MLSQL 一分钟让 Kylin 装备 ETL 能力！

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在上周举办的 Apache Kylin + MLSQL Meetup 中，我们邀请了来自 MLSQL 社区的技术大佬 "},{"type":"text","marks":[{"type":"strong"}],"text":"祝威廉"},{"type":"text","text":" 来进行分享。大家都知道 Kylin 一向以强大的分析能力和丰富的周边生态而备受欢迎，但是 Kylin 自身还欠缺一些 ETL 能力。在本次分享中，祝威廉演示了如何在 Kylin 中快速完成数据处理，用户不用离开 Kylin 即可完成大规模数据分析整个 Pipeline，同时也分享了未来 Kylin 和 MLSQL 之间联动的更多可能性。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"以下是祝威廉在 Kylin Meetup 的演讲实录"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/11\/bf\/11c2e1301d9883b7ea6262ffyyec8dbf.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"本次分享主要分为三部分，首先会宏观地去看一下 Kylin 和 MLSQL 这两者为什么是互补的，其次会进行一个 Demo 的演示，如何在 Kylin 里面通过简单几行代码的修改，就能够支持 MLSQL 的语言，从而完成一些 ETL 的工作。因为 Demo 目前只演示了其中一个场景，我会在第三部分来分享 Kylin 和 MLSQL 之间未来更多的可能性，两者其实是有许多可以联动的地方，大家也可以期待两者联动实现更好的效果。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"Kylin 和 MLSQL 为什么是互补的"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"首先来介绍一下为什么说 Kylin 和 MLSQL 是互补的，如上图所示，目前在整个数据分析领域，主要有三大部分："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/fe\/53\/fe496b6d4060df45cdda5f02609yy053.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一部分是 BI（Business Intelligence）；"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二部分是数据处理，其中包含了目前正热的流处理，以及传统的批处理等；"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第三部分是 AI（Artificial Intelligence）。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Kylin 其实是非常好地覆盖了 BI 这一部分，并且具有非常完美的生态系统，Kylin 对大部分 BI 工具都是支持的，具备一部分的处理功能比如说多维聚合分析 。用户使用 Kylin 去做数据分析，其实就是利用它的数据处理功能。但如果涉及到数据处理中 ETL 相关的功能，Kylin 目前是不太适合去做的，比如一部分流式的数据处理，不过这也正是 MLSQL 可以覆盖的。MLSQL 在 BI 这一部分并不擅长，但是在数据处理和 AI 这两部分是比较强的。如果 Kylin 和 MLSQL 两者进行整合，那么基本就可以完美覆盖大数据分析最主要的三个领域。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"Kylin 简介"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"首先来简单介绍一下 Kylin。Kylin 是一个 OLAP 的系统，支持 "},{"type":"text","marks":[{"type":"strong"}],"text":"高并发、亚秒级查询"},{"type":"text","text":" ，这两点优势其实是许多工具难以超越的。同时，Kylin "},{"type":"text","marks":[{"type":"strong"}],"text":"生态完备"},{"type":"text","text":" ，支持各种 BI 工具，比如像 Tableau PowerBI 等。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/db\/f5\/db5626f906d45877b17388e9f37a10f5.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如上面 Kylin 官网架构图所示，Kylin 有一个 Cube Build Engine，也就是用来构建 Cube 的过程。这个过程意味着它必须有数据来源，这个过程可能存在两个问题："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一是 Build Engine 其实更加侧重的是在数据建模这一块，数据来源需要进行加工才能适合去做建模，所以它其实是缺乏较强的 ETL 能力；"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"二是 Kylin 的很多数据源是要单独去支持的，MLSQL 本身已经支持非常多的数据源了，如果把两者打通，那就意味着这两部分 Kylin 无需自己支持，直接借助一些工具就能够实现，比如前面提到的 CSV 数据源。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"目前有两种方式："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"方式一：大家可以自建一个 ETL 的基础设施；"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"方式二：如果 Kylin 本身可以支持，用户就不用离开 Kylin 就可以完成这件事情了，对用户而言，就能带来更好的使用体验。"}]}]}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"MLSQL 简介"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"大部分 Kylin 用户可能暂时对 MLSQL 不太了解，我先来进行一个简单介绍。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"主要从以下两个方面进行介绍："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/1e\/84\/1ee4f6b2253be6c86e46117d0c53dd84.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"第一方面， MLSQL 是一个面向大数据和 AI 设计的语言。"},{"type":"text","text":" 正常情况下，使用大数据，用户可以用 Scala 或者 JAVA 语言；使用 AI 的用户可能会用 Python 语言；其实数据处理大家真正用的最多的应该是 SQL。那为什么还需要像 MLSQL 这样的语言呢？简单来说，如果用户去用 Scala 写代码，其实是需要深入掌握和应用这门语言的，另外包括 Python 也是存在一定的使用门槛的。SQL 很简单，但是如果想实现 AI 或是更强的 ETL 功能，SQL 就会存在能力欠缺的问题。在这样的情况下，MLSQL 其实就变成很有必要的一种语言了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二方面，前面也提到 MLSQL 是一种语言，任何一种语言都一定要有一个实现，就像我们之前提到的 JAVA语言，就有 JVM 虚拟机的实现。其实 MLSQL 底层也有一个实现这门语言的引擎，本质上底层是基于 Spark 做的引擎。 "},{"type":"text","marks":[{"type":"strong"}],"text":"MLSQL 的目标其实是要做一个真正整合数据管理、商业分析和机器学习的统一的语言和平台。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"接下来介绍 MLSQL 的一些语法，让大家简单熟悉一下。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/f8\/d5\/f82dbae95094e4972feaf9dde7220ed5.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"首先，在加载数据的部分，在 SQL 里必须要有 Catalog 才可以实现，默认数据源可能是 Hive。如果用户想加载其他数据，用 SQL 会比较麻烦。在这一部分，MLSQL 进行了一些拓展，比如使用 MLSQL 的 Load 语句，用户就可以加载任意数据源。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"其次是可以用 SQL 去做标准的数据处理，同时处理完的结果可以进行保存。如上图所示，左侧是 AI 训练，MLSQL 其实是把 Python 和 SQL 进行了融合，用户可以用 Python 去直接使用 SQL 的表， Python 处理后的结果也可以作为表继续在 SQL 里使用。这里只是一些简单介绍，更多内容大家可以关注 MLSQL 的官方文档。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"http:\/\/docs.mlsql.tech\/mlsql-console\/"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"如何不离开 Kylin 就能完成 ETL"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"以上简单介绍了 Kylin 和 MLSQL，要想整合两者，首先必须满足以下几个标准："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"首先，整合成本必须要低。"},{"type":"text","text":" 最差的情况下，是可能需要去修改一丢丢代码的，实际上，我和Kylin社区的Committer张智超也去进行了一些实践。在极致的情况下，只需要修改两行代码就可以完成整合。当然，这个过程中可能会新增一些文件，但是修改的部分可能只有几行代码。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"其次，MLSQL 希望优先补全 Kylin 的 ETL 能力，可以实现在 Kylin 里面就可以直接进行 ETL。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/bd\/32\/bd989db67516490da949381512bf5032.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"考虑到以上两点，大家可以看到上图里做了一个简单注解是 "},{"type":"text","marks":[{"type":"strong"}],"text":"--%mlsql"},{"type":"text","text":" ，后面可以跟一个 MLSQL 引擎地址，有了这个注解，即可把整个框中的脚本内容发到 MLSQL 引擎里面去执行。如果没有这个注解，就会按照 Kylin 的执行逻辑去执行。有了初步思路，我们去做了一个演示 Demo，大家可以一起来看看演示效果。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"Demo 演示"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Demo 中的这个界面大家应该非常熟悉，这是 Kylin 的运行页面。今天分享的 Demo 就是演示用户如何在 Kylin 的产品里，将一个 CSV 文件导入到 Hive 以供 Kylin 进行建模，整个过程仅修改几行代码就能把 ETL 的处理完成。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"http:\/\/mpvideo.qpic.cn\/0b78paaauaaauyajyq2z3bqfa6gdbj4aacqa.f10002.mp4?dis_k=882c3b642a1e2c26f0a2e1e467784012&dis_t=1617263061&spec_id=MzAwODE3ODU5MA%3D%3D1617263062&vid=wxv_1793346667862130694&format_id=10002"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"除了上面展示的 Demo 以外，面向未来，MLSQL 在 Kylin 里可以和 Kylin API 进行对接，或者使用任务调度器 Airflow 等，就可以把写好的脚本保存并运转起来，其实就相当于已经完成了一个 ETL 的工具集了。用户无需自行搭建 ETL 平台，就能在 Kylin 页面完成整个流程。实现以后，用户可以直接对接 Kylin 数据，既可以使用 Kylin 进行加速，比如实现亚秒级响应速度，同时也可以写一个较为复杂的数据处理的流程，去查看数据的情况等。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"Kylin 与 MLSQL 更多联动方式"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"以上分享的这个 Demo 解决的核心问题是 ETL 能力，其实理论上，用户也是可以直接在刚刚的界面上做一些 AI 相关的操作。当用户拿到数据或通过 Kylin 得到一些分析结果，这个时候想去做一些算法相关的工作，就可以用到上面这种最简单的联动方式。其实 Kylin 和 MLSQL 联动的方式存在多种，这里简单介绍以下两种："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"第一种是前置联动"},{"type":"text","text":" ，如果这是一个 Pipeline 的话，那么 MLSQL 应该是在 Kylin 的前半部分， Kylin 是面向用户的或者说面向报表的，MLSQL 就可以帮助 Kylin 完成 ETL 的部分。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"第二种是后置联动"},{"type":"text","text":" ， MLSQL 本身是支持多数据源的，这也就意味着 MLSQL 可以连接 Kylin 的结果，再和 HDFS 或 HIVE 的结果再进行 Join 以及其他计算。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"今天我们所讲的主要是前置联动，也分为两种：浅联动和深联动。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/b9\/ec\/b9dda1218033491249af3c0e1391b8ec.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"本次 Demo 演示的就是浅前置联动，也就是正常情况下进行的 Pipeline，从各种数据源，经过 ETL 处理到 HIVE 或者 Kafka，之后由 Kylin 进行 Cube 的构建，再面向用户进行查询或者面向报表以及其他复杂的操作。这整个 ETL 的过程其实是可以通过 MLSQL Engine 去完成的，当然用户也可以直接使用一个比较好的控制台去 MLSQL Engine 里去执行任务，或者使用其他的引擎等。在本次 Demo 示例里面， "},{"type":"text","marks":[{"type":"strong"}],"text":"Kylin 用户可以在 Kylin UI Query 里面直接下发 ETL 任务到引擎层，去帮助 Cube 事先准备好数据，这就是浅联动"},{"type":"text","text":" ，因为它的影响面比较小。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/60\/7b\/600a70f76d66c380b239907299f9c57b.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二种是深前置联动。相较浅联动，深联动可以把 Pipeline 稍微调整一下，不仅可以下发 ETL 的任务到 MLSQL，甚至可以把 Cube 的构建任务也下发到 MLSQL。目前 Kylin 4.0 其实已经支持 Spark 引擎，所以把 Cube 任务给下沉到 MLSQL 实现起来也比较简单。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此处小剧透，有赞的郑生俊老师会分享在 Kylin 3.0 把 Cube 的任务内置到 Spark 引擎，并且用 K8s 去跑，已经被验证是可行的。这其实是一个比较深的联动，相当于把 Kylin 的 Cube 任务都下沉到 MLSQL 引擎层了。这也是更适合引擎层面去实现的，未来 MLSQL 既可以作为一个常驻的任务，也可以在 Kylin 去启动执行一个任务后，再把资源放掉。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"以上就是对深联动和浅联动的简单介绍。关于后置联动，未来会有相关技术博客和演示分享。从 MLSQL 社区的角度来讲，首先 MLSQL 社区接下来会尝试引入 Cube 的构建功能，也就是前面提及的与 Kylin 的深层次联动；其次，未来 MLSQL 会提供一个更友好的方式以供 Kylin 衔接；当然，我们也有可能给 Kylin 社区去贡献类似嵌入 ETL 能力的插件体系，不仅可以使用 MLSQL，Spark，还可以使用更多其他 ETL 工具，从而实现让用户可以在 Kylin 里就完成整个过程。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"作者介绍"},{"type":"text","text":"："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"祝海林（祝威廉），新一代开源数据与 AI 处理语言 MLSQL 作者，现为 Kyligence 技术合伙人 & 资深架构师，拥有 10+ 年研发经验，近 6 年专注于数据管理、商业分析、机器学习的统一平台的设计和开发。长期活跃在 Spark、Ray、Kylin 等多个开源社区，有多年的开源社区运营经验。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"本文转载自公众号ApacheKylin（ID：apachekylin）。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"原文链接"},{"type":"text","text":"："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/mp.weixin.qq.com\/s?__biz=MzAwODE3ODU5MA==&mid=2653082286&idx=1&sn=3c3b19fcc6ade0fc86cdc72d0a2e40c8&chksm=80a4ac5fb7d32549f25c726b539e99db400b8149a378d13e1d313fc173c98fa91ab67338113d&token=1020922772&lang=zh_CN#rd","title":"","type":null},"content":[{"type":"text","text":"MLSQL 一分钟让 Kylin 装备 ETL 能力！"}]}]}]}