C++学习笔记—find原理—以opencv为例子的实操

1. 更新 locate 数据库 (必做！)

   • 操作: 打开终端，运行 sudo updatedb。
   • 原因: 确保 locate 的数据库包含你刚刚安装的新库的文件信息。否则后续步骤可能什么也找不到。

2. 初步定位：查找核心配置文件 (Config.cmake)

   • 操作: 运行以下命令

     locate -i 'config.cmake' | grep -i 'opencv'
     
     # 结果为
     ~ ❯ locate -i 'config.cmake' | grep -i 'opencv'
     /home/xfy/opencv_versions/4.5.5/build/3rdparty/ade/ade-0.1.1f/sources/ade/cmake/adeConfig.cmake.in
     /home/xfy/opencv_versions/4.5.5/build/CPackConfig.cmake
     /home/xfy/opencv_versions/4.5.5/build/CPackSourceConfig.cmake
     /home/xfy/opencv_versions/4.5.5/build/OpenCVConfig.cmake
     /home/xfy/opencv_versions/4.5.5/build/opencv_python_config.cmake
     /home/xfy/opencv_versions/4.5.5/build/unix-install/OpenCVConfig.cmake
     /home/xfy/opencv_versions/4.5.5/opencv/cmake/OpenCVGenConfig.cmake
     /home/xfy/opencv_versions/4.5.5/opencv/cmake/OpenCVGenPkgconfig.cmake
     /home/xfy/opencv_versions/4.5.5/opencv/cmake/templates/OpenCVConfig.cmake.in
     /home/xfy/opencv_versions/4.5.5/opencv/samples/hal/c_hal/config.cmake
     /home/xfy/opencv_versions/4.5.5/opencv/samples/hal/slow_hal/config.cmake
     /usr/lib/x86_64-linux-gnu/cmake/opencv4/OpenCVConfig.cmake
     /usr/local/opencv/4.5.5/lib/cmake/opencv4/OpenCVConfig.cmake
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/build/3rdparty/ade/ade-0.1.1f/sources/ade/cmake/adeConfig.cmake.in
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/build/CPackConfig.cmake
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/build/CPackSourceConfig.cmake
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/build/OpenCVConfig.cmake
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/build/opencv_python_config.cmake
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/build/unix-install/OpenCVConfig.cmake
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/cmake/OpenCVGenConfig.cmake
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/cmake/OpenCVGenPkgconfig.cmake
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/cmake/templates/OpenCVConfig.cmake.in
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/samples/hal/c_hal/config.cmake
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/home/myslam/3d/opencv-4.2.0/samples/hal/slow_hal/config.cmake
     /var/lib/docker/overlay2/0809ea31aa8a06f4c911b20f04526b19005bfabdd262f23d6bb3a4ee464f4dbb/diff/usr/local/lib/cmake/opencv4/OpenCVConfig.cmake
     /var/lib/docker/overlay2/1f605c5b588ec93f0f1947a29092b374d5fd184eedbfea41eba42bc85f33b932/diff/usr/lib/x86_64-linux-gnu/cmake/opencv4/OpenCVConfig.cmake

     可以看到有很多，但是我知道我是安装到/usr/local/的所以看/usr/local/opencv/4.5.5/lib/cmake/opencv4/OpenCVConfig.cmake
     这一个

3. 确定 find_package 名称

   由OpenCVConfig.cmake可以知道这里的名称大小写是OpenCV，所以应该是find_package(OpenCV REQUIRED)

   这里可以指定模块，也可以不指定，没有太大影响的。之前文章的例子里指定了模块的写法是find_package(OpenCV REQUIRED COMPONENTS core imgcodecs highgui)

4. 然后通过gedit或者vim等等查看这个文件

   #  The OpenCV CMake configuration file
   #
   #             ** File generated automatically, do not modify **
   #
   #  Usage from an external project:
   #    In your CMakeLists.txt, add these lines:
   #
   #    find_package(OpenCV REQUIRED)
   #    include_directories(${OpenCV_INCLUDE_DIRS}) # Not needed for CMake >= 2.8.11
   #    target_link_libraries(MY_TARGET_NAME ${OpenCV_LIBS})
   #
   #    Or you can search for specific OpenCV modules:
   #
   #    find_package(OpenCV REQUIRED core videoio)
   #
   #    You can also mark OpenCV components as optional:
   
   #    find_package(OpenCV REQUIRED core OPTIONAL_COMPONENTS viz)
   #
   #    If the module is found then OPENCV_<MODULE>_FOUND is set to TRUE.
   #
   #    This file will define the following variables:
   #      - OpenCV_LIBS                     : The list of all imported targets for OpenCV modules.
   #      - OpenCV_INCLUDE_DIRS             : The OpenCV include directories.
   #      - OpenCV_COMPUTE_CAPABILITIES     : The version of compute capability.
   #      - OpenCV_ANDROID_NATIVE_API_LEVEL : Minimum required level of Android API.
   #      - OpenCV_VERSION                  : The version of this OpenCV build: "4.5.5"
   #      - OpenCV_VERSION_MAJOR            : Major version part of OpenCV_VERSION: "4"
   #      - OpenCV_VERSION_MINOR            : Minor version part of OpenCV_VERSION: "5"
   #      - OpenCV_VERSION_PATCH            : Patch version part of OpenCV_VERSION: "5"
   #      - OpenCV_VERSION_STATUS           : Development status of this build: ""
   #
   #    Advanced variables:
   #      - OpenCV_SHARED                   : Use OpenCV as shared library
   #      - OpenCV_INSTALL_PATH             : OpenCV location
   #      - OpenCV_LIB_COMPONENTS           : Present OpenCV modules list
   #      - OpenCV_USE_MANGLED_PATHS        : Mangled OpenCV path flag
   #
   #    Deprecated variables:
   #      - OpenCV_VERSION_TWEAK            : Always "0"

   翻译一下

   # ===================================================================================
   #  OpenCV CMake 配置文件
   #
   #             ** 文件自动生成，请勿修改 **
   #
   #  从外部项目使用：
   #    在你的 CMakeLists.txt 中，添加这些行：
   #
   #    find_package(OpenCV REQUIRED)
   #    include_directories(${OpenCV_INCLUDE_DIRS}) # 对于 CMake >= 2.8.11 已不再需要
   #    target_link_libraries(MY_TARGET_NAME ${OpenCV_LIBS})
   #
   #    或者你可以搜索特定的 OpenCV 模块：
   #
   #    find_package(OpenCV REQUIRED core videoio)
   #
   #    你也可以将 OpenCV 组件标记为可选：
   #    find_package(OpenCV REQUIRED core OPTIONAL_COMPONENTS viz)
   #
   #    如果找到了模块，那么 OPENCV_<MODULE>_FOUND 会被设置为 TRUE。
   #
   #    该文件将定义以下变量：
   #      - OpenCV_LIBS                     : OpenCV 模块的所有导入目标 (imported targets) 的列表。
   #      - OpenCV_INCLUDE_DIRS             : OpenCV 的包含目录。
   #      - OpenCV_COMPUTE_CAPABILITIES     : 计算能力 (Compute Capability) 的版本。
   #      - OpenCV_ANDROID_NATIVE_API_LEVEL : 最低要求的 Android API 级别。
   #      - OpenCV_VERSION                  : 此 OpenCV 构建的版本："4.5.5"
   #      - OpenCV_VERSION_MAJOR            : OpenCV_VERSION 的主版本号："4"
   #      - OpenCV_VERSION_MINOR            : OpenCV_VERSION 的次版本号："5"
   #      - OpenCV_VERSION_PATCH            : OpenCV_VERSION 的修订版本号："5"
   #      - OpenCV_VERSION_STATUS           : 此构建的开发状态：""
   #
   #    高级变量：
   #      - OpenCV_SHARED                   : 将 OpenCV 作为共享库使用
   #      - OpenCV_INSTALL_PATH             : OpenCV 的位置
   #      - OpenCV_LIB_COMPONENTS           : 当前存在的 OpenCV 模块列表
   #      - OpenCV_USE_MANGLED_PATHS        : Mangled OpenCV 路径标志
   #
   #    已弃用变量：
   #      - OpenCV_VERSION_TWEAK            : 始终为 "0"

   可以看到这里说明了执行findpackage之后给哪些变量赋予了哪些值，也写了用法

   #  从外部项目使用：
   #    在你的 CMakeLists.txt 中，添加这些行：
   #
   #    find_package(OpenCV REQUIRED)
   #    include_directories(${OpenCV_INCLUDE_DIRS}) # 对于 CMake >= 2.8.11 已不再需要
   #    target_link_libraries(MY_TARGET_NAME ${OpenCV_LIBS})

   这段注释虽然目的是指导用户，但它本身混合了新旧两种风格的痕迹，我们需要仔细解读：

   1. 旧方式 (Variable-Based) 的体现：

      • 明确的指令: 注释中明确给出了以下使用步骤：

        include_directories(${OpenCV_INCLUDE_DIRS})
        target_link_libraries(MY_TARGET_NAME ${OpenCV_LIBS})

      • 这是典型的旧式 CMake 用法。它依赖于 find_package 执行后设置的两个变量：

        • ${OpenCV_INCLUDE_DIRS}: 包含了需要添加到编译器搜索路径的头文件目录。你需要手动调用 include_directories() (或现代的 target_include_directories()) 来使用它。
        • ${OpenCV_LIBS}: 包含了需要链接的库文件列表（可能是库文件的绝对路径，也可能在旧 CMake 中是链接器标志）。你需要手动将这个变量传递给 target_link_libraries()。
   2. 新方式 (Target-Based) 的线索和提示：

      • 关键注释: include_directories(${OpenCV_INCLUDE_DIRS}) # Not needed for CMake >= 2.8.11

        • 这是最强烈的暗示。它告诉你，对于较新版本的 CMake (2.8.11 及以后，实际上现代 CMake 实践通常指 3.x 系列)，手动添加 OpenCV_INCLUDE_DIRS 是不必要的。为什么？因为现代 CMake 的核心是目标 (Target)。当你链接到一个设计良好的导入目标 (Imported Target) 时，这个目标本身就封装了它所需的包含目录、链接库、编译定义等信息。CMake 会自动将这些信息传递给链接它的目标。

      • 变量 OpenCV_LIBS 的描述: OpenCV_LIBS : The list of all **imported targets** for OpenCV modules.

        • 这一点非常重要！虽然旧方式的使用示例中是 target_link_libraries(... ${OpenCV_LIBS})，但这里对 OpenCV_LIBS 变量内容的描述明确指出了它包含的是 “imported targets”。这意味着 ${OpenCV_LIBS} 变量里存放的不再仅仅是库文件名或路径，而是 CMake 能理解的、代表了 OpenCV 各个模块的目标名称列表（例如 OpenCV::core, OpenCV::imgproc 等）。

      • 结合以上两点，可以推断出新方式的用法（尽管这段注释没有直接写出来）

        find_package(OpenCV REQUIRED COMPONENTS core imgproc highgui) # 假设需要这三个
        target_link_libraries(MY_TARGET_NAME PRIVATE OpenCV::core OpenCV::imgproc OpenCV::highgui)
        
        这里find_package可以不指定具体哪个模块而是用
        find_package(OpenCV REQUIRED)
        区别不大
        
        target_link_libraries推荐指定模块，不然就会把所有模块全都链接到项目上
        
        说是这样说，但是我指定模块运行失败，不知道是因为名字不对还是啥
        target_include_directories(my_app PUBLIC include)才成功的

总结：还是用文档里说的吧

#  从外部项目使用：
#    在你的 CMakeLists.txt 中，添加这些行：
#
find_package(OpenCV REQUIRED)
include_directories(${OpenCV_INCLUDE_DIRS}) # 对于 CMake >= 2.8.11 已不再需要
target_link_libraries(MY_TARGET_NAME ${OpenCV_LIBS})

其中include_directories可有可无