[HIVE-16935] Hive should strip comments from input before choosing which CommandProcessor to run.

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.0.0
• Assignee: Andrew Sherman
• Reporter: Andrew Sherman
• Created: 21/Jun/17 20:10
• Updated: 15/Jun/18 23:41
• Resolved: 06/Jul/17 20:43

Description

While using Beeswax, Hue fails to execute statement with following error:

Error while compiling statement: FAILED: ParseException line 3:4 missing KW_ROLE at 'a' near 'a' line 3:5 missing EOF at '=' near 'a'

– comment
SET a=1;
SELECT 1;

The same code works in Beeline and in Impala.
The same code fails in CliDriver

Background

Hive deals with sql comments (“-- to end of line”) in different places.
Some clients attempt to strip comments. For example BeeLine was recently enhanced in https://issues.apache.org/jira/browse/HIVE-13864 to strip comments from multi-line commands before they are executed.
Other clients such as Hue or Jdbc do not strip comments before sending text.
Some tests such as TestCliDriver strip comments before running tests.
When Hive gets a command the CommandProcessorFactory looks at the text to determine which CommandProcessor should handle the command. In the bug case the correct CommandProcessor is SetProcessor, but the comments confuse the CommandProcessorFactory and so the command is treated as sql. Hive’s sql parser understands and ignores comments, but it does not understand the set commands usually handled by SetProcessor and so we get the ParseException shown above.

Issue Links

• breaks: HIVE-18127 Do not strip '--' comments from shell commands issued from CliDriver
• duplicates: HIVE-15822 beeline ignore all sql under comment after semicolon
• supercedes: HIVE-17050 Multiline queries that have comment in middle fail when executed via "beeline -e"
• links to: RB