Table has more than one bucket keys, but "show create table xxx" only displays one

[madeirak]

Apache Iceberg version

1.4.3

Query engine

Spark

Please describe the bug 🐞

Through "select * from xx.xx.partitions" above, it can be seen that this table has two bucket keys. But "show create table xx.xx"as below,only display one bucket key

[manuzhang]

The table has two partition keys from two partition transforms, one of which is bucket.

[madeirak]

The table has two partition keys from two partition transforms, one of which is bucket.

Are these two partition transforms equivalent? name_bucket_10 and id_bucket_10

Are the principle both hash?

[manuzhang]

Sorry, I missed name_bucket_10 part. How did you create your table? With which catalog?

[madeirak]

Sorry, I missed name_bucket_10 part. How did you create your table? With which catalog?

Similar to the following process:

create table   dbxx.tbxx (id INT COMMENT '11', name STRING COMMENT '') USING iceberg PARTITIONED BY (name, bucket(10, name), bucket(10, id ));
insert into tbxx values (1, '1');
show create table dbxx.tbxx ;
select * from dbxx.tbxx.partitions;

[madeirak]

Sorry, I missed name_bucket_10 part. How did you create your table? With which catalog?

With HiveCatalog

[lurnagao-dahua]

create table dbxx.tbxx (id INT COMMENT '11', name STRING COMMENT '') USING iceberg PARTITIONED BY (name, bucket(10, name), bucket(10, id )); insert into tbxx values (1, '1'); show create table dbxx.tbxx ; select * from dbxx.tbxx.partitions;

I am quite puzzled why name is used as both partition and bucket. In this case, all the data under the name partition is in the same bucket, and the bucketing effect is meaningless.

[madeirak]

create table dbxx.tbxx (id INT COMMENT '11', name STRING COMMENT '') USING iceberg PARTITIONED BY (name, bucket(10, name), bucket(10, id )); insert into tbxx values (1, '1'); show create table dbxx.tbxx ; select * from dbxx.tbxx.partitions;

I am quite puzzled why name is used as both partition and bucket. In this case, all the data under the name partition is in the same bucket, and the bucketing effect is meaningless.

This is just an example, not a real table. The main issue is that multiple bucket fields only display one in "show create table xxx"

[manuzhang]

The show create table result is following Spark SQL syntax, which only supports one bucket field.

[madeirak]

The show create table result is following Spark SQL syntax, which only supports one bucket field.

ok, fine. It would be better if it could be as shown in the Iceberg document: ref: https://iceberg.apache.org/docs/latest/spark-ddl/#partitioned-by