Removed table name and id from RFileScanner.

[dlmarion]

Closes #5505

[kevinrr888]

Would still need the withTableId here for the calls env.getPluginEnv().getConfiguration(env.getTableId()) (from javadocs for IteratorEnvironment.getPluginEnv()) so env.getTableId() doesn't throw an exception

[dlmarion]

RFileScannerEnvironmentImpl overrides ClientServiceEnvironmentImpl.getConfiguration(TableId) and returns the tableConf regardless of the TableId.

[dlmarion]

Looking at the 2.1.3 version of RFileScanner, it had an internal class IterEnv which did not override IteratorEnvironment.getTableId, so if it was called it would throw an UnsupportedOperationException.

[kevinrr888]

Yes, that was the reason for #4810 and which was fixed with #5490

[kevinrr888]

So, previously the iter env for rfilescanner did not have getServiceEnv(), getPluginEnv(), or getTableId() implemented. We need all three of these methods to support the uses:

(quoting from my comment #4810 (comment))
The equivalent to
env.getServiceEnv() <-- deprecated in 2.x, but still need in 2.x
would now be
env.getPluginEnv()

and the equivalent to
env.getConfig() <-- deprecated in 2.x, but still need in 2.x
would now be
env.getPluginEnv().getConfiguration(env.getTableId())

[dlmarion]

I modified ClientIteratorEnvironment in f857cb0 to support setting tableId to null. This should avoid throwing an exception and allow RFileScannerEnvironmentImpl.getConfiguration(TableId) to return the table conf.

[kevinrr888]

could have TABLE_ID = null and check that argument for getConfiguration and getTableName is null/TABLE_ID to avoid accepting random values. Argument should be env.getTableId()

[dlmarion]

The RFileScanner is client utility where the user supplies the set of input files, configuration, columns they want to fetch, etc. I don't know that there is utility in validating the table id when this is being used outside of a table context.

[kevinrr888]

Yeah, this is a weird case where we need to pretend there is a table id when there isn't. I'm fine also not validating it, my idea was just that we should expect env.getTableId() as the way of getting the table id when it is needed instead of accepting any table id

[dlmarion]

Does validating the table id and throwing an exception change the behavior of the code between the 2.1.3 and 2.1.4 releases? Are we going to introduce an error into the clients use of our utility by upgrading?

[dlmarion]

Answering my own question - Looking at the 2.1.3 RFileScanner code, IterEnv doesn't override getServiceEnv, getPluginEnv or getTableId so any usage of those methods would throw an UnsupportedOperationException. We could add the validation, or not, it doesn't matter as the methods aren't called in RFileScanner.