It seems some issue with my mailbox, could not able to send mail, so I'm posting here :-
As tested against latest Hive build and Hadoop 20.1.
Kindly provide me the query or the list of queries for which this defect was tested against, as I see no test case in it and when I'm running queries assuming which were being run by He Yongqiang while testing the scenarios, I'm not able to re-produce the scenario.
I'm ran queries in below sequence :-
1 . set hive.exec.drop.ignorenonexistent=false;
2 . create table page_test(view INT, userid INT, page_url STRING) PARTITIONED BY(dt STRING, country STRING) STORED AS A TEXTFILE;
3 . LOAD DATA LOCAL INPATH '/home/test.txt' OVERWRITE INTO TABLE page_test PARTITION(dt='10-10-2010',country='US');
4 . LOAD DATA LOCAL INPATH '/home/test.txt' OVERWRITE INTO TABLE page_test PARTITION(dt='10-12-2010',country='IN');
// This query fails during semantic analysis, since it is invalid partition and don't belong to partition spec of page_test.
5 . ALTER TABLE page_test DROP PARTITION(invalid='23-02-2010',notexist='UK');
and when I turn ON the property hive.exec.drop.ignorenonexistent, hive is silent and OK.
So far good, and no issues.
//However there is another case, in which partition is a valid spec but doesn't exists, as I have only [dt='10-10-2010',country='US'] and [dt='10-12-2010',country='IN']
6 . ALTER TABLE page_test DROP PARTITION (dt='23-02-2010',country='UK');
but it doesn't results in any exception, I'm getting OK response on console.
As per this defect fix, it should actually fail with console error message
Partition : [
] does not exist.
I may be wrong, or not aware of enough use cases covering the scope of this defect, kindly help me providing some insights.