动笔写下这个章节的第一个字时，是我在 GitHub 上泡了一整天之后。Good First Issues 和 What are we doing now 这两个重要的 List 已经近半年没有人维护了，我根据自己还只是 Contributor 时整理的笔记，将零散的代码改进点创建为新的 good first issue，供想要参与开源的新人练手。
令我意外的是，就在我建了两个低难度的 issue 之后，十分钟不到就已经双双有人认领。有新面孔来做 good first issue 的时候仿佛已经过去很久了。Good First Issue List 中的任务大多都早已有人认领，但提交的 PR 在被 Review 之后就杳无音信，又或是没有提交过 PR。
因为开源之夏，你已经认识到了去做开源的这件事情。如果你不是那种需要 deadline 推着走的 type 的话，按照你的节奏去加入社区就可以了。大多数参加开源之夏或者 XX 比赛的同学都只是将课题视为一个任务，没有持续地做下去。事实证明能不能获得 Committer 提名也与比赛有没有获奖根本没有关系。贡献的多少是由社区的大家看在眼里的，只要你能在一个社区有持续性地贡献，都能被认可，并成为相应模块的专家。
《Deep Dive the replication protocol in Kafka》这一场讲的真的特别好，来自台湾的 Luke Chen 老师用最直白的数组深入浅出地讲解了 ISR (In-Sync Replica) 的种种 Corner Case 及其应对机制，让我感受到了 Red Hat Engineer 的魅力。
Sort the User objects list. Only works using method reference, with lambda expression the compiler gives an error:
1 2 3 4
List<User> userList = Arrays.asList(u1, u2, u3); userList.sort(Comparator.comparing(u -> u.getName())); // works userList.sort(Comparator.comparing(User::getName).reversed()); // works userList.sort(Comparator.comparing(u -> u.getName()).reversed()); // Compiler error
Error:
1 2 3 4 5 6
com\java8\collectionapi\CollectionTest.java:35: error: cannot find symbol userList.sort(Comparator.comparing(u -> u.getName()).reversed()); ^ symbol: method getName() location: variable u of type Object 1 error
Reason
This is a weakness in the compiler’s type inferencing mechanism. In order to infer the type of u in the lambda, the target type for the lambda needs to be established. This is accomplished as follows. userList.sort() is expecting an argument of type Comparator<User>. In the first line, Comparator.comparing() needs to return Comparator<User>. This implies that Comparator.comparing() needs a Function that takes a User argument. Thus in the lambda on the first line, u must be of type User and everything works.
In the second and third lines, the target typing is disrupted by the presence of the call to reversed(). Both the receiver and the return type of reversed() are Comparator<T> so it seems like the target type should be propagated back to the receiver, but it isn’t. (Like I said, it’s a weakness.)
Lambdas are divided into implicitly-typed (no manifest types for parameters) and explicitly-typed; method references are divided into exact (no overloads) and inexact. When a generic method call in a receiver position has lambda arguments, and the type parameters cannot be fully inferred from the other arguments, you need to provide either an explicit lambda, an exact method ref, a target type cast, or explicit type witnesses for the generic method call to provide the additional type information needed to proceed.
In the second line, the method reference provides additional type information that fills this gap. This information is absent from the third line, so the compiler infers u to be Object (the inference fallback of last resort), which fails.
Usage
This would work fine if you have not chained the comparators:
However, when comparators are chained, the type of the objects being compared need to be specified explicitly. Obviously if you can use a method reference, do that and it’ll work. Sometimes you can’t use a method reference, for example when you reference non-static methods from static contexts.
If you want to pass an additional parameter, so you have to use a lambda expression. In that case you’d provide type in comparing:
com\java8\collectionapi\CollectionTest.java:35: error: cannot find symbol userList.sort(Comparator.comparing(u -> u.getName()).reversed()); ^ symbol: method getName() location: variable u of type Object 1 error
解释
这是编译器的类型推断机制的一个弱点。为了推断 lambda 中 u 的类型，需要建立 lambda 的目标类型。具体过程是，userList.sort() 期望一个类型为 Comparator<User> 的参数。在第一行中，Comparator.comparing() 需要返回 Comparator<User>。这意味着 Comparator.comparing() 需要一个以 User 为参数的 Function。因此，在第一行的 lambda 表达式中，u 必须是 User 类型。
@Test publicvoidtestGetMappingPattern() { assertEquals("/test/{param1}/path/{param2}", urlMappingPattern.getMappingPattern()); } @Test publicvoidtestCompile()throws NoSuchFieldException, IllegalAccessException { //TODO : Fix me to test the method compile(). It is better using Mockito not PowerMockito. }
@Test publicvoidtestCompile()throws NoSuchFieldException, IllegalAccessException { // Obtain compiledUrlMappingPattern field with reflection FieldcompiledUrlMappingPatternField= UrlMappingPattern.class.getDeclaredField("compiledUrlMappingPattern"); compiledUrlMappingPatternField.setAccessible(true); urlMappingPattern.compile(); // Verify that the compiledUrlMappingPattern field is updated PatterncompiledPattern= (Pattern) compiledUrlMappingPatternField.get(urlMappingPattern); assertNotNull(compiledPattern); // Verify that the mocked pattern is compiled with the expected regex Mockito.verify(urlMappingPattern.compiledUrlMappingPattern) .compile("/test/([%\\w-.\\~!$&'\\(\\)\\*\\+,;=:\\[\\]@]+?)/path/([%\\w-.\\~!$&'\\(\\)\\*\\+,;=:\\[\\]@]+?)(?:\\?.*?)?$"); }
但是跑测试类时，会使别的测试方法报错（虽然报错信息依然指向这里）：
1 2 3 4 5 6 7 8 9 10 11 12 13
org.mockito.exceptions.misusing.UnfinishedVerificationException: Missing method call for verify(mock) here: -> at org.apache.eventmesh.admin.rocketmq.util.UrlMappingPatternTest.testCompile(UrlMappingPatternTest.java:95)
Example of correct verification: verify(mock).doSomething()
Also, this error might show up because you verify either of: final/private/equals()/hashCode() methods. Those methods *cannot* be stubbed/verified. Mocking methods declared on non-public parent classes is not supported.
at org.apache.eventmesh.admin.rocketmq.util.UrlMappingPatternTest$TestUrlMappingPattern.<init>(UrlMappingPatternTest.java:105) at org.apache.eventmesh.admin.rocketmq.util.UrlMappingPatternTest.setUp(UrlMappingPatternTest.java:44)
@Test publicvoidtestCompile()throws NoSuchFieldException, IllegalAccessException { // Obtain compiledUrlMappingPattern field with reflection FieldcompiledUrlMappingPatternField= UrlMappingPattern.class.getDeclaredField("compiledUrlMappingPattern"); compiledUrlMappingPatternField.setAccessible(true);
urlMappingPattern.compile();
// Verify that the compiledUrlMappingPattern field is updated PatterncompiledPattern= (Pattern) compiledUrlMappingPatternField.get(urlMappingPattern); assertNotNull(compiledPattern);
// Verify that the mocked pattern is compiled with the expected regex StringexpectedRegex="/test/([%\\w-.\\~!$&'\\(\\)\\*\\+,;=:\\[\\]@]+?)/path/([%\\w-.\\~!$&'\\(\\)\\*\\+,;=:\\[\\]@]+?)(?:\\?.*?)?$"; PatternexpectedPattern= Pattern.compile(expectedRegex); assertEquals(expectedPattern.pattern(), compiledPattern.pattern()); }
插件化的方式最主要的优点在于方便，可以让所有的使用者根据具体的基础设施方案去使用不同的插件。eventmesh-runtime 是一种微内核的设计，只需要给 runtime 设计好它的运行态，其它的功能都从 API 去使用。定义好不同的 API 之后，就可以进一步定义不同的插件了，比如注册中心模块、存储模块等。
从 EventMesh 内部层次的角度看，消息一开始会进入 Server 层，然后利用 SPI 反射机制，在 API 层利用异步回调去调用存储层的事件存储插件，一方面向事件存储发送消息和回调，另一方面向 API 层回调，等到事件存储返回执行的结果之后，将结果等信息返回给 API，然后再由 API 把执行结果原路返回调用方。
//TODO : Fix me to test the method compile(). It is better using Mockito not PowerMockito.
使用 Mockito，并通过 Java 反射获取私有字段，撰写正则表达式，示例代码如下：
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
@Test publicvoidtestCompile()throws NoSuchFieldException, IllegalAccessException { // Mock the compiledUrlMappingPattern field with reflection PatternmockedPattern= mock(Pattern.class); FieldcompiledUrlMappingPatternField= urlMappingPattern.getClass().getDeclaredField("compiledUrlMappingPattern"); compiledUrlMappingPatternField.setAccessible(true); compiledUrlMappingPatternField.set(urlMappingPattern, mockedPattern);
urlMappingPattern.compile();
// Verify that the compiledUrlMappingPattern field is updated assertEquals(mockedPattern, compiledUrlMappingPatternField.get(urlMappingPattern));
// Verify that the mocked pattern is compiled with the expected regex Mockito.verify(mockedPattern).compile("/test/([%\\w-.\\~!$&'\\(\\)\\*\\+,;=:\\[\\]@]+?)/path/([%\\w-.\\~!$&'\\(\\)\\*\\+,;=:\\[\\]@]+?)(?:\\?.*?)?$"); }
If a thread is interrupted while in a blocked state, in order to prevent subsequent code from making erroneous judgments after handling the interruption, the InterruptedException is thrown and at the same time, the Thread.interrupted() method is called to clear the thread’s interrupt status.
If the dispatcher is interrupted while executing the take() method, it will catch the InterruptedException exception and continue executing the exception handling block. At this point, the information about the thread being interrupted should not be lost. However, because InterruptedException is a checked exception, if it is not handled, it will be propagated to the method caller and may be thrown in EventMeshHTTPServer.java, resulting in a delayed closure of the thread. Moreover, there is no timely exit or restoration of the interrupt status at this point, so the thread may continue executing the retry() method, thereby losing the information about the thread being interrupted.
However, the issue states that InterruptedExceptions should be re-thrown, but I think this handling is inappropriate. This approach does not reset the thread’s interrupt status nor does it communicate the intention and semantics of the thread being interrupted to other developers. It would be better to reset the interrupt status.
捕获 InterruptedException 异常并重新中断线程：
1 2 3 4 5 6
} catch (Exception e) { if (e instanceof InterruptedException) { Thread.currentThread().interrupt(); } retryLogger.error("http-retry-dispatcher error!", e); }
SELECT ( SELECT ID FROM SYS_MENU smps WHERE NAME ='My Report Name' AND HREF ='/path/to/my/report' ) AS MENU_ID, 3AS "TYPE", ( SELECT ID FROM SYS_USER sus WHERE USER_NAME ='myUsername' ) AS OWNER FROM DUMMY;
SELECT ( SELECT ID FROM SYS_MENU smps WHERE NAME ='My Report Name' AND HREF ='/path/to/my/report' LIMIT 1) AS MENU_ID, 3AS "TYPE", ( SELECT ID FROM SYS_USER sus WHERE USER_NAME ='myUsername' LIMIT 1) AS OWNER FROM DUMMY;
SELECT smps.ID AS MENU_ID, 3AS "TYPE", sus.ID AS OWNER FROM SYS_MENU smps LEFTJOIN SYS_USER sus ON sus.USER_NAME ='myUsername' WHERE smps.NAME ='My Report Name' AND smps.HREF ='/path/to/my/report';
然而，我需要进一步的为 smps.ID 和 sus.ID 添加 COALESCE 函数，当 smps.ID 或 sus.ID 为 null 时，可以使用 USER_NAME 或 HREF 作为替代值。使用 LEFT JOIN 后，当右侧条件 WHERE LOWER(USER_NAME) = LOWER('myUsername') 匹配不到结果时，sus_ID 可以显示替代值，当左侧条件 WHERE NAME = 'My Report Name' 在表中没有任何匹配结果时，整个查询只会返回空结果集，就无法用 COALESCE 函数对 MENU_ID 插入替代值了。即使使用 FULL JOIN 或笛卡尔积 CROSS JOIN 也无法解决任何一边的结果集为空的情况。
用 UNION ALL 确保不返回空
为了解决这个问题，可以用 LEFT JOIN 将两个子查询连接在一起，并且 ON 条件始终为 true。然后在每个子查询的末尾添加一个 UNION ALL 子句，如果没有检查到匹配结果，就会通过 DUMMY 表添加一行带有 NULL 值的记录。这样即可确保即使 URL 没有匹配结果，仍然会返回带有替代值的结果。
SELECT COALESCE(TO_CHAR(smps.ID), '/path/to/my/report') AS MENU_ID, COALESCE(TO_CHAR(sus.ID), 'myUsername') AS OWNER FROM (SELECT ID FROM SYS_MENU WHERE NAME ='My Report Name' AND HREF ='/path/to/my/report' UNIONALL SELECTNULLAS ID FROM DUMMY WHERENOTEXISTS ( SELECT ID FROM SYS_MENU WHERE NAME ='My Report Name' AND HREF ='/path/to/my/report' )) AS smps LEFTJOIN (SELECT ID FROM SYS_USER WHERELOWER(USER_NAME) =LOWER('myUsername') UNIONALL SELECTNULLAS ID FROM DUMMY WHERENOTEXISTS ( SELECT ID FROM SYS_USER WHERELOWER(USER_NAME) =LOWER('myUsername') )) AS sus ON1=1;
最终我使用的 SQL 语句如下，再结合 Excel 公式和 SQL 转义，它完美地帮我将一份报表名称与URL重复/不对应、用户名不准确/大小写混淆、存在隐性的/无法匹配的Unicode字符的 xlsx 表格维护的 MENU 与 USER 的关系导入到了数据库中：上万数据一遍跑通，新数据更新旧数据，哪里匹配不上就留下问题定位信息，一对多记录也一并插入，还可以重复导入，最后手工纠正一下替代值，甚是畅快。
MERGE INTO SYS_MENU_USER AS target USING ( SELECT COALESCE(TO_CHAR(smps.ID), '/path/to/my/report') AS MENU_ID, 3AS "TYPE", COALESCE(TO_CHAR(sus.ID), 'myUsername') AS OWNER FROM ( SELECT ID FROM SYS_MENU WHERE NAME ='My Report Name' AND HREF ='/path/to/my/report' UNIONALL SELECT NULLAS ID FROM DUMMY WHERE NOTEXISTS ( SELECT ID FROM SYS_MENU WHERE NAME ='My Report Name' AND HREF ='/path/to/my/report' )) AS smps LEFTJOIN ( SELECT ID FROM SYS_USER WHERE LOWER(USER_NAME) =LOWER('myUsername') UNIONALL SELECT NULLAS ID FROM DUMMY WHERE NOTEXISTS ( SELECT ID FROM SYS_USER WHERE LOWER(USER_NAME) =LOWER('myUsername') )) AS sus ON 1=1) AS SOURCE ON target.MENU_ID = source.MENU_ID AND target."TYPE" = source."TYPE" WHEN MATCHED THEN UPDATE SET OWNER = source.OWNER WHENNOT MATCHED THEN INSERT (MENU_ID, "TYPE", OWNER) VALUES (source.MENU_ID, source."TYPE", source.OWNER);
中间件一般对跨库 join 都支持不好。在分库分表中，要同步更新两个表，这两个表位于不同的物理库中，为了保证数据一致性，一种做法是通过分布式事务中间件将两个更新操作放到一个事务中，但这样的操作一般要加全局锁，性能很捉急，而有些业务能够容忍短暂的数据不一致，怎么做？让它们分别更新呗，但是会存在数据写失败的问题，那就起个定时任务，扫描下 A 表有没有失败的行，然后看看 B 表是不是也没写成功，然后对这两条关联记录做订正，这个时候同样没法用 join 去实现，只能将数据拉到 service 层应用自己来合并了
这些规则都是互联网开发团队总结出来的，适用于高并发、轻写重读、分布式、业务逻辑简单的情况，甚至对数据的一致性要求都不高，允许脏读（毕竟 web 是个非实时、无状态的东西）。
在我的 Team 中，有很多 DB 同事，为 Java 开发写好了 SAP HANA 数据库的存储过程，进行了大量数据清洗，理应利用高效的内存数据库优势。此外，由于 SCI 系统的内存资源比较紧张，一个微服务在容器云上目前最多只能分配 10G，同时业务表数据量比较巨大，一个 version 几千万的数据是很正常的，如果分步查询，每个步骤的获取的结果都需要拿到 JVM 内存中与下一步的结果去做处理，会比较占用内存。因此很多报表的处理思路都是能在 SQL 中解决的就在 SQL 中解决，Java 代码中不做复杂的逻辑。
LR (0)：使用状态机来模拟堆栈 (LR 分析法，它的模型时什么？根据 LR 的分析表来判断四个动作，ACCEPT、空白出错等。规约（必会）：栈顶形成句柄时要将句柄弹出来，将栈顶的字符规约成一个符号，此时需要查 GOTO 表，这决定了新的字符进入新的栈顶之后，新的字符是什么字段) - r 规约 s 移进 accept:1， 空白：出错
移进和规约冲突等，LR (0) 就不行了
SLR
规约与…
LR(1)
推导项目集（作业里有，考试时没有那么难）
LALR
LR (1) 会生成大量的项目集，爆炸式增长
第六章
S 属性、L 属性、为什么要定义 L 属性
L 属性的引入是为了处理依赖于上下文信息的语法结构和语义规则。它允许从产生式的右部符号向左部符号传递属性值，以满足上下文相关的计算需求。在某些语法结构中，仅使用 S 属性无法满足对上下文信息的处理，而 L 属性可以提供更灵活的属性传递机制，支持对上下文信息的分析和计算。
都是平台给需求，中间有个 DT 岗拆解需求，然后给 PA 继续拆解，BA 来做原型逻辑设计，然后是 IT 团队写代码，代码完了再内测，UAT，安全审查等等，一套流程下来一个大需求差不多也得几个月上线哈哈
另外开发都得会 jira，一般都是在 jira 上沟通问题”
岗位简写解释
PM: Product & Marketing 产品经理
RD: Research and Development Engineer 研发工程师。对某种不存在的事物进行系统的研究和开发并具有一定经验的专业工作者，或者对已经存在的事物进行改进以达到优化目的的专业工作者。
QA: Quality Assurance Engineer
DT: Data Technology Business Process Design “根据业务情况梳理客户 LTV 经营业务流程，挖掘客户经营机会和提升点；规划客户 LTV 经营产品平台，包括客户数据分析的数据产品规划和面向运营和营销的平台工具规划”Digital Transformation 数字化转型，通常由 Data Analyst 数据分析岗位负责。
PA: Project Administrator
BA: Business Analyst 商务分析师。Business Analytic 和 Business Analysis 有点不一样，一个偏数据分析，一个偏业务和需求分析。（天津 Site 没有此岗位）
需求处理流程
原型是由 DT 同事和 UI/UE 同事给出来的
BSR: Business Solution Review。业务沟通阶段，一般发生在需求一开始，DT 会和业务沟通看下都是要做什么，这个给出来的文档就是和业务沟通的结果。
ISR: It Solution Review。基于 BSR 和 IT 同事沟通解决方案的文档，比如说涉及到哪些流程，和哪些系统交互，会用到什么东西。
NOHUP(1) User Commands NOHUP(1) NAME nohup - run a command immune to hangups, with output to a non-tty SYNOPSIS nohup COMMAND [ARG]... nohup OPTION DESCRIPTION Run COMMAND, ignoring hangup signals. --help display this help and exit --version output version information and exit
SETSID(8) Linux Programmer’s Manual SETSID(8) NAME setsid - run a program in a new session SYNOPSIS setsid program [ arg ... ] DESCRIPTION setsid runs a program in a new session.
disown [-ar] [-h] [jobspec ...] Without options, each jobspec is removed from the table of active jobs. If the -h option is given, each jobspec is not removed from the table, but is marked so that SIGHUP is not sent to the job if the shell receives a SIGHUP. If no jobspec is present, and neither the -a nor the -r option is supplied, the current job is used. If no jobspec is supplied, the -a option means to remove or mark all jobs; the -r option without a jobspec argument restricts operation to running jobs. The return value is 0 unless a jobspec does not specify a valid job.
用 disown -h JOBSPEC 来使某个作业忽略 HUP 信号。
用 disown -ah 来使所有的作业都忽略 HUP 信号。
用 disown -rh 来使正在运行的作业忽略 HUP 信号。
需要注意的是，当使用过 disown 之后，会将把目标作业从作业列表中移除，我们将不能再使用 jobs 来查看它，但 ps 依然可以查找到它。
这种方法的操作对象是作业，如果需要通过 jobs 命令得到作业列表，在运行命令时应该在结尾添加 & 来使命令成为一个作业并在后台运行。
如果没有把命令作为作业来运行，我们需要按下 Ctrl+Z 将前台进程送入后台并挂起，然后才可以用 jobs 命令查询它的作业号，接着用 bg %JOBSPEC 让它继续运行。需要注意的是，如果挂起会影响当前进程的运行结果，请慎用此方法。
]]><blockquote>
<p>用 ssh 登录远程的 Linux 服务器，某些程序或脚本会在前台持续运行、占用 shell 窗口，且终端连接断开时任务也会中止。如何让运行命令提交后不受本地关闭终端窗口或网络断开连接的干扰呢？本文列举了一些实用的命令。</p>
</blockqu
