Hive是如何解析SQL的呢,首先拿hive的建表语句来举例,比如下面的建表语句

create table test(id int,name string)row format delimited fields terminated by '\t';

然后使用hive的show create table语句来查看创建的表结构,这是一张text表

CREATE TABLE `test`(
`id` int,
`name` string)
ROW FORMAT SERDE
'org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe'
WITH SERDEPROPERTIES (
'field.delim'='\t',
'serialization.format'='\t')
STORED AS INPUTFORMAT
'org.apache.hadoop.mapred.TextInputFormat'
OUTPUTFORMAT
'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat'
LOCATION
'hdfs://master:8020/user/hive/warehouse/test'
TBLPROPERTIES (
'transient_lastDdlTime'='1568561230')

当然还有其他各种建表语句,比如

csv表

CREATE EXTERNAL TABLE `default.test_1`(
`key` string COMMENT 'from deserializer',
`value` string COMMENT 'from deserializer')
ROW FORMAT SERDE
'org.apache.hadoop.hive.serde2.OpenCSVSerde'
WITH SERDEPROPERTIES (
'escapeChar'='\\',
'quoteChar'='\'',
'separatorChar'='\t')
STORED AS INPUTFORMAT
'org.apache.hadoop.mapred.TextInputFormat'
OUTPUTFORMAT
'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat'
LOCATION
'hdfs://master:8020/user/hive/warehouse/test'
TBLPROPERTIES (
'COLUMN_STATS_ACCURATE'='false',
'numFiles'='0',
'numRows'='-1',
'rawDataSize'='-1',
'totalSize'='0',
'transient_lastDdlTime'='xxxx')

parquet表

CREATE TABLE `default.test`(
`time` string,
`server` int,
`id` bigint)
PARTITIONED BY (
`ds` string)
ROW FORMAT SERDE
'org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe'
WITH SERDEPROPERTIES (
'field.delim'='\t',
'serialization.format'='\t')
STORED AS INPUTFORMAT
'org.apache.hadoop.hive.ql.io.parquet.MapredParquetInputFormat'
OUTPUTFORMAT
'org.apache.hadoop.hive.ql.io.parquet.MapredParquetOutputFormat'
LOCATION
'hdfs://master:8020/user/hive/warehouse/test'
TBLPROPERTIES (
'transient_lastDdlTime'='xxxx')

json表

CREATE EXTERNAL TABLE `default.test`(
`titleid` string COMMENT 'from deserializer',
`timestamp` string COMMENT 'from deserializer')
ROW FORMAT SERDE
'org.openx.data.jsonserde.JsonSerDe'
STORED AS INPUTFORMAT
'org.apache.hadoop.mapred.TextInputFormat'
OUTPUTFORMAT
'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat'
LOCATION
'hdfs://master:8020/user/hive/warehouse/test'
TBLPROPERTIES (
'COLUMN_STATS_ACCURATE'='false',
'numFiles'='0',
'numRows'='-1',
'rawDataSize'='-1',
'totalSize'='0',

es表

CREATE EXTERNAL TABLE `default.test`(
`id` string COMMENT 'from deserializer',
`ts` string COMMENT 'from deserializer', ')
PARTITIONED BY (
`ds` string)
ROW FORMAT SERDE
'org.elasticsearch.hadoop.hive.EsSerDe'
STORED BY
'org.elasticsearch.hadoop.hive.EsStorageHandler'
WITH SERDEPROPERTIES (
'serialization.format'='1')
LOCATION
'hdfs://master:8020/user/hive/warehouse/test'
TBLPROPERTIES (
'es.index.auto.create'='yes',
'es.index.read.missing.as.empty'='yes',
'es.nodes'='host1,host2',
'es.port'='9200',
'es.resource'='index1/type1',

使用thrift的binary表

CREATE EXTERNAL TABLE `default.test`(
`bbb` string COMMENT 'from deserializer',
`aaa` string COMMENT 'from deserializer')
COMMENT 'aas'
PARTITIONED BY (
`ds` string COMMENT '日期分区')
ROW FORMAT SERDE
'org.apache.hadoop.hive.serde2.thrift.ThriftDeserializer'
WITH SERDEPROPERTIES (
'serialization.class'='com.xxx.xxx.xxx.tables.v1.XXXX',
'serialization.format'='org.apache.thrift.protocol.TCompactProtocol')
STORED AS INPUTFORMAT
'org.apache.hadoop.mapred.SequenceFileInputFormat'
OUTPUTFORMAT
'org.apache.hadoop.hive.ql.io.HiveSequenceFileOutputFormat'
LOCATION
'hdfs://master:8020/user/hive/warehouse/test'
TBLPROPERTIES (
'transient_lastDdlTime'='xxxxxx')

等等

可以查看show create table的hive源码

https://github.com/apache/hive/blob/68ae4a5cd1b916098dc1deb2bcede5f862afd80e/ql/src/java/org/apache/hadoop/hive/ql/ddl/table/creation/ShowCreateTableOperation.java

其中可以看出hive表的一些基本信息

private static final String CREATE_TABLE_TEMPLATE =
"CREATE <" + TEMPORARY + "><" + EXTERNAL + ">TABLE `<" + NAME + ">`(\n" +
"<" + LIST_COLUMNS + ">)\n" +
"<" + COMMENT + ">\n" +
"<" + PARTITIONS + ">\n" +
"<" + BUCKETS + ">\n" +
"<" + SKEWED + ">\n" +
"<" + ROW_FORMAT + ">\n" +
"<" + LOCATION_BLOCK + ">" +
"TBLPROPERTIES (\n" +
"<" + PROPERTIES + ">)\n"; private String getCreateTableCommand(Table table) {
ST command = new ST(CREATE_TABLE_TEMPLATE); command.add(NAME, desc.getTableName());
command.add(TEMPORARY, getTemporary(table));
command.add(EXTERNAL, getExternal(table));
command.add(LIST_COLUMNS, getColumns(table));
command.add(COMMENT, getComment(table));
command.add(PARTITIONS, getPartitions(table));
command.add(BUCKETS, getBuckets(table));
command.add(SKEWED, getSkewed(table));
command.add(ROW_FORMAT, getRowFormat(table));
command.add(LOCATION_BLOCK, getLocationBlock(table));
command.add(PROPERTIES, getProperties(table)); return command.render();
}

当用户输入一行create table语句的时候,可查看源码

https://github.com/apache/hive/blob/ff98efa7c6f2b241d8fddd0ac8dc55e817ecb234/ql/src/java/org/apache/hadoop/hive/ql/parse/ParseUtils.java

美团点评 Hive SQL的编译过程

https://tech.meituan.com/2014/02/12/hive-sql-to-mapreduce.html

其中可以看到,建表语句首先会使用antlr4将其转换成一颗语法树

public static ASTNode parse(String command) throws ParseException {
return parse(command, null);
}

然后可以使用getTable抽取其中的库名和表名

https://github.com/apache/hive/blob/f37c5de6c32b9395d1b34fa3c02ed06d1bfbf6eb/ql/src/java/org/apache/hadoop/hive/ql/parse/AnalyzeCommandUtils.java

源码

public static Table getTable(ASTNode tree, BaseSemanticAnalyzer sa) throws SemanticException {
String tableName = ColumnStatsSemanticAnalyzer.getUnescapedName((ASTNode) tree.getChild(0).getChild(0));
String currentDb = SessionState.get().getCurrentDatabase();
String [] names = Utilities.getDbTableName(currentDb, tableName);
return sa.getTable(names[0], names[1], true);
}
https://github.com/apache/hive/blob/master/ql/src/java/org/apache/hadoop/hive/ql/parse/ParseDriver.java
public ASTNode parse(String command) throws ParseException {
return parse(command, null);
}

然后比如要提取inputformat,outpurformat,serde和storageHandler

https://github.com/apache/hive/blob/f37c5de6c32b9395d1b34fa3c02ed06d1bfbf6eb/ql/src/java/org/apache/hadoop/hive/ql/parse/StorageFormat.java

源码

要提取字段信息,SkewedValue,表名以及row format

https://github.com/apache/hive/blob/f37c5de6c32b9395d1b34fa3c02ed06d1bfbf6eb/ql/src/java/org/apache/hadoop/hive/ql/parse/BaseSemanticAnalyzer.java

源码

public static List<FieldSchema> getColumns(
ASTNode ast, boolean lowerCase, TokenRewriteStream tokenRewriteStream,
List<SQLPrimaryKey> primaryKeys, List<SQLForeignKey> foreignKeys,
List<SQLUniqueConstraint> uniqueConstraints, List<SQLNotNullConstraint> notNullConstraints,
List<SQLDefaultConstraint> defaultConstraints, List<SQLCheckConstraint> checkConstraints,
Configuration conf) throws SemanticException {
我是源码
}

源码

 /**
* Get the unqualified name from a table node.
*
* This method works for table names qualified with their schema (e.g., "db.table")
* and table names without schema qualification. In both cases, it returns
* the table name without the schema.
*
* @param node the table node
* @return the table name without schema qualification
* (i.e., if name is "db.table" or "table", returns "table")
*/
public static String getUnescapedUnqualifiedTableName(ASTNode node) {
assert node.getChildCount() <= 2; if (node.getChildCount() == 2) {
node = (ASTNode) node.getChild(1);
} return getUnescapedName(node);
}

源码

  protected void analyzeRowFormat(ASTNode child) throws SemanticException {
child = (ASTNode) child.getChild(0);
int numChildRowFormat = child.getChildCount();
for (int numC = 0; numC < numChildRowFormat; numC++) {
ASTNode rowChild = (ASTNode) child.getChild(numC);
switch (rowChild.getToken().getType()) {
case HiveParser.TOK_TABLEROWFORMATFIELD:
fieldDelim = unescapeSQLString(rowChild.getChild(0)
.getText());
if (rowChild.getChildCount() >= 2) {
fieldEscape = unescapeSQLString(rowChild
.getChild(1).getText());
}
break;
case HiveParser.TOK_TABLEROWFORMATCOLLITEMS:
collItemDelim = unescapeSQLString(rowChild
.getChild(0).getText());
break;
case HiveParser.TOK_TABLEROWFORMATMAPKEYS:
mapKeyDelim = unescapeSQLString(rowChild.getChild(0)
.getText());
break;
case HiveParser.TOK_TABLEROWFORMATLINES:
lineDelim = unescapeSQLString(rowChild.getChild(0)
.getText());
if (!lineDelim.equals("\n")
&& !lineDelim.equals("10")) {
throw new SemanticException(SemanticAnalyzer.generateErrorMessage(rowChild,
ErrorMsg.LINES_TERMINATED_BY_NON_NEWLINE.getMsg()));
}
break;
case HiveParser.TOK_TABLEROWFORMATNULL:
nullFormat = unescapeSQLString(rowChild.getChild(0)
.getText());
break;
default:
throw new AssertionError("Unkown Token: " + rowChild);
}
}
}
}

分区信息,首先通过取得Map对象,

https://github.com/apache/hive/blob/6f18bbbc2e030ce7d446b2475037203cbd4f860d/ql/src/java/org/apache/hadoop/hive/ql/parse/AnalyzeCommandUtils.java

源码

  public static Map<String,String> getPartKeyValuePairsFromAST(Table tbl, ASTNode tree,
HiveConf hiveConf) throws SemanticException {
ASTNode child = ((ASTNode) tree.getChild(0).getChild(1));
Map<String,String> partSpec = new HashMap<String, String>();
if (child != null) {
partSpec = DDLSemanticAnalyzer.getValidatedPartSpec(tbl, child, hiveConf, false);
} //otherwise, it is the case of analyze table T compute statistics for columns;
return partSpec;
}

再转换成List<Partition>对象

https://github.com/apache/hive/blob/556531182dc989e12fd491d951b353b4df13fd47/ql/src/java/org/apache/hadoop/hive/ql/parse/BaseSemanticAnalyzer.java

源码

public Map<String, String> partSpec; // has to use LinkedHashMap to enforce order
public List<Partition> partitions; // involved partitions in TableScanOperator/FileSinkOperator
partitions = db.getPartitions(table, partSpec);

location信息,parsedLocation

https://github.com/apache/hive/blob/0213afb8a31af1f48d009edd41cec9e6c8942354/ql/src/java/org/apache/hadoop/hive/ql/parse/ImportSemanticAnalyzer.java

Hive学习笔记——parse的更多相关文章

  1. hive学习笔记之十:用户自定义聚合函数(UDAF)

    欢迎访问我的GitHub 这里分类和汇总了欣宸的全部原创(含配套源码):https://github.com/zq2599/blog_demos 本篇概览 本文是<hive学习笔记>的第十 ...

  2. hive学习笔记之一:基本数据类型

    欢迎访问我的GitHub https://github.com/zq2599/blog_demos 内容:所有原创文章分类汇总及配套源码,涉及Java.Docker.Kubernetes.DevOPS ...

  3. hive学习笔记之三:内部表和外部表

    欢迎访问我的GitHub https://github.com/zq2599/blog_demos 内容:所有原创文章分类汇总及配套源码,涉及Java.Docker.Kubernetes.DevOPS ...

  4. hive学习笔记之四:分区表

    欢迎访问我的GitHub https://github.com/zq2599/blog_demos 内容:所有原创文章分类汇总及配套源码,涉及Java.Docker.Kubernetes.DevOPS ...

  5. hive学习笔记之五:分桶

    欢迎访问我的GitHub https://github.com/zq2599/blog_demos 内容:所有原创文章分类汇总及配套源码,涉及Java.Docker.Kubernetes.DevOPS ...

  6. hive学习笔记之六:HiveQL基础

    欢迎访问我的GitHub https://github.com/zq2599/blog_demos 内容:所有原创文章分类汇总及配套源码,涉及Java.Docker.Kubernetes.DevOPS ...

  7. hive学习笔记之七:内置函数

    欢迎访问我的GitHub https://github.com/zq2599/blog_demos 内容:所有原创文章分类汇总及配套源码,涉及Java.Docker.Kubernetes.DevOPS ...

  8. hive学习笔记之九:基础UDF

    欢迎访问我的GitHub https://github.com/zq2599/blog_demos 内容:所有原创文章分类汇总及配套源码,涉及Java.Docker.Kubernetes.DevOPS ...

  9. hive学习笔记之十一:UDTF

    欢迎访问我的GitHub https://github.com/zq2599/blog_demos 内容:所有原创文章分类汇总及配套源码,涉及Java.Docker.Kubernetes.DevOPS ...

随机推荐

  1. Android adb连接后adb devices显示unauthorized解决办法

    adb链接手机后,显示unauthorized.原因是命令所在目录不对,这个adb devices命令是在夜神模拟器目录下执行的,所以出现该问题,切换目录后解决: 如果切换目录后仍然不行:重启adb. ...

  2. JPA(java持久化API)的环境的搭建

    因为我使用的是java工程 所以需要引入的依赖有: <properties> <project.build.sourceEncoding>UTF-8</project.b ...

  3. Graylog-centos安装

    graylog安装 1.先决条件 yum install java-1.8.0-openjdk-headless.x86_64 -y #安装java软件包 yum install epel-relea ...

  4. Beta冲刺博客汇总(校园帮-追光的人)

    所属课程 软件工程1916 作业要求 Beta冲刺博客汇总 团队名称 追光的人 作业目标 汇总Beta阶段的博客,方便查看 冲刺日志 Beta之前-凡事预则立(校园帮-追光的人)5-22 Beta冲刺 ...

  5. NameValueCollectionValueProvider

    NameValueCollectionValueProvider provider = new NameValueCollectionValueProvider(nameValueCollection ...

  6. 2019牛客暑期多校训练营(第九场)B:Quadratic equation (二次剩余求mod意义下二元一次方程)

    题意:给定p=1e9+7,A,B.  求一对X,Y,满足(X+Y)%P=A; 且(X*Y)%P=B: 思路:即,X^2-BX+CΞ0;  那么X=[B+-sqrt(B^2-4C)]/2: 全部部分都要 ...

  7. 前端性能----CDN

    Content Distribute Network(内容分发网络)是构建在网络之上的内容分发网络,依靠部署在各地的边缘服务器,通过中心平台的负载均衡.内容分发.调度等功能模块,使用户就近获取所需内容 ...

  8. idea 将java导出为可执行jar及导入jar依赖

    使用maven可以很好的帮助我们进行依赖的管理,也可以使用maven的jar包打包插件构建出可运行的jar.那针对不是用maven进行管理的普通java项目,可以通过以下方式导出可执行的jar包以及导 ...

  9. 【每天学一点linux】后台进程不打印日志

    command > out.file 2>&1 & 将文件输出到指定的文件中

  10. CF1245D: Shichikuji and Power Grid

    CF1245D: Shichikuji and Power Grid 题意描述: 给定\(n\)个点\((n\leq2000)\),在第\(i\)个点上建立一个基站需要\(c_i\)的代价,连接两个点 ...