[IMPALA-252] Empty string partition value causes metastore update failure

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: Impala 0.6
• Fix Version/s: Impala 1.0
• Assignee: Alexander Behm
• Reporter: Henry Robinson
• Created: 15/Apr/13 02:20
• Updated: 29/Apr/13 04:04
• Resolved: 26/Apr/13 00:08

Description

[localhost:21000] > create table s_part(col int) partitioned by (s string);
Query: create table s_part(col int) partitioned by (s string)
[localhost:21000] > insert into s_part partition(s)  select 1, "" from functional.alltypes limit 1;
Query: insert into s_part partition(s)  select 1, "" from functional.alltypes limit 1
Remote error

[localhost:21000] > insert into s_part partition(s="")  select 1 from functional.alltypes limit 1;
Query: insert into s_part partition(s="")  select 1 from functional.alltypes limit 1
Remote error

Hive's behaviour is arguably worse; empty string in partition columns get silently mapped to _HIVE_DEFAULT_PARTITION_ - and then that stays the value when that partition is selected!

We should map empty strings to our own _IMPALA_EMPTY_STRING_ constant, and then do the right thing when selecting them back. Same as for NULL values.