Boost.log:如何防止输出在使用add_file_log()函数时被复制到所有添加的流？

Question

我使用该add_file_log()函数初始化将日志记录存储到文本文件中的日志记录接收器.当我定义几个接收器时,我观察到:

• 为每个接收器创建一个文件.
• 输出将复制到所有文件.

这是我的记录器:

class logger
{
public:
  logger(const logger&) =delete;
  logger(logger&&) =delete;
  logger& operator=(const logger&) =delete;
  logger& operator=(logger&&) =delete;

  static logger& get_instance(
    const std::string& file,
    bool console
  )
  {
      boost::log::register_simple_formatter_factory<
                                                    boost::log::trivial::severity_level,
                                                    char
                                                   >("Severity");

      std::string the_format = "[%TimeStamp%] (%LineID%) [%Severity%]: %Message%";

      if(!file.empty()) {
        boost::log::add_file_log(
          boost::log::keywords::file_name = file + "_%N.log",
          boost::log::keywords::rotation_size = 10 * 1024 * 1024,
          boost::log::keywords::time_based_rotation =
            boost::log::sinks::file::rotation_at_time_point(0, 0, 0),
          boost::log::keywords::auto_flush = true,
          //boost::log::keywords::open_mode = (std::ios::out | std::ios::app),
          boost::log::keywords::format = the_format
        );
      }

      boost::log::add_common_attributes();
      static logger instance{ the_format, console };
      return instance;
  }

  void log(
    const std::string& msg
  )
  {
      BOOST_LOG_SEV ( m_log_, boost::log::trivial::info ) << msg;
  }

private:
  boost::log::sources::severity_logger<
                                       boost::log::trivial::severity_level
                                      > m_log_;

  logger(
    const std::string& format,
    bool console
  )
  {
      if(console) {
        boost::log::add_console_log(
          std::clog,
          boost::log::keywords::format = format
        );
      }
  }
}; // logger

这是我的main()功能:

void test(
  const std::string& file
)
{
  logger& lg1 = logger::get_instance( file, false );
  lg1.log( "Hello" );
  lg1.log( "World" );
  lg1.log( "Bye" );
} // test

int main()
{
  unsigned char result = EXIT_SUCCESS;

  try
  {
    std::string file1 = "a.txt",
                file2 = "b.txt";
    logger& lg = logger::get_instance( file1, false );

    for(int i = 1; i<=10; i++) {
       lg.log( std::to_string(i) );
       if(i == 5) {
         test( file2 );
       }
    }
  }
  catch ( std::exception& e )
  {
    std::cerr << "Error: " << e.what() << std::endl;
    result = EXIT_FAILURE;
  }
  return result;
}

运行示例后,文件包含:

a.txt_0.log

[2016-Aug-31 11:49:48.584353] (1) [info]: 1
[2016-Aug-31 11:49:48.585376] (2) [info]: 2
[2016-Aug-31 11:49:48.585418] (3) [info]: 3
[2016-Aug-31 11:49:48.585442] (4) [info]: 4
[2016-Aug-31 11:49:48.585462] (5) [info]: 5
[2016-Aug-31 11:49:48.585505] (6) [info]: Hello  <-- 
[2016-Aug-31 11:49:48.585610] (7) [info]: World  <-- Generated by second logger
[2016-Aug-31 11:49:48.585672] (8) [info]: Bye    <--
[2016-Aug-31 11:49:48.585709] (9) [info]: 6
[2016-Aug-31 11:49:48.585744] (10) [info]: 7
[2016-Aug-31 11:49:48.585777] (11) [info]: 8
[2016-Aug-31 11:49:48.585813] (12) [info]: 9
[2016-Aug-31 11:49:48.585842] (13) [info]: 10

b.txt_0.log

[2016-Aug-31 11:49:48.585505] (6) [info]: Hello
[2016-Aug-31 11:49:48.585610] (7) [info]: World
[2016-Aug-31 11:49:48.585672] (8) [info]: Bye
[2016-Aug-31 11:49:48.585709] (9) [info]: 6    <--
[2016-Aug-31 11:49:48.585744] (10) [info]: 7   <--
[2016-Aug-31 11:49:48.585777] (11) [info]: 8   <-- Generated by the first logger
[2016-Aug-31 11:49:48.585813] (12) [info]: 9   <--
[2016-Aug-31 11:49:48.585842] (13) [info]: 10  <--

我怎么能防止这种行为？我希望每个文件只存储其相关记录器生成的信息.

Answer 1

您似乎对Boost.Log的工作方式存在误解.

有源和汇.源接收数据,例如字符串,并使用它创建条目.然后将该条目提供给核心,核心将其分配给所有接收器.然后,接收器可以过滤,格式化并将条目输出到他们想要的任何位置,例如stdout文件.

一个示例源将是severity_logger您正在使用.您可能习惯使用术语"记录器"而不是"源",但"记录器"不是很精确,因为记录是一个多阶段过程.

您通常不必创建多个源("记录器").相反,您可以添加多个全局接收器.在您的情况下,每个文件都需要一个过滤后的接收器.

                                  +--------------+
                            +---> | console sink | ----> stdout
                            |     +--------------+
                            |
+--------+      +------+    |     +--------------+
| source | ---> | core | ---+---> | file sink    | ----> log1.txt
+--------+      +------+    |     +--------------+
                            |
                            |     +--------------+
                            +---> | file sink    | ----> log2.txt
                                  +--------------+

现在,您可以拥有多个源,每个源都有自己的线程模型,属性,字符类型等,但它们仍然会生成条目并将它们提供给核心.在你的情况下,它不会非常有用.

让我们把标题放在一边:

#include <string>
#include <fstream>
#include <boost/log/sinks.hpp>
#include <boost/log/utility/setup/formatter_parser.hpp>
#include <boost/log/sources/severity_channel_logger.hpp>
#include <boost/log/trivial.hpp>
#include <boost/log/utility/setup/file.hpp>
#include <boost/log/utility/setup/common_attributes.hpp>
#include <boost/log/utility/setup/console.hpp>
#include <boost/log/expressions.hpp>
#include <boost/log/attributes/scoped_attribute.hpp>

namespace bl = boost::log;

开始吧:

BOOST_LOG_ATTRIBUTE_KEYWORD(tag_attr, "Tag", std::string);

日志条目具有可在每次记录某些内容时设置的属性.这些属性通常用于格式化(例如"[%TimeStamp%] [%Message%]"),但我们将添加一个新属性以允许区分不同的文件.我称之为"标签"属性.

using logger_type = bl::sources::severity_logger<bl::trivial::severity_level>;
static logger_type g_logger;

const std::string g_format = "[%TimeStamp%] (%LineID%) [%Severity%] [%Tag%]: %Message%";

现在,在此示例中,实际的boost记录器是一个全局对象(g_logger).您可能希望限制其范围并将其传递给您自己的logger对象.我还将格式设为全局常量.因人而异.

这是logger班级:

class logger
{
public:
    logger(std::string file)
        : tag_(file)
    {
        using backend_type = bl::sinks::text_file_backend;
        using sink_type = bl::sinks::synchronous_sink<backend_type>;
        namespace kw = bl::keywords;

        auto backend = boost::make_shared<backend_type>(
            kw::file_name = file + "_%N.log",
            kw::rotation_size = 10 * 1024 * 1024,
            kw::time_based_rotation = bl::sinks::file::rotation_at_time_point(0, 0, 0),
            kw::auto_flush = true);

        auto sink = boost::make_shared<sink_type>(backend);
        sink->set_formatter(bl::parse_formatter(g_format));
        sink->set_filter(tag_attr == tag_);

        bl::core::get()->add_sink(sink);
    }

    void log(const std::string& s)
    {
        BOOST_LOG_SCOPED_THREAD_TAG("Tag", tag_);
        BOOST_LOG_SEV(g_logger, bl::trivial::info) << s;
    }

private:
    const std::string tag_;
};

我已经将文件名用作标记,但只要它是唯一的,它就可以是其他任何东西.每个日志条目都将此标记作为属性,将在接收器过滤器中使用.

首先,text_file_backend创建a并将其提供给新的接收器,然后将其添加到核心.这实际上就是你打电话时发生的事情add_file_log(),它只是一个辅助功能.我重用了你在例子中使用的相同参数(文件名模式,旋转等)

有趣的是这一行:

sink->set_filter(tag_attr == tag_);

这里,tag_attr上面定义为属性关键字.关键字在Boost.Log中有点不寻常:它们可用于创建将在运行时评估的表达式.在这种情况下,接收器只接受其中的条目tag_attr == tag_.因此,当记录器记录某些内容时,它会将自己的标记设置为属性,而接收器将忽略任何没有此标记的内容.在log(),您可以看到"Tag"正在设置的属性.

这是main():

int main()
{
    bl::register_simple_formatter_factory<bl::trivial::severity_level, char>("Severity");
    boost::log::add_common_attributes();

    bl::add_console_log(std::clog, bl::keywords::format=g_format);

    logger lg1("1");
    logger lg2("2");

    lg1.log("a");
    lg1.log("b");
    lg1.log("c");

    lg2.log("d");
    lg2.log("e");
    lg2.log("f");
}

你会看到我已经将常见的东西移到了外面,logger因为它并不真正属于那里.条目"a","b"并且"c"将被写入"1_0.txt"和"d","e"并"f"以"2_0.txt".所有六个条目都将写入控制台.

     +--------------+
     | lg1.log("a") |
     +--------------+
            |
            v
+-------------------------+
| Entry:                  |
|   Timestamp: 1472660811 |
|   Message:   "a"        |
|   LineID:    1          |
|   Severity:  info       |
|   Tag:       "1"        |
+-------------------------+
            |            
            v                 +----------------------+
         +------+             | console sink         |
         | core | -----+----> |   file: stdout       |  --> written
         +------+      |      |   filter: none       |
                       |      +----------------------+
                       |      
                       |      +----------------------+
                       |      | file sink            |
                       +----> |   file: "1_0.txt"    |  --> written
                       |      |   filter: tag == "1" |
                       |      +----------------------+
                       |      
                       |      +----------------------+
                       |      | file sink            |
                       +----> |   file: "2_0.txt"    |  --> discarded
                              |   filter: tag == "2" |
                              +----------------------+