在线文档 - Google 文档的数据协议设计

Google 文档作为 G Suite 重要的产品套件之一，作为优秀的在线协作文档而经常被开发者所讨论，在 Google 文档背后，有着一整套优秀的相关架构设计支撑，数据协议设计就是其中之一，非常具有学习和研究价值。

前言

截至 2020 年，Google 旗下的 G Suite 用户量达到 20 亿，而 Google 文档作为其重要的产品套件之一，作为优秀的在线协作文档而经常被开发者所讨论，在Google文档背后，有着一整套优秀的相关架构设计支撑，数据协议设计就是其中之一，非常具有学习和研究价值，本文旨在向研发同学详细介绍 Google 文档的数据协议设计精髓。
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-747W8POL-1642041334639)(https://docs.corp.kuaishou.com/d/loadimage/5343388864566933907)]

定义

在线协作文档的内容分类

在设计一个数据协议之前，需要将一个在线协作文档所包含的全部功能进行整理：

文字

文字内容
特殊转义符号
换行符，制表符等等
人类文字
文字样式
大小
字体
加粗
…

文档基本信息

背景色
纸张尺寸
纸张内外间距

插件

图片
超链接
评论
目录
表格
…

在线文档的用户操作分类

在对用户操作进行分类之前，我们需要引入操作指令（command）这个概念。简单来说，用户每次改变文档的操作，都将抽象成一次 command 发送到服务端，再由服务端将这次 command 分发给其他协作者的客户端。

在线协作文档有查看历史、协作、撤销的产品特性，故用户的每一个command都需要被原子化。我们对 Google文档的 command 进行分类（用户的每一次操作都应该可以通过这些类别的 command 组合来清晰的表达）：

create 创建

create 创建图片、列表项等实体
insertAfter 在指定的位置后插入
insertBefore 在指定的位置前插入

update 更新

更新已有的属性

delete 删除

deleteAt 在指定的位置删除

tether 绑定

将某些内容和create的实体绑定起来

分析用户对 Google 文档的操作

打开一个 Google 文档，正文书写为 ”快手，拥抱每一种生活“，并对文本进行一些样式修改，例如加粗，修改文本颜色等
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-ylYGet8F-1642041334641)(https://docs.corp.kuaishou.com/d/loadimage/-1598048658281776554)]

然后，通过其提供的版本历史记录功能，我们可以获取查看一个文档的历史版本的请求接口

https://docs.google.com/document/u/0/d/${docId}/showrevision

剔除掉一些非关键信息，我们得到以下 json 结构，很明显这是一个数组，接下来我们简称这个数组为 commands 并逐个进行分析

[{"ty":"is","ibi":1,"s":"快手，拥抱每一种生活"},{"ty":"as","st":"document","si":0,"ei":0,"sm":{"ds_pw":595.4399999999999,"ds_lhs":1,"ds_ph":841.68}},{"ty":"as","st":"headings","si":0,"ei":0,"sm":{"hs_h3":{"sdef_ps":{"ps_sb_i":false,"ps_sb":16},"sdef_ts":{"ts_fgc":"#434343","ts_bd":false,"ts_fgc_i":false,"ts_bd_i":false}},"hs_t":{"sdef_ps":{"ps_sb_i":false,"ps_sa":3,"ps_sa_i":false,"ps_sb":0},"sdef_ts":{"ts_bd":false,"ts_bd_i":true,"ts_fs":26,"ts_fs_i":false}},"hs_h2":{"sdef_ps":{"ps_sa":6,"ps_sa_i":false},"sdef_ts":{"ts_bd":false,"ts_bd_i":false,"ts_fs":16,"ts_fs_i":false}},"hs_h1":{"sdef_ps":{"ps_sb_i":false,"ps_sb":20},"sdef_ts":{"ts_bd":false,"ts_bd_i":true,"ts_fs":20,"ts_fs_i":false}},"hs_nt":{"sdef_ps":{"ps_lslm":1,"ps_lslm_i":false,"ps_sm":0,"ps_sm_i":false}},"hs_st":{"sdef_ps":{"ps_sb_i":false,"ps_sa":16,"ps_sa_i":false,"ps_sb":0},"sdef_ts":{"ts_ff_i":false,"ts_it":false,"ts_fs":15,"ts_ff":"Arial","ts_it_i":false,"ts_fs_i":false}},"hs_h6":{"sdef_ps":{"ps_sb_i":false,"ps_sa":4,"ps_sa_i":false,"ps_sb":12},"sdef_ts":{"ts_fgc":"#666666","ts_fgc_i":false,"ts_it":true,"ts_bd_i":true,"ts_fs":11,"ts_it_i":false,"ts_bd":false,"ts_fs_i":false}},"hs_h5":{"sdef_ps":{"ps_sb_i":false,"ps_sa":4,"ps_sa_i":false,"ps_sb":12},"sdef_ts":{"ts_fgc":"#666666","ts_bd":false,"ts_fgc_i":false,"ts_bd_i":true}},"hs_h4":{"sdef_ps":{"ps_sb_i":false,"ps_sa":4,"ps_sa_i":false,"ps_sb":14},"sdef_ts":{"ts_fgc":"#666666","ts_bd":false,"ts_fgc_i":false,"ts_bd_i":true}}}},{"ty":"as","st":"language","si":0,"ei":0,"sm":{"lgs_l":"zh_CN"}},{"ty":"as","st":"paragraph","si":11,"ei":11,"sm":{"ps_klt_i":true,"ps_awao_i":true,"ps_sm_i":true,"ps_ls_i":true,"ps_il_i":true,"ps_ir_i":true,"ps_al_i":true,"ps_bl_i":true,"ps_sd_i":true,"ps_sb_i":true,"ps_sa_i":true,"ps_lslm_i":true,"ps_br_i":true,"ps_bbtw_i":true,"ps_kwn_i":true,"ps_bt_i":true,"ps_ifl_i":true,"ps_bb_i":true}},{"ty":"as","st":"text","si":0,"ei":11,"sm":{"ts_un":false,"ts_un_i":true,"ts_sc":false,"ts_st_i":true,"ts_bgc":null,"ts_fs_i":true,"ts_bgc_i":true,"ts_ff_i":true,"ts_bd_i":true,"ts_va_i":true,"ts_fs":11,"ts_ff":"Arial","ts_bd":false,"ts_tw":400,"ts_it_i":true,"ts_fgc":"#000000","ts_fgc_i":true,"ts_it":false,"ts_va":"nor","ts_st":false,"ts_sc_i":true}},{"ty":"as","st":"text","si":1,"ei":2,"sm":{"ts_un":true,"ts_fgc":"#00796b","ts_un_i":false,"ts_fgc_i":false,"ts_bd_i":false,"ts_st":false,"ts_bd":true,"ts_st_i":false}},{"ty":"as","st":"text","si":3,"ei":10,"sm":{"ts_fgc":"#00796b","ts_st":false,"ts_fgc_i":false,"ts_st_i":false}},{"ty":"as","st":"text","si":11,"ei":11,"sm":{"ts_fgc":"#ff9900","ts_fgc_i":false}}
]

创建字符

commands[0]

{"ty":"is","ibi":1,"s":"快手，拥抱每一种生活"}

首先 “ty” 是 “type” 的缩写, “is” 是 “insertSpacers” 的缩写，然后 “ibi” 是 “insertBeforeIndex” 的缩写, “s” 是 “spacers” 的缩写,那么这个重新理解下这个 command

{"type":"insertSpacers","insertBeforeIndex":1,"spacers":"快手，拥抱每一种生活"
}

含义：在文档字符内容索引 1 的位置前插入 “快手，拥抱每一种生活”

创建文档基本信息

command[1]

{"ty":"as","st":"document","si":0,"ei":0,"sm":{"ds_pw":595.4399999999999,"ds_lhs":1,"ds_ph":841.68}
}

我们继续分析下一个 command ，“as” 是 “applyStyle” 的缩写，“st” 是 “styleType” 的缩写, “si” 是 “startIndex” 的缩写,“ei” 是 “endIndex” 的缩写, “sm” 是 “styleMap” 的缩写，“ds_pw” 是 “documentStyle_pageWidth” 的缩写，“ds_pw” 是 “documentStyle_pageHeight” 的缩写，“ds_lhs” 是 “documentStyle_lineHeightStrategy” 的缩写。

{"type":"applyStyle","styleType":"document","startIndex":0,"endIndex":0,"styleMap":{"documentStyle_pageWidth":595.4399999999999,"documentStyle_lineHeightStrategy":1,"documentStyle_pageHeight":841.68}
}

含义：这是一个文档全局配置，描述文档的纸张宽度为 595 point，高度为 841 point，因为要兼容不同设备的尺寸。

看到这，大家可能好奇为什么我们能够一眼就看出这个协议的含义。实际上，通过 debugger 调试 google docs 压缩后的代码，能够比较快的找到线索。后面的 command 我们就直接写翻译结果了。

https://docs.google.com/static/document/client/js/3556551332-client_js_prod_kix_core__zh_cn.js

创建标题样式(默认)

commands[2]

{"type":"applyStyle","styleType":"headings","startIndex":0,"endIndex":0,"styleMap":{"headStyle_h3":{"styleDefault_paragraphStyle":{"paragraphStyle_spacingBefore_inherit":false,"paragraphStyle_spacingBefore":16},"styleDefault_textStyle":{"textStyle_foregroundColor":"#434343","textStyle_bold":false,"textStyle_foregroundColor_inherit":false,"textStyle_bold_inherit":false}},"headStyle_title":{"styleDefault_paragraphStyle":{"paragraphStyle_spacingBefore_inherit":false,"paragraphStyle_spacingAfter":3,"paragraphStyle_spacingAfter_inherit":false,"paragraphStyle_spacingBefore":0},"styleDefault_textStyle":{"textStyle_bold":false,"textStyle_bold_inherit":true,"textStyle_fontSize":26,"textStyle_fontSize_inherit":false}},"headStyle_h2":{"styleDefault_paragraphStyle":{"paragraphStyle_spacingAfter":6,"paragraphStyle_spacingAfter_inherit":false},"styleDefault_textStyle":{"textStyle_bold":false,"textStyle_bold_inherit":false,"textStyle_fontSize":16,"textStyle_fontSize_inherit":false}},"headStyle_h1":{"styleDefault_paragraphStyle":{"paragraphStyle_spacingBefore_inherit":false,"paragraphStyle_spacingBefore":20},"styleDefault_textStyle":{"textStyle_bold":false,"textStyle_bold_inherit":true,"textStyle_fontSize":20,"textStyle_fontSize_inherit":false}},"headStyle_normalText":{"styleDefault_paragraphStyle":{"paragraphStyle_lslm":1,"paragraphStyle_lslm_i":false,"paragraphStyle_spacingMode":0,"paragraphStyle_spacingMode_inherit":false}},"headStyle_subTitle":{"styleDefault_paragraphStyle":{"paragraphStyle_spacingBefore_inherit":false,"paragraphStyle_spacingAfter":16,"paragraphStyle_spacingAfter_inherit":false,"paragraphStyle_spacingBefore":0},"styleDefault_textStyle":{"textStyle_fontFamily_inherit":false,"textStyle_italic":false,"textStyle_fontSize":15,"textStyle_fontFamily":"Arial","textStyle_italic_inherit":false,"textStyle_fontSize_inherit":false}},"headStyle_h6":{"styleDefault_paragraphStyle":{"paragraphStyle_spacingBefore_inherit":false,"paragraphStyle_spacingAfter":4,"paragraphStyle_spacingAfter_inherit":false,"paragraphStyle_spacingBefore":12},"styleDefault_textStyle":{"textStyle_foregroundColor":"#666666","textStyle_foregroundColor_inherit":false,"textStyle_italic":true,"textStyle_bold_inherit":true,"textStyle_fontSize":11,"textStyle_italic_inherit":false,"textStyle_bold":false,"textStyle_fontSize_inherit":false}},"headStyle_h5":{"styleDefault_paragraphStyle":{"paragraphStyle_spacingBefore_inherit":false,"paragraphStyle_spacingAfter":4,"paragraphStyle_spacingAfter_inherit":false,"paragraphStyle_spacingBefore":12},"styleDefault_textStyle":{"textStyle_foregroundColor":"#666666","textStyle_bold":false,"textStyle_foregroundColor_inherit":false,"textStyle_bold_inherit":true}},"headStyle_h4":{"styleDefault_paragraphStyle":{"paragraphStyle_spacingBefore_inherit":false,"paragraphStyle_spacingAfter":4,"paragraphStyle_spacingAfter_inherit":false,"paragraphStyle_spacingBefore":14},"styleDefault_textStyle":{"textStyle_foregroundColor":"#666666","textStyle_bold":false,"textStyle_foregroundColor_inherit":false,"textStyle_bold_inherit":true}}}}

含义：这个 command 比较大，也比较特殊，它用来描述标题的默认样式。你可以将其理解为一个配置项，是标题、副标题等的默认样式。

创建语言配置

commands[3]

 {"type":"applyStyle","styleType":"language","startIndex":0,"endexIndex":0,"styleMap":{"language_locale":"zh_CN"}}

含义：当前文档的语言

创建段落样式

commands[4]

{"type":"applyStyle","styleType":"paragraph","startIndex":11,"endIndex":11,"styleMap":{"paragraphStyle_keepLineTogether_inherit":true,"paragraphStyle_avoidWindowaAndOrphan_inherit":true,"paragraphStyle_spacingMode_inherit":true,"paragraphStyle_lineSpacing_inherit":true,"paragraphStyle_indentLeft_inherit":true,"paragraphStyle_indentRight_inherit":true,"paragraphStyle_alignment_inherit":true,"paragraphStyle_borderLeft_inherit":true,"paragraphStyle_styleDefault_inherit":true,"paragraphStyle_spacingBefore_inherit":true,"paragraphStyle_spacingAfter_inherit":true,"paragraphStyle_lineSpacing_lm_inherit":true,"paragraphStyle_borderRight_inherit":true,"paragraphStyle_borderBottomtw_inherit":true,"paragraphStyle_keepWidthNext_inherit":true,"paragraphStyle_borderTop_inherit":true,"paragraphStyle_ifl_inherit":true,"paragraphStyle_borderBottom_inherit":true}
}

含义：在以文档索引 11 位置为开始的段落样式

创建文本样式

commands[5] ~ commands[8], 我们通过 commands[5] 来介绍创建文本样式的数据结构

 {"type":"applyStyle","styleType":"text","startIndex":0,"endIndex":11,"styleMap":{"textStyle_underline":false,"textStyle_underline_inherit":true,"textStyle_small_caps":false,"textStyle_strikethough_inherit":true,"textStyle_backgroundcolor":null,"textStyle_fontSize_inherit":true,"textStyle_backgroundcolor_inherit":true,"textStyle_fontFamily_inherit":true,"textStyle_bold_inherit":true,"textStyle_verticalAligment_inherit":true,"textStyle_fontSize":11,"textStyle_fontFamily":"Arial","textStyle_bold":false,"textStyle_textWeight":400,"textStyle_italic_inherit":true,"textStyle_foregroundColor":"#000000","textStyle_foregroundColor_inherit":true,"textStyle_italic":false,"textStyle_verticalAligment":"nor","textStyle_strikethough":false,"textStyle_small_caps_inherit":true}
}

**含义：**在文档内容索引0~ 11 的位置创建一个文本样式，此时我们能发现文档的层级自上到下是：

文档→ 段落→ 文本

创建图片

这是通过 chrome 调试器查看 save 接口的请求体

https://docs.google.com/document/d/1VPHd0n4xbjMgRH8RWaG7nus7mU0s84TUztZaY4aUXwA/save

很明显，这是由多个 command 组合的，我们还是翻译一下

[{"type":"multi","multiCommands":[{"type":"insertSpacer","insertBeforeIndex":12,"spacers":"*"},{"type":"addEntity","entityType":"inline","id":"kix.mzzo7hjigmkz","entityPropertyMap":{"entity_embedded_object":{"embedded_object_marginLeft":9,"embedded_object_marginRight":9,"embedded_object_marginTop":9,"embedded_object_marginBottom":9,"embedded_object_type":0,"image_width":369.75,"image_height":272.25,"image_src":"","image_cid":"PLACEHOLDER_1d402a7d7b3610f3_0"}}},{"type":"TETHER_ENTITY","id":"kix.mzzo7hjigmkz","spaceIndex":12}]}
]

这个地方我们需要说明下插入图片（插件）操作的设计，实际这个操作是由 3 个 command 组合而成的，首先在文档内容的 12 位置插入一个特殊字符 “*”，然后给 12 位置创建一个”tether“类的command，然后这个“tether”类的command和 "addEntity"类的 command 建立链接，通过 "id： kix.mzzo7hjigmkz"建立关联
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-ERzYccp8-1642041334642)(https://docs.corp.kuaishou.com/d/loadimage/-8658999341642970419)]

创建表格

创建一个 2 行 3 列的表格
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-s2MzucNf-1642041334642)(https://docs.corp.kuaishou.com/d/loadimage/-889372112303852054)]

还是通过 Google Docs 的 Save 接口获得创建表格的 commands

[{"type": "multi","mts": [{"type": "insertSpacers","insertBeforeIndex": 14,"spacers": "\n\u0010\u0012\u001c\n\u001c\n\u001c\n\u0012\u001c\n\u001c\n\u001c\n\u0011"},cellTextStyleCommand1,cellTextStyleCommand2,cellTextStyleCommand3,cellTextStyleCommand...]
}]

由于这个 commands 非常大，需要把对单元格创建文本样式的操作都简化为 cellTextStyleCommand1,.cellTextStyleCommand , 创建表格的创建字符内容非常特殊，是 “\n\u0010\u0012\u001c\n\u001c\n\u001c\n\u0012\u001c\n\u001c\n\u001c\n\u0011”, 我们把它换种形式去看，他就是这样的

【TableStart】\u0010【Row】\u0012【CELL】\u001c\n 【CELL】\u001c\n 【CELL】 \u001c\n【Row】\u0012【CELL】\u001c\n 【CELL】\u001c\n 【CELL】 \u001c\n【TableEnd】\u0011

显而易见， Google 文档通过转义字符来代表表格起始、行起始、单元格起始以及表格结束。对表格内进行文字输入，实际就是在这些特殊的转义字符后添加字符串。

删除字符

假如我在文档上删除 “快手，拥抱每一种生活” 中的 “每一种”

{"type":"deleteSpacers","startIndex":6,"endIndex":8
}

含义：将文档索引 6 到 8 的位置删除

用户操作的合并

实际上用户通过对一个文档不断的操作，产生越来越多的 commands ，页面重新打开，Google 文档不会直接返回全量的 commands，而是对所有操作的 commands 进行一次合并，例如

{"type":"insertSpacers","insertBeforeIndex":1,"spacers":"快手","createdTime":"2020-11-03"
}

和

{"type":"insertSpacers","insertBeforeIndex":3,"spacers":"，拥抱每一种生活","createdTime":"2020-11-04"
}

最终合并

{"type":"insertSpacers","insertBeforeIndex":1,"spacers":"快手，拥抱每一种生活","createdTime":"2020-11-05"
}

总结

通过对抓取到的 command 进行翻译，我们能够理解到： Google文档设计的前后端交互协议*（command）*能够在保障涵盖对一篇文档全部内容的同时，最大限度的减轻服务端压力。通过这次对Google文档的系统分析，我们可以对在线协作文档的前后端交互协议设计（本文中的command）得出如下建议：