mirror of
https://github.com/apache/impala.git
synced 2026-01-07 09:02:19 -05:00
9b507b6ed6
Creating Kudu tables shouldn't allow types not supported by Kudu (e.g. VARCHAR/CHAR, DECIMAL, TIMESTAMP, collection types). The behavior is inconsistent: for some types it throws in the catalog, for VARCHAR/CHAR these become strings. This changes behavior so that all fail during analysis. Analysis tests were added. Similarly, external tables cannot contain Kudu types that Impala doesn't support (e.g. UNIXTIME_MICROS, BINARY). Tests were added to validate this behavior. Note that this required upgrading the python Kudu client. This also fixes a small corner case with ALTER TABLE: ALTER TABLE shouldn't allow Kudu tables to change the storage descriptor tblproperty, otherwise the table metadata gets in an inconsistent state. Tests were added for all of the above. Change-Id: I475273cbbf4110db8d0f78ddf9a56abfc6221e3e Reviewed-on: http://gerrit.cloudera.org:8080/4857 Reviewed-by: Dimitris Tsirogiannis <dtsirogiannis@cloudera.com> Tested-by: Tim Armstrong <tarmstrong@cloudera.com>
81 lines
2.3 KiB
Plaintext
81 lines
2.3 KiB
Plaintext
====
|
|
---- QUERY
|
|
create table t primary key (id) distribute by hash (id) into 3 buckets
|
|
stored as kudu
|
|
as select id, int_col from functional.alltypestiny;
|
|
select * from t;
|
|
---- RESULTS
|
|
0,0
|
|
1,1
|
|
2,0
|
|
3,1
|
|
4,0
|
|
5,1
|
|
6,0
|
|
7,1
|
|
---- TYPES
|
|
INT,INT
|
|
====
|
|
---- QUERY
|
|
# Boolean primary key column
|
|
create table tab (x int, y boolean, primary key(x, y))
|
|
distribute by hash (x) into 3 buckets stored as kudu
|
|
---- CATCH
|
|
NonRecoverableException: Key column may not have type of BOOL, FLOAT, or DOUBLE
|
|
====
|
|
---- QUERY
|
|
# Float primary key column
|
|
create table tab (x int, y float, primary key(x, y))
|
|
distribute by hash (x) into 3 buckets stored as kudu
|
|
---- CATCH
|
|
NonRecoverableException: Key column may not have type of BOOL, FLOAT, or DOUBLE
|
|
====
|
|
---- QUERY
|
|
# Primary keys should be declared first
|
|
# TODO: See KUDU-1709 for improving Kudu error messages.
|
|
create table tab (x int, y int, primary key(y))
|
|
distribute by hash (y) into 3 buckets stored as kudu
|
|
---- CATCH
|
|
NonRecoverableException: Got out-of-order key column: name: "y" type: INT32 is_key: true is_nullable: false cfile_block_size: 0
|
|
====
|
|
---- QUERY
|
|
# Small number of hash buckets
|
|
create table tab (a int, b int, c int, d int, primary key(a, b, c))
|
|
distribute by hash(a,b) into 8 buckets, hash(c) into 1 buckets stored as kudu
|
|
---- CATCH
|
|
NonRecoverableException: must have at least two hash buckets
|
|
====
|
|
---- QUERY
|
|
# Same column in multiple hash based distributions
|
|
create table tab (a int, b int, primary key (a))
|
|
distribute by hash (a) into 3 buckets, hash (a) into 2 buckets stored as kudu
|
|
---- CATCH
|
|
NonRecoverableException: hash bucket schema components must not contain columns in common
|
|
====
|
|
---- QUERY
|
|
# Same column referenced multiple times in the same hash-based distribution
|
|
create table tab (a int primary key) distribute by hash (a, a, a) into 3 buckets
|
|
stored as kudu
|
|
---- CATCH
|
|
NonRecoverableException: hash bucket schema components must not contain columns in common
|
|
====
|
|
---- QUERY
|
|
# Kudu table that uses Impala keywords as table name and column names
|
|
create table `add`(`analytic` int, `function` int, primary key(`analytic`, `function`))
|
|
distribute by hash (`analytic`) into 4 buckets, range (`function`) split rows ((1), (10))
|
|
stored as kudu;
|
|
insert into `add` select id, int_col from functional.alltypestiny;
|
|
select * from `add`
|
|
---- RESULTS
|
|
0,0
|
|
1,1
|
|
2,0
|
|
3,1
|
|
4,0
|
|
5,1
|
|
6,0
|
|
7,1
|
|
---- TYPES
|
|
INT,INT
|
|
====
|