[DRILL-6122] Order by clause works inconsistently when sorting columns with NaN

• Type: Bug
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 1.13.0
• Component/s: Storage - JSON
• Assignee: Volodymyr Tkach
• Reporter: Alexander Malashevsky
• Created: 30/Jan/18 17:52
• Updated: 30/Jan/18 18:22

Description

AFFECTED_VERSION: drill-1.13.0-SNAPSHOT

AFFECTED_FUNCTIONALITY: ORDER BY, DESC

ISSUE_DESCRIPTION: 'ORDER BY/DESC' clause behaves in different ways when sorting columns containing NaN values. In one case it considers NaN to be the largest value, in another - the smallest one.
Steps:

• Select from the attached test file (orderBy.json, attached)

  SELECT name, attr4 from dfs.tmp.`orderBy.json` order by name, attr4

• Check the attached screen shot (orderByIssue.jpg):
  EXPECTED_RESULT: It was expected the 'ORDER BY' clause to sort attr4 columns data in the same way (most probably NaN should be the largest, see NB)
  ACTUAL_RESULT: attr4 column's values were sorted in different ways: for 'obj1'/'obj3' NaN is the largest, for 'obj2'/'obj4' NaN is the smallest.

NB: Postgres as well as Java's sorting (Collection.sort() / Arrays.sort() methods) treats NaN as the largest value

Issue Links

• is caused by: DRILL-5919 Add non-numeric support for JSON processing