Add hive configs for supported read and write formats

[pramodsatya]

Description

Adds hive configs hive.read-formats and hive.write-formats to configure the file formats supported by hive connector for read and write operations respectively.

Motivation and Context

Presto C++ only supports reading of tables with DWRF, ORC and PARQUET formats, and writing to tables with DWRF and PARQUET formats, with the hive connector. Using these hive configs will allow to fail-fast at coordinator when attempting to read from and write to tables with unsupported file formats in Presto C++.
Currently attempting to read from tables with unsupported file formats in Presto C++ fails at the worker:

it != readerFactories().end() ReaderFactory is not registered for format text

Release Notes

== RELEASE NOTES ==
Hive Connector Changes
* Adds :ref:`connector/hive:Hive Configuration Properties` `hive.read-formats` and `hive.write-formats` to allow users to set file formats supported for read and write operations by hive connector.

[aditi-pandit]

@pramodsatya : Thanks for this code. Should we add a check for the file formats applicable at the Writer side as well ? Native execution only supports DWRF and Parquet writers.

[pramodsatya]

Thanks for the feedback @tdcmeehan, @aditi-pandit . Added hive configs for supported read and write formats, and validated read/write operations fail for unsupported formats when these configs are set. Could you please take another look?

[aditi-pandit]

Thanks @pramodsatya

[steveburnett]

Should we have documentation for these new properties? https://github.com/prestodb/presto/blob/master/presto-docs/src/main/sphinx/presto_cpp/properties.rst

[steveburnett]

LGTM! (docs)

Pull branch, local doc build, looks good. Thanks!

[tdcmeehan]

Can you mention that the default behavior is to allow all built-in support of read and write formats?

[pramodsatya]

Added this to the doc, could you please take another look?

[aditi-pandit]

@pramodsatya : This code looks good. Though can you write e2e tests with queries ... maybe set read/write formats to only parquet in the queryRunner and show that reading.writing any other format fails.

[pramodsatya]

@pramodsatya : This code lookso good. Though can you write e2e tests with queries ... maybe set read/write formats to only parquet in the queryRunner and show that reading.writing any other format fails.

Thanks @aditi-pandit, added tests for these configs with different file formats. Could you please take another look?

[steveburnett]

Suggest adding a link to the doc in the release note entry, like so:

== RELEASE NOTES ==

Hive Connector Changes
* Adds :ref:`connector/hive:Hive Configuration Properties` `hive.read-formats` and `hive.write-formats` to allow users to set file formats supported for read and write operations by hive connector.

[aditi-pandit]

Don't think we should delay checking the tablelayout for storageFormat until split scheduling. We want to avoid scheduling the fragment at all. Is it possible to do this earlier in BasePlanFragmenter
https://github.com/prestodb/presto/blob/master/presto-main-base/src/main/java/com/facebook/presto/sql/planner/BasePlanFragmenter.java#L249 ? The TableLayout is available at this point.

@tdcmeehan : wdyt ?

[aditi-pandit]

Even this code is called during scheduling stage executions... Its worthwhile only if the check is done sooner in logical planning or BasePlanFragmenter.

I think extractHiveStorageFormat from a table can be called much sooner. Can you investigate further ?