通常用户在HIVE中用SELECT语句出来结果，无法确定结果是来自哪个文件或者具体位置信息，HIVE中考虑到了这点，在Virtual Column虚列中可以指定三个静态列:

1. INPUT__FILE__NAME map任务读入File的全路径

2. BLOCK__OFFSET__INSIDE__FILE 如果是RCFile或者是SequenceFile块压缩格式文件则显示Block file Offset，也就是当前快在文件的第一个字偏移量，如果是TextFile，显示当前行的第一个字节在文件中的偏移量

3. ROW__OFFSET__INSIDE__BLOCK RCFile和SequenceFile显示row number, textfile显示为0

注：若要显示ROW__OFFSET__INSIDE__BLOCK ，必须设置set hive.exec.rowoffset=true;

测试：

1.

table: test_virtual_columns

InputFormat: org.apache.hadoop.mapred.TextInputFormat

query:

select a, INPUT__FILE__NAME,BLOCK__OFFSET__INSIDE__FILE,ROW__OFFSET__INSIDE__BLOCK from test_virtual_columns;

result:

qweqwe  hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t3.txt        0       0
dfdf    hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t3.txt        7       0
sdafsafsaf      hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t3.txt        12      0
dfdffd  hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t3.txt        23      0
dsf     hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t3.txt        30      0
1       hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t1.txt        0       0
2       hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t1.txt        2       0
3       hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t1.txt        4       0
4       hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t1.txt        6       0
5       hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t1.txt        8       0
6       hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t1.txt        10      0
7       hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t1.txt        12      0
8       hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t2.txt        0       0
9       hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t2.txt        2       0
10      hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t2.txt        4       0
11      hdfs://10.2.6.102/user/hive/warehouse/tmp.db/test_virtual_columns/t2.txt        7       0

2.

table: nginx

InputFormat: org.apache.hadoop.hive.ql.io.RCFileInputFormat

query:

select hostname, INPUT__FILE__NAME,BLOCK__OFFSET__INSIDE__FILE,ROW__OFFSET__INSIDE__BLOCK from nginx where dt='2013-09-01' limit 10;

result:

10.1.2.162      hdfs://10.2.6.102/share/data/log/nginx_rcfile/2013-09-01/000000_0       537155468       0
10.1.2.162      hdfs://10.2.6.102/share/data/log/nginx_rcfile/2013-09-01/000000_0       537155468       1
10.1.2.162      hdfs://10.2.6.102/share/data/log/nginx_rcfile/2013-09-01/000000_0       537155468       2
10.1.2.162      hdfs://10.2.6.102/share/data/log/nginx_rcfile/2013-09-01/000000_0       537155468       3
10.1.2.162      hdfs://10.2.6.102/share/data/log/nginx_rcfile/2013-09-01/000000_0       537155468       4
10.1.2.162      hdfs://10.2.6.102/share/data/log/nginx_rcfile/2013-09-01/000000_0       537155468       5
10.1.2.162      hdfs://10.2.6.102/share/data/log/nginx_rcfile/2013-09-01/000000_0       537155468       6
10.1.2.162      hdfs://10.2.6.102/share/data/log/nginx_rcfile/2013-09-01/000000_0       537155468       7
10.1.2.162      hdfs://10.2.6.102/share/data/log/nginx_rcfile/2013-09-01/000000_0       537155468       8
10.1.2.162      hdfs://10.2.6.102/share/data/log/nginx_rcfile/2013-09-01/000000_0       537155468       9

如果碰到有脏数据或者结果异常的时候，可以通过select这三个值来定位出错的原始文件和具体所在位置，很方便。

本文链接 http://blog.csdn.net/lalaguozhe/article/details/11467593，转载请注明

ITPub博客

Hive中查看数据来源文件和具体位置方法

最新文章