Hive使用十六进制分隔符异常分析

Posted 2021-04-29 Hadoop实操

温馨提示：要看高清无码套图，请使用手机打开并单击图片放大查看。

1.问题描述

通过sqoop抽取mysql表数据到hive表，发现hive表所有列显示为null

Hive表的分隔符为“\u001B”,sqoop指定的分隔符也是“\u001B”

通过命令show create table test_hive_delimiter查看建表语句如下:

0: jdbc:hive2://localhost:10000/> show create table test_hive_delimiter;
...
INFO  : OK
+----------------------------------------------------+--+
|                   createtab_stmt                   |
+----------------------------------------------------+--+
| CREATE EXTERNAL TABLE `test_hive_delimiter`(       |
|   `id` int,                                        |
|   `name` string,                                   |
|   `address` string)                                |
| ROW FORMAT SERDE                                   |
|   'org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe'  |
| WITH SERDEPROPERTIES (                             |
|   'field.delim'='\u0015',                          |
|   'serialization.format'='\u0015')                 |
| STORED AS INPUTFORMAT                              |
|   'org.apache.hadoop.mapred.TextInputFormat'       |
| OUTPUTFORMAT                                       |
|   'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat' |
| LOCATION                                           |
|   'hdfs://ip-172-31-6-148.fayson.com:8020/fayson/test_hive_delimiter' |
| TBLPROPERTIES (                                    |
|   'COLUMN_STATS_ACCURATE'='false',                 |
|   'numFiles'='0',                                  |
|   'numRows'='-1',                                  |
|   'rawDataSize'='-1',                              |
|   'totalSize'='0',                                 |
|   'transient_lastDdlTime'='1504705887')            |
+----------------------------------------------------+--+
22 rows selected (0.084 seconds)
0: jdbc:hive2://localhost:10000/> 

发现Hive的原始建表语句中的分隔符是“\u001B”而通过show create table test_hive_delimiter命令查询出来的分隔符为“\u0015”，分隔符被修改了。

2.问题复现

1.创建Hive表test_hive_delimiter，使用“\u001B”分隔符

create external table test_hive_delimiter
(
  id int,
  name string,
  address string
)
row format delimited fields terminated by '\u001B'
stored as textfile location '/fayson/test_hive_delimiter';

2.使用sqoop抽取MySQL中test表数据到hive表（test_hive_delimiter）

 [root@ip-172-31-6-148 ~]# sqoop import --connect jdbc:mysql://ip-172-31-6-148.fayson.com:3306/fayson -username root -password 123456 --table test -m 1 --hive-import --fields-terminated-by "\0x001B" --target-dir /fayson/test_hive_delimiter --hive-table test_hive_delimiter

数据抽取成功：

[root@ip-172-31-6-148 ~]# hadoop fs -ls /fayson/test_hive_delimiter
Found 2 items
-rw-r--r--   3 fayson supergroup          0 2017-09-06 13:46 /fayson/test_hive_delimiter/_SUCCESS
-rwxr-xr-x   3 fayson supergroup         56 2017-09-06 13:46 /fayson/test_hive_delimiter/part-m-00000
[root@ip-172-31-6-148 ~]# hadoop fs -ls /fayson/test_hive_delimiter/part-m-00000
-rwxr-xr-x   3 fayson supergroup         56 2017-09-06 13:46 /fayson/test_hive_delimiter/part-m-00000
[root@ip-172-31-6-148 ~]# 

3.查看test_hive_delimiter表数据

[root@ip-172-31-6-148 ~]# beeline 
Beeline version 1.1.0-cdh5.12.1 by Apache Hive
beeline> !connect jdbc:hive2://localhost:10000/;principal=hive/[e-mail]
...
Transaction isolation: TRANSACTION_REPEATABLE_READ
0: jdbc:hive2://localhost:10000/> select * from test_hive_delimiter;
...
INFO  : OK
+-------------------------+---------------------------+------------------------------+--+
| test_hive_delimiter.id  | test_hive_delimiter.name  | test_hive_delimiter.address  |
+-------------------------+---------------------------+------------------------------+--+
| NULL                    | NULL                      | NULL                         |
| NULL                    | NULL                      | NULL                         |
| NULL                    | NULL                      | NULL                         |
+-------------------------+---------------------------+------------------------------+--+
3 rows selected (0.287 seconds)
0: jdbc:hive2://localhost:10000/> 

4.Hive表的建表语句如下

3.解决方法

分隔符“\u001B”为十六进制，而Hive的分隔符实际是八进制，所以在使用十六进制的分隔符时会被Hive转义，所以出现使用“\u001B”分隔符创建hive表后显示的分隔符为“\u0015”。

在不改变数据文件分隔符的情况下，要先将十六进制分隔符转换成八进制分隔符来创建Hive表。

1.将十六进制分隔符转换为八进制分隔符

“\u001B”转换八进制为“\033”,在线转换工具：http://tool.lu/hexconvert/

2.修改建表语句使用八进制“\033”作为分隔符

create external table test_hive_delimiter
(
  id int,
  name string,
  address string
)
row format delimited fields terminated by '\033'
stored as textfile location '/fayson/test_hive_delimiter';

使用命令show create table test_hive_delimiter查看建表语句

0: jdbc:hive2://localhost:10000/> show create table test_hive_delimiter;
...
INFO  : OK
+----------------------------------------------------+--+
|                   createtab_stmt                   |
+----------------------------------------------------+--+
| CREATE EXTERNAL TABLE `test_hive_delimiter`(       |
|   `id` int,                                        |
|   `name` string,                                   |
|   `address` string)                                |
| ROW FORMAT SERDE                                   |
|   'org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe'  |
| WITH SERDEPROPERTIES (                             |
|   'field.delim'='\u001B',                          |
|   'serialization.format'='\u001B')                 |
| STORED AS INPUTFORMAT                              |
|   'org.apache.hadoop.mapred.TextInputFormat'       |
| OUTPUTFORMAT                                       |
|   'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat' |
| LOCATION                                           |
|   'hdfs://ip-172-31-6-148.fayson.com:8020/fayson/test_hive_delimiter' |
| TBLPROPERTIES (                                    |
|   'COLUMN_STATS_ACCURATE'='false',                 |
|   'numFiles'='0',                                  |
|   'numRows'='-1',                                  |
|   'rawDataSize'='-1',                              |
|   'totalSize'='0',                                 |
|   'transient_lastDdlTime'='1504707693')            |
+----------------------------------------------------+--+
22 rows selected (0.079 seconds)
0: jdbc:hive2://localhost:10000/> 

3.查询test_hive_delimiter表数据

0: jdbc:hive2://localhost:10000/> select * from test_hive_delimiter;
...
INFO  : OK
+-------------------------+---------------------------+------------------------------+--+
| test_hive_delimiter.id  | test_hive_delimiter.name  | test_hive_delimiter.address  |
+-------------------------+---------------------------+------------------------------+--+
| 1                       | fayson                    | guangdong                    |
| 2                       | zhangsan                  | shenzheng                    |
| 3                       | lisi                      | shanghai                     |
+-------------------------+---------------------------+------------------------------+--+
3 rows selected (0.107 seconds)
0: jdbc:hive2://localhost:10000/> 

将十六进制的”\u001B”转换为八进制的”\033”建表，问题解决。

4.备注

1. Hive建表时使用十六进制分割符需要注意，部分分隔符会被转义（如：001B/001C等）

2. Sqoop指定十六进制分隔符，为什么是“\0x001B”而不是“\u001B”,可参考Sqoop

官网说明：https://sqoop.apache.org/docs/1.4.6/SqoopUserGuide.html#_file_formats

醉酒鞭名马，少年多浮夸！ 岭南浣溪沙，呕吐酒肆下！挚友不肯放，数据玩的花！

温馨提示：要看高清无码套图，请使用手机打开并单击图片放大查看。

推荐关注Hadoop实操，第一时间，分享更多Hadoop干货，欢迎转发和分享。