截取是一个逻辑上简单，实现起来却相当麻烦的部分，而且还是必不可少的部分。笔者通过 [Spy++](https://docs.microsoft.com/en-us/visualstudio/debugger/how-to-start-spy-increment) 得到了以下两点信息："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class_name = \"TMain\"                  # ms_arbiter.exe的主窗体类别\n",
    "title_name = \"Minesweeper Arbiter\"    # ms_arbiter.exe的主窗体名称"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "注意到了么？主窗体的名称后面有个空格。正是这个空格让笔者困扰了一会儿，只有加上这个空格，win32gui 才能够正常的获取到窗体的句柄。(提示：在助教的机器上不加空格反而有效)\n",
    "\n",
    "运行 `ms_arbiter.exe`, 本项目采用了 win32gui 来获取窗体的位置信息，具体代码如下："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import win32gui\n",
    "\n",
    "class_name = \"TMain\"                  # ms_arbiter.exe的主窗体类别\n",
    "title_name = \"Minesweeper Arbiter\"    # ms_arbiter.exe的主窗体名称\n",
    "\n",
    "hwnd = win32gui.FindWindow(class_name, title_name)            # Handle to A Window\n",
    "if hwnd:\n",
    "    left, top, right, bottom = win32gui.GetWindowRect(hwnd)\n",
    "    print(\"Window find.\")\n",
    "    print(\"left: \" + str(left))\n",
    "    print(\"top: \" + str(top))\n",
    "    print(\"right: \" + str(right))\n",
    "    print(\"bottom: \" + str(bottom))\n",
    "else:\n",
    "    print(\"Window not find.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "你可以移动窗体的位置，重复运行上面的代码，看得到的位置数据是否有变化。\n",
    "\n",
    "通过以上代码，我们得到了窗体相对于整块屏幕的位置。\n",
    "\n",
    "之后我们需要通过 PIL 来进行扫雷界面的棋盘截取："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from PIL import ImageGrab\n",
    "\n",
    "# 微调参数，得到棋盘的相对位置，仅在我的电脑环境下测试通过\n",
    "# 这一步如果需要重复运行，需要把上面的代码也一起重复运行 (参数更新)\n",
    "left += 15\n",
    "top += 111\n",
    "right -= 15\n",
    "bottom -= 31\n",
    "\n",
    "# 棋盘截取\n",
    "rect = (left, top, right, bottom)\n",
    "img = ImageGrab.grab().crop(rect)             # 扫雷界面不能被遮挡\n",
    "img.show()  # 通过快照的位置调整上面的四个参数，实际调用中没有这一行\n",
    "\n",
    "# 得到棋盘的宽度和高度\n",
    "width = right - left\n",
    "height = bottom - top"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "其实有着更好的寻找扫雷棋盘相对位置的方法，但是这里希望大家进行人为参数寻找。\n",
    "\n",
    "这种无法由算法给出，需要人类给定的参数叫作“超参数”，这一概念后面还会见到。\n",
    "\n",
    "### 雷块分割\n",
    "\n",
    "在进行雷块分割之前，我们事先需要了解雷块的尺寸以及它的边框大小。经过笔者的测量，在 `ms_arbiter` 下，每一个雷块的尺寸为 16px*16px."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "block_width, block_height = 16, 16\n",
    "blocks_x = int((right - left) / block_width)      # 宽度上块的个数，高级为 30\n",
    "blocks_y = int((bottom - top) / block_height)     # 高度上块的个数，高级为 15\n",
    "print(str(blocks_x) + \",\" + str(blocks_y))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "之后，我们建立一个二维数组用于存储每一个雷块的图像，并且进行图像分割，保存在之前建立的数组中。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def crop_block(hole_img, x, y):\n",
    "    # 基于棋盘进行单个块的截取\n",
    "    x1, y1 = x * block_width, y * block_height\n",
    "    x2, y2 = x1 + block_width, y1 + block_height\n",
    "    return hole_img.crop((x1, y1, x2, y2))\n",
    "\n",
    "blocks_img = [[0 for i in range(blocks_y)] for i in range(blocks_x)]\n",
    "    \n",
    "for y in range(blocks_y):\n",
    "    for x in range(blocks_x):\n",
    "        blocks_img[x][y] = crop_block(img, x, y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "将整个图像获取、分割的部分封装成一个库，随时调用就OK啦~在笔者的实现中，笔者将这一部分封装成了`imageProcess.py`, 其中函数 `get_frame()` 用于完成上述的图像获取、分割过程："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import win32gui\n",
    "import numpy\n",
    "from PIL import ImageGrab\n",
    "import cv2\n",
    "\n",
    "block_width, block_height = 16, 16\n",
    "\n",
    "def crop_block(hole_img, x, y):\n",
    "    x1, y1 = x * block_width, y * block_height\n",
    "    x2, y2 = x1 + block_width, y1 + block_height\n",
    "    return hole_img.crop((x1, y1, x2, y2))\n",
    "\n",
    "def pil_to_cv(img):\n",
    "    return cv2.cvtColor(numpy.asarray(img), cv2.COLOR_RGB2BGR)\n",
    "\n",
    "def get_frame():\n",
    "    class_name = \"TMain\"\n",
    "    title_name = \"Minesweeper Arbiter\"\n",
    "\n",
    "    hwnd = win32gui.FindWindow(class_name, title_name)\n",
    "    if hwnd:\n",
    "        left, top, right, bottom = win32gui.GetWindowRect(hwnd)\n",
    "\n",
    "        left += 15\n",
    "        top += 111\n",
    "        right -= 15\n",
    "        bottom -= 31\n",
    "\n",
    "        width = right - left\n",
    "        height = bottom - top\n",
    "\n",
    "        rect = (left, top, right, bottom)\n",
    "        img = ImageGrab.grab().crop(rect)\n",
    "\n",
    "        blocks_x = int((right - left) / block_width)\n",
    "        blocks_y = int((bottom - top) / block_height)\n",
    "\n",
    "        blocks_img = [[0 for i in range(blocks_y)] for i in range(blocks_x)]\n",
    "\n",
    "        for y in range(blocks_y):\n",
    "            for x in range(blocks_x):\n",
    "                blocks_img[x][y] = crop_block(img, x, y)\n",
    "\n",
    "        return img, blocks_img, (blocks_x, blocks_y), (width, height), (left, top, right, bottom)\n",
    "\n",
    "    return -1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "img = get_frame()          # 单元测试该方法\n",
    "# img[0].show()            # 显示整个棋盘\n",
    "img[1][0][0].show()      # 显示指定的 [1][x][y] 块，可人为点击改变块的状态 "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "同样地，我们这里采用了笨笨的分割方法，如果学习过特征提取，我们能够使用更好的方法进行图像分割。在本节练习中，我们将利用图像颜色特征的知识对雷块的属性进行识别。\n",
    "\n",
    "我们现在来设计一个类(Class)，下面是已经写好的源码，你能不能试着直接读懂它们：\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class BoomMine:\n",
    "    \n",
    "    __inited = False\n",
    "    blocks_x, blocks_y = -1, -1\n",
    "    width, height = -1, -1\n",
    "    img_cv, img = -1, -1\n",
    "    blocks_img = [[-1 for i in range(blocks_y)] for i in range(blocks_x)]\n",
    "    blocks_num = [[-3 for i in range(blocks_y)] for i in range(blocks_x)]\n",
    "    blocks_is_mine = [[0 for i in range(blocks_y)] for i in range(blocks_x)]\n",
    "\n",
    "    is_new_start = True\n",
    "\n",
    "    next_steps = []\n",
    "\n",
    "    @staticmethod                         # Python 内置函数 staticmethod() \n",
    "    def rgb_to_bgr(rgb):\n",
    "        return rgb[2], rgb[1], rgb[0]\n",
    "\n",
    "    @staticmethod                         # 该方法不强制要求传递参数，无需实例化\n",
    "    def equal(arr1, arr2):\n",
    "        if arr1[0] == arr2[0] and arr1[1] == arr2[1] and arr1[2] == arr2[2]:\n",
    "            return True\n",
    "        return False\n",
    "\n",
    "    def is_in_form(self, location):\n",
    "        x, y = location[0], location[1]\n",
    "        if x < self.left or x > self.right or y < self.top or y > self.bottom:\n",
    "            return False\n",
    "        return True\n",
    "\n",
    "    def iterate_blocks_image(self, func):\n",
    "        if self.__inited:\n",
    "            for y in range(self.blocks_y):\n",
    "                for x in range(self.blocks_x):\n",
    "                    # args are: self, [0]singleBlockImage, [1]location(as an array)\n",
    "                    func(self, self.blocks_img[x][y], (x, y))\n",
    "\n",
    "    def iterate_blocks_number(self, func):\n",
    "        if self.__inited:\n",
    "            for y in range(self.blocks_y):\n",
    "                for x in range(self.blocks_x):\n",
    "                    # args are: self, [0]singleBlockNumber, [1]location(as an array)\n",
    "                    func(self, self.blocks_num[x][y], (x, y))\n",
    "\n",
    "    def analyze_block(self, block, location):\n",
    "        x, y = location[0], location[1]\n",
    "        now_num = self.blocks_num[x][y]\n",
    "\n",
    "        # if 1:\n",
    "        if not now_num == -2 and not 0 < now_num < 9:\n",
    "\n",
    "            block = imageProcess.pil_to_cv(block)\n",
    "            block_color = block[8, 8]\n",
    "\n",
    "            # -1:Not opened\n",
    "            # -2:Opened but blank\n",
    "            # -3:Un initialized\n",
    "\n",
    "            # Opened\n",
    "            if self.equal(block_color, self.rgb_to_bgr((192, 192, 192))):\n",
    "                if not self.equal(block[8, 1], self.rgb_to_bgr((255, 255, 255))):\n",
    "                    self.blocks_num[x][y] = -2\n",
    "                    self.is_started = True\n",
    "                else:\n",
    "                    self.blocks_num[x][y] = -1\n",
    "\n",
    "            elif self.equal(block_color, self.rgb_to_bgr((0, 0, 255))):\n",
    "                self.blocks_num[x][y] = 1\n",
    "\n",
    "            elif self.equal(block_color, self.rgb_to_bgr((0, 128, 0))):\n",
    "                self.blocks_num[x][y] = 2\n",
    "\n",
    "            elif self.equal(block_color, self.rgb_to_bgr((255, 0, 0))):\n",
    "                self.blocks_num[x][y] = 3\n",
    "\n",
    "            elif self.equal(block_color, self.rgb_to_bgr((0, 0, 128))):\n",
    "                self.blocks_num[x][y] = 4\n",
    "\n",
    "            elif self.equal(block_color, self.rgb_to_bgr((128, 0, 0))):\n",
    "                self.blocks_num[x][y] = 5\n",
    "\n",
    "            elif self.equal(block_color, self.rgb_to_bgr((0, 128, 128))):\n",
    "                self.blocks_num[x][y] = 6\n",
    "\n",
    "            elif self.equal(block_color, self.rgb_to_bgr((0, 0, 0))):\n",
    "                if self.equal(block[6, 6], self.rgb_to_bgr((255, 255, 255))):\n",
    "                    # Is mine\n",
    "                    self.blocks_num[x][y] = 9\n",
    "                elif self.equal(block[5, 8], self.rgb_to_bgr((255, 0, 0))):\n",
    "                    # Is flag\n",
    "                    self.blocks_num[x][y] = 0\n",
    "                else:\n",
    "                    self.blocks_num[x][y] = 7\n",
    "\n",
    "            elif self.equal(block_color, self.rgb_to_bgr((128, 128, 128))):\n",
    "                self.blocks_num[x][y] = 8\n",
    "            else:\n",
    "                self.blocks_num[x][y] = -3\n",
    "                self.is_mine_form = False\n",
    "\n",
    "            if self.blocks_num[x][y] == -3 or not self.blocks_num[x][y] == -1:\n",
    "                self.is_new_start = False\n",
    "\n",
    "    def detect_mine(self, block, location):\n",
    "\n",
    "        def generate_kernel(k, k_width, k_height, block_location):\n",
    "            ls = []\n",
    "            loc_x, loc_y = block_location[0], block_location[1]\n",
    "            for now_y in range(k_height):\n",
    "                for now_x in range(k_width):\n",
    "\n",
    "                    if k[now_y][now_x]:\n",
    "                        rel_x, rel_y = now_x - 1, now_y - 1\n",
    "                        ls.append((loc_y + rel_y, loc_x + rel_x))\n",
    "            return ls\n",
    "\n",
    "        def count_unopen_blocks(blocks):\n",
    "            count = 0\n",
    "            for single_block in blocks:\n",
    "                if self.blocks_num[single_block[1]][single_block[0]] == -1:\n",
    "                    count += 1\n",
    "            return count\n",
    "\n",
    "        def mark_as_mine(blocks):\n",
    "            for single_block in blocks:\n",
    "                if self.blocks_num[single_block[1]][single_block[0]] == -1:\n",
    "                    self.blocks_is_mine[single_block[1]][single_block[0]] = 1\n",
    "\n",
    "        x, y = location[0], location[1]\n",
    "\n",
    "        if self.blocks_num[x][y] > 0:\n",
    "\n",
    "            kernel_width, kernel_height = 3, 3\n",
    "\n",
    "            # Kernel mode:[Row][Col]\n",
    "            kernel = [[1, 1, 1], [1, 1, 1], [1, 1, 1]]\n",
    "\n",
    "            # Left border\n",
    "            if x == 0:\n",
    "                for i in range(kernel_height):\n",
    "                    kernel[i][0] = 0\n",
    "\n",
    "            # Right border\n",
    "            if x == self.blocks_x - 1:\n",
    "                for i in range(kernel_height):\n",
    "                    kernel[i][kernel_width - 1] = 0\n",
    "\n",
    "            # Top border\n",
    "            if y == 0:\n",
    "                for i in range(kernel_width):\n",
    "                    kernel[0][i] = 0\n",
    "\n",
    "            # Bottom border\n",
    "            if y == self.blocks_y - 1:\n",
    "                for i in range(kernel_width):\n",
    "                    kernel[kernel_height - 1][i] = 0\n",
    "\n",
    "            # Generate the search map\n",
    "            to_visit = generate_kernel(kernel, kernel_width, kernel_height, location)\n",
    "\n",
    "            unopen_blocks = count_unopen_blocks(to_visit)\n",
    "            if unopen_blocks == self.blocks_num[x][y]:\n",
    "                mark_as_mine(to_visit)\n",
    "\n",
    "    def detect_to_click_block(self, block, location):\n",
    "\n",
    "        def generate_kernel(k, k_width, k_height, block_location):\n",
    "            ls = []\n",
    "            loc_x, loc_y = block_location[0], block_location[1]\n",
    "            for now_y in range(k_height):\n",
    "                for now_x in range(k_width):\n",
    "\n",
    "                    if k[now_y][now_x]:\n",
    "                        rel_x, rel_y = now_x - 1, now_y - 1\n",
    "                        ls.append((loc_y + rel_y, loc_x + rel_x))\n",
    "            return ls\n",
    "\n",
    "        def count_mines(blocks):\n",
    "            count = 0\n",
    "            for single_block in blocks:\n",
    "                if self.blocks_is_mine[single_block[1]][single_block[0]] == 1:\n",
    "                    count += 1\n",
    "            return count\n",
    "\n",
    "        def mark_to_click_block(blocks):\n",
    "            for single_block in blocks:\n",
    "\n",
    "                # Not Mine\n",
    "                if not self.blocks_is_mine[single_block[1]][single_block[0]] == 1:\n",
    "\n",
    "                    # Click-able\n",
    "                    if self.blocks_num[single_block[1]][single_block[0]] == -1:\n",
    "\n",
    "                        # Source Syntax: [y][x] - Converted\n",
    "                        if not (single_block[1], single_block[0]) in self.next_steps:\n",
    "                            self.next_steps.append((single_block[1], single_block[0]))\n",
    "\n",
    "        x, y = location[0], location[1]\n",
    "\n",
    "        if block > 0:\n",
    "\n",
    "            kernel_width, kernel_height = 3, 3\n",
    "\n",
    "            # Kernel mode:[Row][Col]\n",
    "            kernel = [[1, 1, 1], [1, 1, 1], [1, 1, 1]]\n",
    "\n",
    "            # Left border\n",
    "            if x == 0:\n",
    "                for i in range(kernel_height):\n",
    "                    kernel[i][0] = 0\n",
    "\n",
    "            # Right border\n",
    "            if x == self.blocks_x - 1:\n",
    "                for i in range(kernel_height):\n",
    "                    kernel[i][kernel_width - 1] = 0\n",
    "\n",
    "            # Top border\n",
    "            if y == 0:\n",
    "                for i in range(kernel_width):\n",
    "                    kernel[0][i] = 0\n",
    "\n",
    "            # Bottom border\n",
    "            if y == self.blocks_y - 1:\n",
    "                for i in range(kernel_width):\n",
    "                    kernel[kernel_height - 1][i] = 0\n",
    "\n",
    "            # Generate the search map\n",
    "            to_visit = generate_kernel(kernel, kernel_width, kernel_height, location)\n",
    "\n",
    "            mines_count = count_mines(to_visit)\n",
    "\n",
    "            if mines_count == block:\n",
    "                mark_to_click_block(to_visit)\n",
    "\n",
    "    def rel_loc_to_real(self, block_rel_location):\n",
    "        return self.left + 16 * block_rel_location[0] + 8, self.top + 16 * block_rel_location[1] + 8\n",
    "\n",
    "    def __init__(self):\n",
    "        self.next_steps = []\n",
    "        self.left = 0\n",
    "        self.top = 0\n",
    "        self.right = 0\n",
    "        self.bottom = 0\n",
    "        self.continue_random_click = False\n",
    "        self.is_mine_form = True\n",
    "        self.is_started = False\n",
    "        self.have_solve = False\n",
    "        if self.process_once():\n",
    "            self.__inited = True\n",
    "\n",
    "    def show_map(self):\n",
    "        if self.__inited:\n",
    "            for y in range(self.blocks_y):\n",
    "                line = \"\"\n",
    "                for x in range(self.blocks_x):\n",
    "                    if self.blocks_num[x][y] == -1:\n",
    "                        line += \"  \"\n",
    "                    else:\n",
    "                        line += str(self.blocks_num[x][y]) + \" \"\n",
    "                print(line)\n",
    "\n",
    "    def show_mine(self):\n",
    "        if self.__inited:\n",
    "            for y in range(self.blocks_y):\n",
    "                line = \"\"\n",
    "                for x in range(self.blocks_x):\n",
    "                    if self.blocks_is_mine[x][y] == 0:\n",
    "                        line += \"  \"\n",
    "                    else:\n",
    "                        line += str(self.blocks_is_mine[x][y]) + \" \"\n",
    "                print(line)\n",
    "\n",
    "    def process_once(self):\n",
    "\n",
    "        # Initialize\n",
    "        result = imageProcess.get_frame()\n",
    "        if result == -1:\n",
    "            print(\"Minesweeper Arbiter Window Not Detected!\")\n",
    "            return False\n",
    "        self.img, self.blocks_img, form_size, img_size, form_location = result\n",
    "\n",
    "        self.blocks_num = [[-1 for i in range(self.blocks_y)] for i in range(self.blocks_x)]\n",
    "        self.blocks_is_mine = [[0 for i in range(self.blocks_y)] for i in range(self.blocks_x)]\n",
    "        self.next_steps = []\n",
    "        self.is_new_start = True\n",
    "        self.is_mine_form = True\n",
    "\n",
    "        self.blocks_x, self.blocks_y = form_size[0], form_size[1]\n",
    "        self.width, self.height = img_size[0], img_size[1]\n",
    "        self.img_cv = imageProcess.pil_to_cv(self.img)\n",
    "        self.left, self.top, self.right, self.bottom = form_location\n",
    "\n",
    "        # Analyze the number of blocks\n",
    "        self.iterate_blocks_image(BoomMine.analyze_block)\n",
    "\n",
    "        # Mark all mines\n",
    "        self.iterate_blocks_number(BoomMine.detect_mine)\n",
    "\n",
    "        # Calculate where to click\n",
    "        self.iterate_blocks_number(BoomMine.detect_to_click_block)\n",
    "\n",
    "        self.have_solve = False\n",
    "        if len(self.next_steps) > 0:\n",
    "            self.have_solve = True\n",
    "\n",
    "        if self.is_in_form(mouseOperation.get_mouse_point()):\n",
    "\n",
    "            for to_click in self.next_steps:\n",
    "                on_screen_location = self.rel_loc_to_real(to_click)\n",
    "                mouseOperation.mouse_move(on_screen_location[0], on_screen_location[1])\n",
    "                mouseOperation.mouse_click()\n",
    "\n",
    "                # If your computer's performance is not high, enable this:\n",
    "                # time.sleep(0.001)\n",
    "\n",
    "        if not self.have_solve and self.is_mine_form:\n",
    "\n",
    "            rand_location = (random.randint(0, self.blocks_x - 1), random.randint(0, self.blocks_y - 1))\n",
    "            rand_x, rand_y = rand_location[0], rand_location[1]\n",
    "            iter_times = 0\n",
    "\n",
    "            if len(self.blocks_is_mine) > 0:\n",
    "\n",
    "                while self.blocks_is_mine[rand_x][rand_y] or not self.blocks_num[rand_x][rand_y] == -1 and iter_times < 50:\n",
    "                    rand_location = (random.randint(0, self.blocks_x - 1), random.randint(0, self.blocks_y - 1))\n",
    "                    rand_x, rand_y = rand_location[0], rand_location[1]\n",
    "                    iter_times += 1\n",
    "\n",
    "            screen_location = self.rel_loc_to_real((rand_location[0], rand_location[1]))\n",
    "            if self.is_in_form(mouseOperation.get_mouse_point()):\n",
    "                mouseOperation.mouse_move(screen_location[0], screen_location[1])\n",
    "                mouseOperation.mouse_click()\n",
    "            else:\n",
    "                self.is_mine_form = False\n",
    "\n",
    "        cv2.imshow(\"Sweeper Screenshot\", self.img_cv)\n",
    "\n",
    "        if cv2.waitKey(1) & 0xFF == ord('q'):\n",
    "            return False\n",
    "        return True"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "我们先不进行单元测试，先集成测试看一下整体效果："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import random\n",
    "import cv2\n",
    "import time\n",
    "import mouseOperation    # 鼠标操作，无需知道细节\n",
    "import imageProcess      # 这个文件我们之前实现了所有功能"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "miner = BoomMine()\n",
    "\n",
    "while 1:\n",
    "    try:\n",
    "        miner.process_once()\n",
    "    except:\n",
    "        pass\n",
    "# miner.show_map()\n",
    "# print(miner.next_steps)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "这个项目在一些机器上可能存在着 Bug, 试一下你能不能理解代码，找到出错的原因。\n",
    "\n",
    "### 扫雷算法实现\n",
    "\n",
    "简单起见，这里使用了一种比较暴力的扫雷逻辑：\n",
    "\n",
    "1. 遍历每一个已经有数字的雷块，判断在它周围的九宫格内未被打开的雷块数量是否和本身数字相同，如果相同则表明周围九宫格内全部都是地雷，进行标记；\n",
    "2. 再次遍历每一个有数字的雷块，取九宫格范围内所有未被打开的雷块，去除已经被上一次遍历标记为地雷的雷块，记录并且点开；\n",
    "3. 如果以上方式无法继续进行，那么说明遇到了死局，选择在当前所有未打开的雷块中随机点击，这意味着容易出现失败\n",
    "\n",
    "在实现它之后，你可以使用动态规划或者搜索的方式，改进这个扫雷算法，根据你的期望重构这个项目，发布到你的 GitHub 上面。"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}