从RocksDB 6.0开始弃用和删除
好处
如何使用?
例子
配置 RocksLuaCompactionFilterOptions
- 配置Lua脚本(RocksLuaCompactionFilter::Script)
- Error Log 配置 (RocksLuaCompactionFilterOptions::error_log)
动态更新Lua脚本

从RocksDB 6.0开始弃用和删除

RocksDB CompactionFilter提供了一种基于后台自定义逻辑删除/过滤过期键/值对的方法。现在我们在它上面实现了一个扩展，允许用户在Lua中实现自定义的CompactionFilter。这个特性在RocksDB 5.0中可用。

好处

在Lua中开发CompactionFilter有以下好处:
动态更改: 可以在不关闭RocksDB实例的情况下动态更新Lua压缩过滤器。
单个二进制单元: 由于可以动态地更新Lua压缩过滤器，因此不再需要按照C++压缩过滤器的要求重新构建二进制。
             这对于构建在RocksDB之上的服务来说是一个巨大的好处，RocksDB为客户维护定制的CompactionFilters。
更安全: CompactionFilter中的错误将不再导致核心转储。Lua 引擎捕获所有异常。

如何使用?

使用RocksLuaCompactionFilter很简单。你需要做的就是以下步骤:
将LUA_PATH设置为Lua的根目录，构建RocksDB。
用lua脚本配置RocksLuaCompactionFilterOptions。（更多细节将在下一节描述）
使用步骤1中的RocksLuaCompactionFilterOptions构建了一个 RocksLuaCompactionFilterFactory。
将步骤2中的RocksLuaCompactionFilterFactory传递到ColumnFamilyOptions::compaction_filter_factory。

例子

这里面有一个简单的RocksLuaCompactionFilter，过滤掉任何键的初始值小于r:

lua::RocksLuaCompactionFilterOptions lua_opt; // removes all keys whose initial is less than ‘r’ lua_opt.lua_script =
```
 "function Filter(level, key, existing_value)\n"
 "  if key:sub(1,1) < 'r' then\n"
 "    return true, false, \"\"\n"
 "  end\n"
 "  return false, false, \"\"\n"
 "end\n"
 "\n"
 "function FilterMergeOperand(level, key, operand)\n"
 "  return false\n"
 "end\n"
 "function Name()\n"
 "  return \"KeepsAll\"\n"
 "end\n";
```
// specify error log. auto* rocks_logger = new facebook::rocks::RocksLogger(
```
 "RocksLuaTest",
 true,  // print error message in GLOG
 true,  // print error message to scribe, available in LogView `RocksDB ERROR`
 nullptr);
```
// Create RocksLuaCompactionFilter with the above lua_opt and pass it to Options rocksdb::Options options; options.compaction_filter_factory =
```
 std::make_shared<rocksdb::lua::RocksLuaCompactionFilterFactory>(lua_opt);
```
… // open FbRocksDB with the above options rocksdb::DB* db; auto status = openRocksDB(options, “RocksDBWithLua”, &db);

配置 RocksLuaCompactionFilterOptions

在这里，我们将更详细地介绍如何配置RocksLuaCompactionFilterOptions。
RocksLuaCompactionFilterOptions的定义可以在include/rocksdb/utilities/lua/rocks_lua_compaction_filter.h中找到。

配置Lua脚本(RocksLuaCompactionFilter::Script)

第一个也是最重要的参数是RocksLuaCompactionFilterOptions::script，您的Lua压缩过滤器将在这里实现。Lua脚本必须实现所需的函数，即Name()和Filter()。

  // The lua script in string that implements all necessary CompactionFilter
  // virtual functions.  The specified lua_script must implement the following
  // functions, which are Name and Filter, as described below.
  //
  // 0. The Name function simply returns a string representing the name of
  //    the lua script.  If there's any erorr in the Name function, an
  //    empty string will be used.
  //    --- Example
  //      function Name()
  //        return "DefaultLuaCompactionFilter"
  //      end
  //
  //
  // 1. The script must contains a function called Filter, which implements
  //    CompactionFilter::Filter() , takes three input arguments, and returns
  //    three values as the following API:
  //
  //   function Filter(level, key, existing_value)
  //     ...
  //     return is_filtered, is_changed, new_value
  //   end
  //
  //   Note that if ignore_value is set to true, then Filter should implement
  //   the following API:
  //
  //   function Filter(level, key)
  //     ...
  //     return is_filtered
  //   end
  //
  //   If there're any error in the Filter() function, then it will keep
  //   the input key / value pair.
  //
  //   -- Input
  //   The function must take three arguments (integer, string, string),
  //   which map to "level", "key", and "existing_value" passed from
  //   RocksDB.
  //
  //   -- Output
  //   The function must return three values (boolean, boolean, string).
  //     - is_filtered: if the first return value is true, then it indicates
  //       the input key / value pair should be filtered.
  //     - is_changed: if the second return value is true, then it indicates
  //       the existing_value needs to be changed, and the resulting value
  //       is stored in the third return value.
  //     - new_value: if the second return value is true, then this third
  //       return value stores the new value of the input key / value pair.
  //
  //   -- Examples
  //     -- a filter that keeps all key-value pairs
  //     function Filter(level, key, existing_value)
  //       return false, false, ""
  //     end
  //
  //     -- a filter that keeps all keys and change their values to "Rocks"
  //     function Filter(level, key, existing_value)
  //       return false, true, "Rocks"
  //     end
  std::string lua_script;

一个不使用value(RocksLuaCompactionFilter::ignore_value)的优化，以防您的CompactionFilter从不使用value来决定是保留还是丢弃一个键/值对,然后设置RocksLuaCompactionFilterOptions::ignore_value=true并实现简化的Filter() API。我们的结果表明，这种优化可以节省高达40%的CPU开销引入LuaCompactionFilter:

 // If set to true, then existing_value will not be passed to the Filter
 // function, and the Filter function only needs to return a single boolean
 // flag indicating whether to filter out this key or not.
 //
 //   function Filter(level, key)
 //     ...
 //     return is_filtered
 //   end
 bool ignore_value = false;

简化的Filter() API只接受两个输入参数，并且只返回一个布尔标志，指示是保留还是放弃输入键。

Error Log 配置 (RocksLuaCompactionFilterOptions::error_log)

* 当RocksLuaCompactionFilter 遇到热河错误时，它将作为no-op执行，并且对于任何导致错误的键/值对总是返回false。开发人员可以配置RocksLuaCompactionFilterOptions::error_log 来记录任何Lua错误:
 // When specified a non-null pointer, the first "error_limit_per_filter"
 // errors of each CompactionFilter that is lua related will be included
 // in this log.
 std::shared_ptr<Logger> error_log;

注意，对于每个压缩作业，我们只记录error_log中的前几个Lua错误，以避免生成太多的错误消息，并且每个压缩作业报告的错误数量可以通过error_limit_per_filter配置。默认值是1。

 // The number of errors per CompactionFilter will be printed
 // to error_log.
 int error_limit_per_filter = 1;

动态更新Lua脚本

要在运行RocksDB数据库时更新Lua脚本，只需调用RocksLuaCompactionFilterFactory的SetScript() API:

// Change the Lua script so that the next compaction after this
// function call will use the new Lua script.
void SetScript(const std::string& new_script);