//// Copyright (c) 2019 Vinnie Falco (vinnie.falco@gmail.com) Copyright (c) 2025 Dmitry Arkhipov (grisumbras@yandex.ru) Distributed under the Boost Software License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) Official repository: https://github.com/boostorg/json //// [pagelevels=1,toclevels=1] = 基准测试 本节将 Boost.JSON 与两个广泛使用的具有类似功能的库进行性能比较：以高性能著称的 RapidJSON 和以功能丰富著称的JSON for Modern C++ 库（nlohmann/json）。bench 程序测量解析和序列化一组代表典型工作负载的 JSON 的吞吐量。评估的实现包括： .Implementations |=== |Name|Description | *boost(pool)* | Parses the input using a <>, 该资源针对解析后不进行后续修改的场景进行了优化。<> 对象在各次试验之间被复用，使得实现所分配的临时内存得以保留，从而提升性能。 | *boost* | Parses the input using the <> object is reused between trials, allowing temporary memory allocated by the implementation to persist and improve performance. | *rapidjson(pool)* | Parses the input using an instance of https://rapidjson.org/classrapidjson_1_1_memory_pool_allocator.html[`MemoryPoolAllocator`] 的实例实现的，该分配器针对“仅解析、不修改”的场景进行了优化。持有临时内存的 https://rapidjson.org/classrapidjson_1_1_generic_document.html[`Document`] 对象在多次测试之间不重复使用，否则内存消耗将无限增长，从而导致基准测试结果无效。 | *rapidjson* | Parses the input using an instance of 使用 https://rapidjson.org/classrapidjson_1_1_crt_allocator.html[`CrtAllocator`] 的实例来实现的，该分配器使用标准 C++ 分配器，适用于通用场景，包括解析后对文档的修改。持有临时内存的 https://rapidjson.org/classrapidjson_1_1_generic_document.html[`Document`] 对象在多次测试之间不重复使用，否则内存消耗将无限增长，导致基准测试结果无效。 | *nlohmann* | Parses the input using the static member function https://nlohmann.github.io/json/classnlohmann_1_1basic__json_ab330c13ba254ea41fbc1c52c5c610f45.html[`json::parse`] 来实现的，该函数使用默认的 `std` 分配器，适用于通用场景，包括解析后对文档的修改。该库未提供复用解析或序列化过程中所用临时存储的接口，因此无法在后续操作中重用这些临时内存。 == 方法论 首先加载所有输入文件。随后，每种配置会运行足够多的试验次数，以确保总运行时间不少于 5 秒。每次试验会记录所用时间、调用次数（解析或序列化）以及传输的字节数，并以此生成一个样本，同时计算出以 MB/s（兆字节每秒）为单位的吞吐量。每种配置会生成多个样本（目前为五个），剔除其中非中间两个的样本后，对剩余的两个样本取平均值，作为该配置的最终基准测试结果。 输入文件位于 bench/data 目录，按如下方式布局： .Input Files |=== |Name|Size|Description | <> | 125KB | Data from the Apache Jenkins installation. | <> | 2.2MB | The largest file, containing a large number of 2-element arrays holding （该二元数组是由）浮点坐标对（组成的） | <> | 1.69MB | A large JSON with a variety of nesting, types, and lengths. | <> | 64KB | An export of data from the Github Events API. | <> | 3.25MB | Google Summer of Code 2018 data. | <> | 216KB | An array of large objects. | <> | 2.91MB | A three.js example model serialized to JSON. | <> | 707KB | A JSON representing a 3D mesh. Contains many floating-point numbers. | <> | 1.54MB | mesh.json with whitespace added. | <> | 147KB | A array containing only floating-point numbers. | <> | 499KB | A JSON with lots of Cyrillic characters. | <> | 617KB | An export of data from Twitter's API. | <> | 550KB | twitter.json with whitespace removed and non-ASCII characters replaced with Unicode 转义符的。 | <> | 521KB | An export of data from Twitter's API. |=== 测试所用硬件： **Intel(R) Core(TM) i7-7700K CPU @ 4.20GHz**, Windows 10, 32GB RAM. 编译器及优化选项: gcc 8.1 (-O3), clang 12.0 (-O3), and msvc 19.26 (/O2). == 解析 === 解析 apache_builds.json image::images/parse_apache_builds.png[width=668,align="left"] === 解析 canada.json image::images/parse_canada.png[width=668,align="left"] === 解析 citm_catalog.json image::images/parse_citm_catalog.png[width=668,align="left"] === 解析 github_events.json image::images/parse_github_events.png[width=668,align="left"] === 解析 gsoc-2018.json image::images/parse_gsoc_2018.png[width=668,align="left"] === 解析 instruments.json image::images/parse_instruments.png[width=668,align="left"] === 解析 marine_ik.json image::images/parse_marine_ik.png[width=668,align="left"] === 解析 mesh.json image::images/parse_mesh.png[width=668,align="left"] === 解析 mesh.pretty.json image::images/parse_mesh_pretty.png[width=668,align="left"] === 解析 numbers.json image::images/parse_numbers.png[width=668,align="left"] === 解析 random.json image::images/parse_random.png[width=668,align="left"] === 解析 twitter.json image::images/parse_twitter.png[width=668,align="left"] === 解析 twitterescaped.json image::images/parse_twitterescaped.png[width=668,align="left"] === 解析 update-center.json image::images/parse_update_center.png[width=668,align="left"] == 序列化 === 序列化 canada.json image::images/serialize_canada.png[width=668,align="left"] === 序列化 citm_catalog.json image::images/serialize_citm_catalog.png[width=668,align="left"] === 序列化 github_events.json image::images/serialize_github_events.png[width=668,align="left"] === 序列化 gsoc-2018.json image::images/serialize_gsoc_2018.png[width=668,align="left"] === 序列化 instruments.json image::images/serialize_instruments.png[width=668,align="left"] === 序列化 marine_ik.json image::images/serialize_marine_ik.png[width=668,align="left"] === 序列化 mesh.json image::images/serialize_mesh.png[width=668,align="left"] === 序列化 mesh.pretty.json image::images/serialize_mesh_pretty.png[width=668,align="left"] === 序列化 numbers.json image::images/serialize_numbers.png[width=668,align="left"] === 序列化 random.json image::images/serialize_random.png[width=668,align="left"] === 序列化 twitter.json image::images/serialize_twitter.png[width=668,align="left"] === 序列化 twitterescaped.json image::images/serialize_twitterescaped.png[width=668,align="left"] === 序列化 update-center.json image::images/serialize_update_center.png[width=668,align="left"]