[
https://issues.apache.org/jira/browse/HIVE-626?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zheng Shao updated HIVE-626:
----------------------------
Attachment: HIVE-626.1.showinfo.patch
I added some instrumentation to the code (see HIVE-626.1.showinfo.patch) The
result of "explain extended" (below) shows that the order of the output column
of the JoinOperator does not match that of the FileSinkOperator:
{code}
hive> explain extended
> select zshao_foo.foo_name, zshao_bar.bar_name, n from zshao_foo join
zshao_bar on zshao_foo.foo_id =
> zshao_bar.foo_id join zshao_count on zshao_count.bar_id =
zshao_bar.bar_id;
OK
ABSTRACT SYNTAX TREE:
(TOK_QUERY (TOK_FROM (TOK_JOIN (TOK_JOIN (TOK_TABREF zshao_foo) (TOK_TABREF
zshao_bar) (= (. (TOK_TABLE_OR_COL zshao_foo) foo_id) (. (TOK_TABLE_OR_COL
zshao_bar) foo_id))) (TOK_TABREF zshao_count) (= (. (TOK_TABLE_OR_COL
zshao_count) bar_id) (. (TOK_TABLE_OR_COL zshao_bar) bar_id)))) (TOK_INSERT
(TOK_DESTINATION (TOK_DIR TOK_TMP_FILE)) (TOK_SELECT (TOK_SELEXPR (.
(TOK_TABLE_OR_COL zshao_foo) foo_name)) (TOK_SELEXPR (. (TOK_TABLE_OR_COL
zshao_bar) bar_name)) (TOK_SELEXPR (TOK_TABLE_OR_COL n)))))
STAGE DEPENDENCIES:
Stage-1 is a root stage
Stage-2 depends on stages: Stage-1
Stage-0 is a root stage
STAGE PLANS:
Stage: Stage-1
Map Reduce
Alias -> Map Operator Tree:
...
Reduce Operator Tree:
Join Operator
condition map:
Inner Join 0 to 1
condition expressions:
0 {VALUE._col1}
1 {VALUE._col0} {VALUE._col4}
output names: _col1, _col6, _col10
File Output Operator
compressed: true
GlobalTableId: 0
directory: hdfs://xxx:9000/tmp/hive-zshao/1413634235/10002
table:
input format: org.apache.hadoop.mapred.SequenceFileInputFormat
output format:
org.apache.hadoop.hive.ql.io.HiveSequenceFileOutputFormat
properties:
name binary_table
serialization.ddl struct binary_table { string _col1, string
_col10, i32 _col6}
serialization.format
com.facebook.thrift.protocol.TBinaryProtocol
name: binary_table
Stage: Stage-2
Map Reduce
Alias -> Map Operator Tree:
$INTNAME
...
{code}
The output of the join has the order: output names: _col1, _col6, _col10
The FileSinkOperator expects: struct binary_table { string _col1, string
_col10, i32 _col6}
> Typecast bug in Join operator
> -----------------------------
>
> Key: HIVE-626
> URL: https://issues.apache.org/jira/browse/HIVE-626
> Project: Hadoop Hive
> Issue Type: Bug
> Reporter: Zheng Shao
> Attachments: HIVE-626.1.showinfo.patch
>
>
> There is a type cast error in Join operator. Produced by the following steps:
> {code}
> create table zshao_foo (foo_id int, foo_name string, foo_a string, foo_b
> string,
> foo_c string, foo_d string) row format delimited fields terminated by ','
> stored as textfile;
> create table zshao_bar (bar_id int, bar_0 int, foo_id int, bar_1 int, bar_name
> string, bar_a string, bar_b string, bar_c string, bar_d string) row format
> delimited fields terminated by ',' stored as textfile;
> create table zshao_count (bar_id int, n int) row format delimited fields
> terminated by ',' stored as textfile;
> Each table has a single row as follows:
> zshao_foo:
> 1,foo1,a,b,c,d
> zshao_bar:
> 10,0,1,1,bar10,a,b,c,d
> zshao_count:
> 10,2
> load data local inpath 'zshao_foo' overwrite into table zshao_foo;
> load data local inpath 'zshao_bar' overwrite into table zshao_bar;
> load data local inpath 'zshao_count' overwrite into table zshao_count;
> explain extended
> select zshao_foo.foo_name, zshao_bar.bar_name, n from zshao_foo join
> zshao_bar on zshao_foo.foo_id =
> zshao_bar.foo_id join zshao_count on zshao_count.bar_id = zshao_bar.bar_id;
> {code}
> The case is from David Lerman.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.