Ensure that bucketing and sort column names correspond to table column names

[findinpath]

Description

In the metastore, the bucketing and sorting column names can differ in case from its corresponding table column names. This change makes certain that, even though a table can be delivered by the metastore with such inconsistencies, Trino will make use of exactly the same bucketing and sort column names as their corresponding data column names.

Additional context and related issues

Reproduction scenario

testing/bin/ptl env up --environment 'singlenode-spark-hive'  --config 'config-default'

Connect to jdbc:hive2://localhost:10213/default
Spark

create table bucket_table(
                                    `row_id` int,
                                    `SEGMENT_ID` int
)
    partitioned by (`part` string)
    clustered by (`SEGMENT_ID`) into 10 buckets;

➜  ~ docker container exec -it ptl-hadoop-master /bin/bash
(reverse-i-search)`': ^C
[root@hadoop-master /]# mysql -D metaastore -u root -proot
MariaDB [metastore]> select * from COLUMNS_V2 where cd_id = 56;
+-------+---------+-------------+-----------+-------------+
| CD_ID | COMMENT | COLUMN_NAME | TYPE_NAME | INTEGER_IDX |
+-------+---------+-------------+-----------+-------------+
|    56 | NULL    | row_id      | int       |           0 |
|    56 | NULL    | segment_id  | int       |           1 |
+-------+---------+-------------+-----------+-------------+
2 rows in set (0.00 sec)


MariaDB [metastore]> select * from BUCKETING_COLS;
+-------+-----------------+-------------+
| SD_ID | BUCKET_COL_NAME | INTEGER_IDX |
+-------+-----------------+-------------+
|    56 | SEGMENT_ID      |           0 |
+-------+-----------------+-------------+

The inconsistency in case for the data column segment_id and the bucketing column SEGMENT_ID was causing in Trino the issue:

java.lang.IllegalArgumentException: Cannot find column 'SEGMENT_ID' in bucket_table at io.trino.plugin.hive.util.HiveBucketing.lambda$isSupportedBucketing$4(HiveBucketing.java:324)

Release notes

( ) This is not user-visible or docs only and no release notes are required.
( ) Release notes are required, please propose a release note for me.
(x) Release notes are required, with the following suggested text:

# Hive
* Ensure that bucketing and sort column names correspond to table column names. ({issue}`issuenumber`)

[findepi]

there should be exactly one such

also, linear search over column list is OK provided that there are only few bucketing columns, which is a reasonable assumption (otherwise we would build a set of data column names). add a code comment.

however, i wonder whether we have to do this validation here at all
i think it should be sufficient to lowercase the column names.

storageDescriptor.getBucketCols().stream()
 .map(name -> name.toLowerCase(ENGLISH))
 .collect(toImmList())

[findinpath]

I followed your suggestion and lowercased the bucketing and sorting columns to target them matching the data column names

[findepi]

is it legal to bucket on a partitioning column? (i know it doesn't make sense)

[findinpath]

CI hit #14637

[findinpath]

CI hit #12818