[
https://issues.apache.org/jira/browse/IMPALA-6503?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16722479#comment-16722479
]
Quanlong Huang commented on IMPALA-6503:
----------------------------------------
Actually, the two assumptions are wrong... While testing I found these:
{code}
Tuple(id=1 size=9 slots=[Slot(id=2 type=DOUBLE col_path=[2,0,2,0] offset=0
null=(offset=8 mask=1) slot_idx=0 field_idx=-1)] tuple_path=[2, 0, 2])
Tuple(id=0 size=25 slots=[Slot(id=0 type=ARRAY col_path=[2,0,2]
collection_item_tuple_id=1 offset=0 null=(offset=24 mask=1) slot_idx=0
field_idx=-1), Slot(id=1 type=BIGINT col_path=[2,1] offset=16 null=(offset=24
mask=2) slot_idx=1 field_idx=-1)] tuple_path=[2])
{code}
a. The slots in the top level tuple (id=0) have different length of col_path.
b. The top level tuple path is [2] while the col_path of its first slot is
[2,0,2].
The table is created by Hive:
{code:sql}
hive> create table struct_array (
id int, name string,
structs array<struct<sid:int, sname:string, scores:array<double>>>
) stored as orc;
hive> insert into struct_array select 1, "abcdefg", array(
named_struct("sid", 0, "sname", "inner_name1", "scores", array(0.1, 0.2, 0.3,
0.4)),
named_struct("sid", 1, "sname", "inner_name2", "scores", array(100, 99.9,
98.9)),
named_struct("sid", 2, "sname", "inner_name3", "scores", array(3.14159,
2.71828, 0.00000001)));
{code}
And the Impala query is:
{code:sql}
[localhost:21000] default> select outer_array.pos, s.item from
struct_array.structs outer_array, outer_array.item.scores s;
{code}
> Support reading complex types from ORC format files
> ---------------------------------------------------
>
> Key: IMPALA-6503
> URL: https://issues.apache.org/jira/browse/IMPALA-6503
> Project: IMPALA
> Issue Type: New Feature
> Components: Backend, Frontend
> Reporter: Quanlong Huang
> Assignee: Quanlong Huang
> Priority: Major
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
