fix(sql): unhandled ArrayIndexOutOfBoundsException in ASOF and LT JOIN on empty table

[puzpuzpuz]

Fixes #3301

[jerrinot]

If I understand the problem well, it's caused by the fact the record.getRowId() has undefined behaviour when the Record does not point to anything, because there was no cursor.hasNext() returning true called. Is that right?

Do you think it would make sense to define the getRow() contract that way? currently, it says this:

    /**
     * Gets the numeric ID of this row. This can be not real table row id
     *
     * @return numeric ID of the current row
     */

how about:

    /**
     * Gets the numeric ID of this row. This can be not real table row id. 
     *
     * @return numeric ID of the current row or Long.LONG_NaN when the record is not associated with a valid row. 
     */

Or do you think LONG_NaN could be used to represent an actual row?

[puzpuzpuz]

Do you think it would make sense to define the getRow() contract that way?

To me, getRowId() contact should be the same as for getStr(int) or any other column accessor method, i.e. it should be called only when the record is initialized with a previous hasNext() call that returned true. Introducing a special contract for getRowId() only would make the Record API less consistent.

@jerrinot WDYT?

[jerrinot]

@puzpuzpuz that's a good point.

I thought the getRowId() was more meta-method than e.g getStr() and thus it should always return a well-defined result. but I understand your view too.

[puzpuzpuz]

@jerrinot thanks for the review!

[ideoma]

[PR Coverage check]

😍 pass : 6 / 6 (100.00%)

file detail

	path	covered line	new line	coverage
🔵	io/questdb/griffin/engine/join/LtJoinLightRecordCursorFactory.java	3	3	100.00%
🔵	io/questdb/griffin/engine/join/AsOfJoinLightRecordCursorFactory.java	3	3	100.00%