Manish Gupta created CARBONDATA-1094:
----------------------------------------
Summary: Wrong results returned by the query in case inverted index is not created on a column
Key: CARBONDATA-1094
URL:
https://issues.apache.org/jira/browse/CARBONDATA-1094 Project: CarbonData
Issue Type: Bug
Reporter: Manish Gupta
Assignee: Manish Gupta
Fix For: 1.2.0
While creating a table if a column is not specified as sort column or included as no inverted index then the column will not be sorted during data load. Unsorted data will have incorrect min/max values and inverted index will not be created for that column.
During query, if filter exists for that column it gives incorrect results as binary search cannot be applied on the unsorted data.
Commands to reproduce
-----------------------------------------
CREATE TABLE IF NOT EXISTS index1 (id Int, name String, city String) STORED BY 'org.apache.carbondata.format' TBLPROPERTIES('NO_INVERTED_INDEX'='name,city', 'DICTIONARY_EXCLUDE'='city')
LOAD DATA LOCAL INPATH '<csv file path>' into table index1
SELECT * FROM index1 WHERE city >= 'Shanghai'
+---+------+----------+
| id| name| city|
+---+------+----------+
| 11| James|Washington|
| 5| John| Beijing|
| 20| Kevin| Singapore|
| 17| Lisa| Hangzhou|
| 12| Maria| Berlin|
| 2| Mark| Paris|
| 9| Mary| Tokyo|
| 6|Michel| Chicago|
| 16| Paul| Shanghai|
| 14| Peter| Boston|
| 7|Robert| Houston|
| 4| Sara| Tokyo|
| 8| Sunny| Boston|
+---+------+----------+
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)