[HIVE-13685] Vectorization: Date*ColScalar parses the scalar in every loop

• Type: Improvement
• Status: Patch Available
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.1.0
• Component/s: Vectorization
• Target Version/s: 2.1.2
• Assignee: Gopal Vijayaraghavan
• Reporter: Gopal Vijayaraghavan
• Created: 04/May/16 07:42
• Updated: 20/Nov/16 20:31

Description

@Override
  public void evaluate(VectorizedRowBatch batch) {
...
switch (inputTypes[0]) {
...
      case STRING:
      case CHAR:
      case VARCHAR:
        try {
          baseDate = formatter.parse(new String(stringValue, "UTF-8"));
          break;

This is parsing the constant scalar for every evaluate() call.

Parsing the constant in the initialize op will save thousands of expensive calls into Calendar operations.