Display the output column name in InvalidNullByteException

[LakshSingla]

Description

InvalidNullByteFault for MSQ displayed the query column name (the one that is used internally while planning and executing the query) instead of the output column name (the one that is given by the user). Sometimes the two are similar, however, for columns involving expressions, it is not the case.
It can be misleading for someone who is not familiar with this quirk, and even then it would require reading the native plan to figure out the culprit column.

This PR maps the query column to the output column name while surfacing the fault since that is readily visible to the user while executing the query.

---

For a query like:

WITH "ext" AS (SELECT *
FROM TABLE(
  EXTERN(
    '{"type":"inline","data":"{\"desc\":\"Normal row\",\"text\":\"Hi I am a normal row\"}\n{\"desc\":\"Row with NULL\",\"text\":\"There is a null in\\u0000 here somewhere\"}\n{\"desc\":\"Anther normal row\",\"text\":\"Nice\"}\n"}',
    '{"type":"json"}'
  )
) EXTEND ("desc" VARCHAR, "text" VARCHAR))
SELECT
  "desc",
  REPLACE("text", 'a', 'A') AS "text"
FROM "ext"

Original Behaviour:

InvalidNullByte: Invalid null byte at source [external input source: InlineInputSource{data='{"desc":"Normal row","text":"Hi I am a normal row"} {"desc":"Row with NULL","text":"There is a null in\u0000 here somewhere"} {"desc":"Anther normal row","text":"Nice"} '}], rowNumber [2], column[v0], value[There is A null in here somewhere], position[18]. Consider sanitizing the string using REPLACE("v0", U&'\0000', '') AS v0.

Notice it refers to the column as 'v0'

Modified Behaviour
InvalidNullByte: Invalid null byte at source [external input source: InlineInputSource{data='{"desc":"Normal row","text":"Hi I am a normal row"} {"desc":"Row with NULL","text":"There is a null in\u0000 here somewhere"} {"desc":"Anther normal row","text":"Nice"} '}], rowNumber [2], column[text], value[There is A null in here somewhere], position[18]. Consider sanitizing the string using REPLACE("text", U&'\0000', '') AS text.

---

Note, for columns without alias, it will display it as EXPR$<number> which is still more informative than something like v0.

WITH "ext" AS (SELECT *
FROM TABLE(
  EXTERN(
    '{"type":"inline","data":"{\"desc\":\"Normal row\",\"text\":\"Hi I am a normal row\"}\n{\"desc\":\"Row with NULL\",\"text\":\"There is a null in\\u0000 here somewhere\"}\n{\"desc\":\"Anther normal row\",\"text\":\"Nice\"}\n"}',
    '{"type":"json"}'
  )
) EXTEND ("desc" VARCHAR, "text" VARCHAR))
SELECT
  "desc",
  REPLACE("text", 'a', 'A')
FROM "ext"

Modified Behaviour
InvalidNullByte: Invalid null byte at source [external input source: InlineInputSource{data='{"desc":"Normal row","text":"Hi I am a normal row"} {"desc":"Row with NULL","text":"There is a null in\u0000 here somewhere"} {"desc":"Anther normal row","text":"Nice"} '}], rowNumber [2], column[EXPR$1], value[There is A null in here somewhere], position[18]. Consider sanitizing the string using REPLACE("EXPR$1", U&'\0000', '') AS EXPR$1.

---

This PR has:

• been self-reviewed.
  • using the concurrency checklist (Remove this item if the PR doesn't have any relation to concurrency.)
• added documentation for new or modified features or behaviors.
• a release note entry in the PR description.
• added Javadocs for most classes and all non-trivial methods. Linked related entities via Javadoc links.
• added or updated version, license, or notice information in licenses.yaml
• added comments explaining the "why" and the intent of the code wherever would not be obvious for an unfamiliar reader.
• added unit tests or modified existing tests to cover new code paths, ensuring the threshold for code coverage is met.
• added integration tests.
• been tested in a test Druid cluster.

[abhishekagarwal87]

can "Consider sanitizing the string using" be changed to "Consider sanitizing the input string using"

[adarshsanjeev]

nit: Should we handle the case where it is more than 1 separately for neatness? That said, since this is code to handle a different type of exception, I am okay with the current code.

[LakshSingla]

I think it should be fine since even if there are multiple columns for a single query column, to surface the exception, we can only display one name, and displaying the first one seems okay.

If the code which throws InvalidNullByteException instead showed all the columns with values \0 in them (which I think won't be possible), we can modify it, however as the code currently is, we throw on encountering the first value with \0.

[cryptoe]

Left one comment.
Rest all LGTM.

[cryptoe]

IMHO: this change should go into :

druid/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerImpl.java

Lines 739 to 747 in 3ddcaa0

	public void workerError(MSQErrorReport errorReport)
	{
	if (workerTaskLauncher.isTaskCanceledByController(errorReport.getTaskId()) ||
	!workerTaskLauncher.isTaskLatest(errorReport.getTaskId())) {
	log.info("Ignoring task %s", errorReport.getTaskId());
	} else {
	workerErrorRef.compareAndSet(null, errorReport);
	}
	}

[LakshSingla]

Makes much more sense, incorporated the feedback in my latest commit.

[LakshSingla]

Missed out on the latest comments. Thanks for the reviews @abhishekagarwal87 @adarshsanjeev @cryptoe!!