[core] Improve the performance of show tables with hive metastore

[xleoken]

Purpose

When store the catalog in s3 object store, as the number of tables increases, the performance of the show tables deteriorates.

HiveCatalog#listTablesImpl will check each table in filesystem, will send many list request to the s3 server.

reason as #4592

Tests

List 102 tables

Before patch

abc098
abc099
Time taken: 25.643 seconds, Fetched 102 row(s)

After patch

abc098
abc099
Time taken: 1.711 seconds, Fetched 102 row(s)

API and Format

Documentation

[wwj6591812]

Hi, @xleoken thanks for your prepare this pr.
I have a question:
Why change AbstractCatalog#tableSchemaInFileSystem to AbstractCatalog#tableExistsInFileSystem can Improve the performance.? They both need call "listVersionedFiles(fileIO, schemaDirectory(), SCHEMA_PREFIX)" only once.

[xleoken]

hi @wwj6591812, AbstractCatalog#tableExistsInFileSystem will check schema-0 first, it read the s3 key directly.

[wwj6591812]

hi @wwj6591812, AbstractCatalog#tableExistsInFileSystem will check schema-0 first, it read the s3 key directly.

OK, After I git pull the new code, I found it .
I am agree with you.

[wwj6591812]

+1

[xleoken]

+1

thanks

[JingsongLi]

+1