Rajeshbabu Chintaguntla created PHOENIX-3374:
------------------------------------------------
Summary: Wrong data row key is getting generated for local indexes
for functions with fixed non null columns
Key: PHOENIX-3374
URL: https://issues.apache.org/jira/browse/PHOENIX-3374
Project: Phoenix
Issue Type: Bug
Reporter: Rajeshbabu Chintaguntla
Assignee: Rajeshbabu Chintaguntla
Fix For: 4.9.0, 4.8.2
When we create local indexes on functions then while joining back missing
columns from data table is generating wrong data row key.
When we generated index row key schema we are initializing isNullableToBe to
true always even if the expression might be not null. Because of this we might
change the column type to other castable one and may consider actual data as
separator bytes.
{noformat}
...
if (indexField == null) {
Expression e = expressionItr.next();
isNullableToBe = true;
dataTypeToBe = IndexUtil.getIndexColumnDataType(isNullableToBe,
e.getDataType());
sortOrderToBe = descIndexColumnBitSet.get(i) ? SortOrder.DESC :
SortOrder.ASC;
maxLengthToBe = e.getMaxLength();
scaleToBe = e.getScale();
...
{noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
