[jira] [Created] (HIVE-19247) StatsOptimizer: Missing stats fast-path for Date/Timestamp

["Gopal V (JIRA)" <ji...@apache.org>]

Gopal V created HIVE-19247:
------------------------------

             Summary: StatsOptimizer: Missing stats fast-path for Date/Timestamp
                 Key: HIVE-19247
                 URL: https://issues.apache.org/jira/browse/HIVE-19247
             Project: Hive
          Issue Type: Bug
          Components: Statistics
    Affects Versions: 2.3.2, 2.2.0, 3.0.0
            Reporter: Gopal V


{code}
2018-04-19T18:57:24,268 DEBUG [67259108-c184-4c92-9e18-9e2969244442 HiveServer2-Handler-Pool: Thread-73]: optimizer.StatsOptimizer (StatsOptimizer.java:process(614)) - Unsupported type: date encountered in metadata optimizer for column : jour
{code}

{code}
if (udaf instanceof GenericUDAFMin) {
            ExprNodeColumnDesc colDesc = (ExprNodeColumnDesc)exprMap.get(((ExprNodeColumnDesc)aggr.getParameters().get(0)).getColumn());
            String colName = colDesc.getColumn();
            StatType type = getType(colDesc.getTypeString());
            if (!tbl.isPartitioned()) {
              if (!StatsSetupConst.areColumnStatsUptoDate(tbl.getParameters(), colName)) {
                Logger.debug("Stats for table : " + tbl.getTableName() + " column " + colName
                    + " are not up to date.");
                return null;
              }
              ColumnStatisticsData statData = hive.getMSC().getTableColumnStatistics(
                  tbl.getDbName(), tbl.getTableName(), Lists.newArrayList(colName))
                  .get(0).getStatsData();
              String name = colDesc.getTypeString().toUpperCase();
              switch (type) {
                case Integeral: {
                  LongSubType subType = LongSubType.valueOf(name);
                  LongColumnStatsData lstats = statData.getLongStats();
                  if (lstats.isSetLowValue()) {
                    oneRow.add(subType.cast(lstats.getLowValue()));
                  } else {
                    oneRow.add(null);
                  }
                  break;
                }
                case Double: {
                  DoubleSubType subType = DoubleSubType.valueOf(name);
                  DoubleColumnStatsData dstats = statData.getDoubleStats();
                  if (dstats.isSetLowValue()) {
                    oneRow.add(subType.cast(dstats.getLowValue()));
                  } else {
                    oneRow.add(null);
                  }
                  break;
                }
                default: // unsupported type
                  Logger.debug("Unsupported type: " + colDesc.getTypeString() + " encountered in " +
                      "metadata optimizer for column : " + colName);
                  return null;
              }
            }
{code}

{code}
    enum StatType{
      Integeral,
      Double,
      String,
      Boolean,
      Binary,
      Unsupported
    }

    enum LongSubType {
      BIGINT { @Override
      Object cast(long longValue) { return longValue; } },
      INT { @Override
      Object cast(long longValue) { return (int)longValue; } },
      SMALLINT { @Override
      Object cast(long longValue) { return (short)longValue; } },
      TINYINT { @Override
      Object cast(long longValue) { return (byte)longValue; } };

      abstract Object cast(long longValue);
    }
{code}

Date/Timestamp are stored as Integral stats (& also the typo there).



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)