[
https://issues.apache.org/jira/browse/HIVE-15388?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15860480#comment-15860480
]
Ashutosh Chauhan commented on HIVE-15388:
-----------------------------------------
I agree current behavior is indeed ambiguous. Requiring interval for all
expressions is less ambiguous than what we currently have. In that case select
(1) second will throw exception. However, since we do allow interval being
optional for constants, I am not sure if this can be handled at parser level.
In semantic analysis phase, we can add a check to throw in these cases.
For vectorization change, not between is now generating tokens as not between
instead of between not. As a result, vectorization which has a logic to handle
between not which used to generate a single udf is not getting triggered, but
instead doing two evaluations as between followed by not.
> HiveParser spends lots of time in parsing queries with lots of "("
> ------------------------------------------------------------------
>
> Key: HIVE-15388
> URL: https://issues.apache.org/jira/browse/HIVE-15388
> Project: Hive
> Issue Type: Improvement
> Affects Versions: 2.2.0
> Reporter: Rajesh Balamohan
> Assignee: Pengcheng Xiong
> Attachments: HIVE-15388.01.patch, HIVE-15388.02.patch,
> HIVE-15388.03.patch, HIVE-15388.04.patch, HIVE-15388.05.patch,
> HIVE-15388.06.patch, HIVE-15388.07.patch, hive-15388.stacktrace.txt
>
>
> Branch: apache-master (applicable with previous releases as well)
> Queries generated via tools can have lots of "(" for "AND/OR" conditions.
> This causes huge delays in parsing phase when the number of expressions are
> high.
> e.g
> {noformat}
> SELECT `iata`,
> `airport`,
> `city`,
> `state`,
> `country`,
> `lat`,
> `lon`
> FROM airports
> WHERE
> ((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((`airports`.`airport`
> = "Thigpen"
>
> OR `airports`.`airport` = "Astoria Regional")
>
> OR `airports`.`airport` = "Warsaw Municipal")
>
> OR `airports`.`airport` = "John F Kennedy Memorial")
>
> OR `airports`.`airport` = "Hall-Miller Municipal")
>
> OR `airports`.`airport` = "Atqasuk")
> OR
> `airports`.`airport` = "William B Hartsfield-Atlanta Intl")
> OR
> `airports`.`airport` = "Artesia Municipal")
> OR
> `airports`.`airport` = "Outagamie County Regional")
> OR
> `airports`.`airport` = "Watertown Municipal")
> OR
> `airports`.`airport` = "Augusta State")
> OR
> `airports`.`airport` = "Aurora Municipal")
> OR
> `airports`.`airport` = "Alakanuk")
> OR
> `airports`.`airport` = "Austin Municipal")
> OR
> `airports`.`airport` = "Auburn Municipal")
> OR
> `airports`.`airport` = "Auburn-Opelik")
> OR
> `airports`.`airport` = "Austin-Bergstrom International")
> OR
> `airports`.`airport` = "Wausau Municipal")
> OR
> `airports`.`airport` = "Mecklenburg-Brunswick Regional")
> OR
> `airports`.`airport` = "Alva Regional")
> OR
> `airports`.`airport` = "Asheville Regional")
> OR
> `airports`.`airport` = "Avon Park Municipal")
> OR
> `airports`.`airport` = "Wilkes-Barre/Scranton Intl")
> OR
> `airports`.`airport` = "Marana Northwest Regional")
> OR
> `airports`.`airport` = "Catalina")
> OR
> `airports`.`airport` = "Washington Municipal")
> OR
> `airports`.`airport` = "Wainwright")
> OR `airports`.`airport`
> = "West Memphis Municipal")
> OR `airports`.`airport`
> = "Arlington Municipal")
> OR `airports`.`airport` =
> "Algona Municipal")
> OR `airports`.`airport` =
> "Chandler")
> OR `airports`.`airport` =
> "Altus Municipal")
> OR `airports`.`airport` =
> "Neil Armstrong")
> OR `airports`.`airport` =
> "Angel Fire")
> OR `airports`.`airport` =
> "Waycross-Ware County")
> OR `airports`.`airport` =
> "Colorado City Municipal")
> OR `airports`.`airport` =
> "Hazelhurst")
> OR `airports`.`airport` =
> "Kalamazoo County")
> OR `airports`.`airport` =
> "Granville")
> OR `airports`.`airport` = "Silver
> Springs")
> OR `airports`.`airport` = "Whitford")
> OR `airports`.`airport` = "Biddeford
> Municipal")
> OR `airports`.`airport` = "Sugarloaf
> Regional")
> OR `airports`.`airport` = "Barnes
> Municipal")
> OR `airports`.`airport` = "Columbus
> Municipal")
> OR `airports`.`airport` = "Battle
> Mountain")
> OR `airports`.`airport` = "Huron County
> Memorial")
> OR `airports`.`airport` = "New Braunfels
> Municipal")
> OR `airports`.`airport` = "Benson Municipal")
> OR `airports`.`airport` = "Curtis")
> OR `airports`.`airport` = "Marlboro County")
> OR `airports`.`airport` = "Broken Bow
> Municipal")
> OR `airports`.`airport` = "Virginia Tech")
> OR `airports`.`airport` = "Bryce Canyon")
> OR `airports`.`airport` = "Black River Falls Area")
> OR `airports`.`airport` = "Boca Raton")
> OR `airports`.`airport` = "Birchwood")
> OR `airports`.`airport` = "Baudette International")
> OR `airports`.`airport` = "Blanding Muni")
> OR `airports`.`airport` = "Bradley International")
> OR `airports`.`airport` = "Morrilton Municipal")
> OR `airports`.`airport` = "Igor I Sikorsky Memorial")
> OR `airports`.`airport` = "Broadus")
> OR `airports`.`airport` = "Laurence G Hanscom")
> OR `airports`.`airport` = "Southwest Michigan Regional")
> OR `airports`.`airport` = "Bethel")
> OR `airports`.`airport` = "Bradford Reg")
> OR `airports`.`airport` = "Scotts Bluff County")
> OR `airports`.`airport` = "Boeing Field/King County Intl")
> OR `airports`.`airport` = "Buffalo Municipal")
> OR `airports`.`airport` = "Meadows")
> OR `airports`.`airport` = "Mobile Downtown")
> OR `airports`.`airport` = "Virgil I Grissom Municipal")
> OR `airports`.`airport` = "Silver Bay Municipal");
> {noformat}
> This query would take more than minutes to parse and compile.
> Temp Workaround: If all "("s in "where" condition are removed manually, this
> query would complete in 2-3 seconds.
> Problem exists in older versions as well (but in older versions this can make
> HS2 unresponsive for large amount of time, as Driver ends up taking global
> compiler lock in this phase).
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
