yucai created HIVE-21042:
----------------------------
Summary: varchar and string will behave differently for
non-standard utf8 characters
Key: HIVE-21042
URL: https://issues.apache.org/jira/browse/HIVE-21042
Project: Hive
Issue Type: Bug
Reporter: yucai
If data contains non-standard utf8 characters, varchar and string will behave
differently.
The content in /tmp/hex_data is
"6130373530633166313366306B35*B0A5*46386A*8DAEAB*62B4526F273464613936".
*B0A5 and* *8DAEAB* are non-standard utf8 characters, they are encoded to
EFBFBD if the data type is varchar, which string will not change it. So:
VARCHAR shows:
6130373530633166313366306B35EFBFBDEFBFBD46386AEFBFBDEFBFBDEFBFBD62EFBFBD526F273464613936
STRING shows: 6130373530633166313366306B35B0A546386A8DAEAB62B4526F273464613936
See details:
{code:java}
hive> DROP TABLE TBL_S;
OK
Time taken: 0.562 seconds
hive> CREATE TABLE TBL_S
> (
> GUID STRING
> )
> row format delimited fields terminated by '\177' stored as textfile
> LOCATION
> '/tmp/hex_data'
> tblproperties('serialization.null.format'='', 'timestamp.formats' =
'yyyy-MM-dd HH:mm:ss')
> ;
OK
Time taken: 3.074 seconds
hive>
> DROP TABLE TBL_V;
OK
Time taken: 0.894 seconds
hive> CREATE TABLE TBL_V
> (
> GUID VARCHAR(32)
> )
> row format delimited fields terminated by '\177' stored as textfile
> LOCATION
> '/tmp/hex_data'
> tblproperties('serialization.null.format'='', 'timestamp.formats' =
'yyyy-MM-dd HH:mm:ss')
> ;
OK
Time taken: 0.242 seconds
hive> SELECT GUID, hex(GUID) FROM TBL_S;
OK
a0750c1f13f0k5��F8j���b�Ro'4da96
6130373530633166313366306B35B0A546386A8DAEAB62B4526F273464613936
Time taken: 1.581 seconds, Fetched: 1 row(s)
hive> SELECT GUID, hex(GUID) FROM TBL_V;
OK
a0750c1f13f0k5��F8j���b�Ro'4da96
6130373530633166313366306B35EFBFBDEFBFBD46386AEFBFBDEFBFBDEFBFBD62EFBFBD526F273464613936
{code}
Is it expected?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
