Produce RunLengthEncodedBlock in VariableWidthBlockBuilder when all values are null

[lukasz-stec]

Description

Add support in VariableWidthBlockBuilder for producing RunLengthEncodedBlock if all positions are null.

Is this change a fix, improvement, new feature, refactoring, or other?

improvement

Is this a change to the core query engine, a connector, client library, or the SPI interfaces? (be specific)

only VariableWidthBlockBuilder class

How would you describe this change to a non-technical end user or system administrator?

Related issues, pull requests, and links

Documentation

( ) No documentation is needed.
( ) Sufficient documentation is included in this PR.
( ) Documentation PR is available with #prnumber.
( ) Documentation issue #issuenumber is filed, and can be handled later.

Release notes

( ) No release notes entries required.
( ) Release notes entries required with the following suggested text:

# Section
* Fix some things. ({issue}`issuenumber`)

[findepi]

Add support in VariableWidthBlockBuilder for producing RunLengthEncodedBlock if all positions are null.

Or should the caller be responsible to do this instead?
If not the caller, are you planning on adding similar functionality to all the block builders?

cc @dain

[lukasz-stec]

Add support in VariableWidthBlockBuilder for producing RunLengthEncodedBlock if all positions are null.

Or should the caller be responsible to do this instead? If not the caller, are you planning on adding similar functionality to all the block builders?

cc @dain

It's already there for most of BlockBuilkders (all simple types except this).
I'm adding RowBlockBuilder support in another PR.

[sopel39]

assertIsNullRle

[sopel39]

% comments

[sopel39]

testBuildProducesNullRle

[sopel39]

getRegion doesn't really copy anything.

Maybe just inline these two test methods and call it, testBuilderProducesNullRleForNullRows

[lukasz-stec]

ok, i merged the methods into testBuilderProducesNullRleForNullRows

[lukasz-stec]

io.trino.execution.buffer.TestPagesSerde#testVarcharSerializedSize was failing. I fixed the test but it's quite interesting that for blocks with just nulls, up until something like ~300 positions, a normal, flat block will serialize to a smaller number of bytes than RLE (because of RLE overhead + good compression of flat block).
This means that this change could theoretically increase network traffic for some special cases (i.e. small, null blocks).