Carter Shanklin created HIVE-17023:
--------------------------------------
Summary: HPL/SQL: NPE creating table if datatype lacks a precision
and hplsql.conn.convert.hiveconn is false
Key: HIVE-17023
URL: https://issues.apache.org/jira/browse/HIVE-17023
Project: Hive
Issue Type: Bug
Components: hpl/sql
Reporter: Carter Shanklin
This bug is part of a series of issues and surprising behavior I encountered
writing a reporting script that would aggregate values and give rows different
classifications based on an the aggregate. Addressing some or all of these
issues would make HPL/SQL more accessible to newcomers.
Consider this script:
create table if not exists test1(col1 integer);
create table if not exists test2(col1 double);
create table if not exists test3(col1 decimal(10, 4));
create table if not exists test4(col1 string);
create table if not exists test5(col1 varchar(20));
If you run with this config:
<configuration>
<property>
<name>hplsql.conn.default</name>
<value>hiveconn</value>
</property>
<property>
<name>hplsql.conn.hiveconn</name>
<value>org.apache.hive.jdbc.HiveDriver;jdbc:hive2://</value>
</property>
</configuration>
You will get exceptions like this for some of the types:
Exception in thread "main" java.lang.NullPointerException
at org.apache.hive.hplsql.Exec.evalPop(Exec.java:2398)
at org.apache.hive.hplsql.Stmt.createTableDefinition(Stmt.java:169)
at org.apache.hive.hplsql.Stmt.createTable(Stmt.java:142)
at org.apache.hive.hplsql.Exec.visitCreate_table_stmt(Exec.java:1366)
at org.apache.hive.hplsql.Exec.visitCreate_table_stmt(Exec.java:52)
at
org.apache.hive.hplsql.HplsqlParser$Create_table_stmtContext.accept(HplsqlParser.java:4198)
at
org.antlr.v4.runtime.tree.AbstractParseTreeVisitor.visitChildren(AbstractParseTreeVisitor.java:70)
at org.apache.hive.hplsql.Exec.visitStmt(Exec.java:1013)
at org.apache.hive.hplsql.Exec.visitStmt(Exec.java:52)
at
org.apache.hive.hplsql.HplsqlParser$StmtContext.accept(HplsqlParser.java:1018)
...
FAIL: create table test1 (col1 integer);
FAIL: create table test2 (col1 double);
SUCCESS: create table test3 (col1 decimal(10, 4));
FAIL: create table test4 (col1 string);
SUCCESS: create table test5(col1 varchar(20));
Problem boils down to this line in Stmt.java:
exec.append(sql, exec.evalPop(col.dtype(), col.dtype_len()),
col.column_name().getStop(), col.dtype().getStart());
Because col.dtype_len() returns null
I have to wonder if not converting is ever a good idea? If it's always a bad
idea(tm) why even have it?
Version = 3.0.0-SNAPSHOT r71f52d8ad512904b3f2c4f04fe39a33f2834f1f2
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
