Jesse Santangelo created HIVE-17264:
---------------------------------------
Summary: BaseSemanticAnalyzer.escapeSQLString escapes \0 unsafely
Key: HIVE-17264
URL: https://issues.apache.org/jira/browse/HIVE-17264
Project: Hive
Issue Type: Bug
Components: Query Planning
Reporter: Jesse Santangelo
Priority: Minor
Currently escapeSQLString() converts the null character to \0. If there are
numerical values after the null character, calling unescapeSQLString() can
treat the resulting value as an octal literal.
Example:
"FOOBAR_EVENT\u00002017"
unescapeSQLString converts \u0000 to the null character. Calling
escapeSQLString() on that resulting string currently gives
"FOOBAR_EVENT\02017"
A third call of unescapeSQLString() will then see \020 and convert it to 0x10.
This is an issue for the Hbase connector where key values are processed twice
(I'm unfamiliar with the hive codebase so I don't know why) and so the
connector gets the wrong key value if it's formatted like the example.
In my version I've changed escapeSQLString to replace the null character with
'\u0000' instead of '\0' and that seems to have fixed the issue, but I don't
know if that might break anything else.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
