Examples with GroupByDesc org.apache.hadoop.hive.ql.plan.GroupByDesc used on opensource projects

Example 36 with GroupByDesc

use of org.apache.hadoop.hive.ql.plan.GroupByDesc in project hive by apache.

the class SemanticAnalyzer method genMapGroupByForSemijoin.

private Operator genMapGroupByForSemijoin(QB qb, ArrayList<ASTNode> fields, Operator<?> input, GroupByDesc.Mode mode) throws SemanticException {
    RowResolver groupByInputRowResolver = opParseCtx.get(input).getRowResolver();
    RowResolver groupByOutputRowResolver = new RowResolver();
    ArrayList<ExprNodeDesc> groupByKeys = new ArrayList<ExprNodeDesc>();
    ArrayList<String> outputColumnNames = new ArrayList<String>();
    ArrayList<AggregationDesc> aggregations = new ArrayList<AggregationDesc>();
    Map<String, ExprNodeDesc> colExprMap = new HashMap<String, ExprNodeDesc>();
    for (int i = 0; i < fields.size(); ++i) {
        // get the group by keys to ColumnInfo
        ASTNode colName = fields.get(i);
        String[] nm;
        String[] nm2;
        ExprNodeDesc grpByExprNode = genExprNodeDesc(colName, groupByInputRowResolver);
        if (grpByExprNode instanceof ExprNodeColumnDesc) {
            // In most of the cases, this is a column reference
            ExprNodeColumnDesc columnExpr = (ExprNodeColumnDesc) grpByExprNode;
            nm = groupByInputRowResolver.reverseLookup(columnExpr.getColumn());
            nm2 = groupByInputRowResolver.getAlternateMappings(columnExpr.getColumn());
        } else if (grpByExprNode instanceof ExprNodeConstantDesc) {
            // However, it can be a constant too. In that case, we need to track
            // the column that it originated from in the input operator so we can
            // propagate the aliases.
            ExprNodeConstantDesc constantExpr = (ExprNodeConstantDesc) grpByExprNode;
            String inputCol = constantExpr.getFoldedFromCol();
            nm = groupByInputRowResolver.reverseLookup(inputCol);
            nm2 = groupByInputRowResolver.getAlternateMappings(inputCol);
        } else {
            // of the left semijoin
            return input;
        }
        groupByKeys.add(grpByExprNode);
        // generate output column names
        String field = getColumnInternalName(i);
        outputColumnNames.add(field);
        ColumnInfo colInfo2 = new ColumnInfo(field, grpByExprNode.getTypeInfo(), "", false);
        groupByOutputRowResolver.put(nm[0], nm[1], colInfo2);
        if (nm2 != null) {
            groupByOutputRowResolver.addMappingOnly(nm2[0], nm2[1], colInfo2);
        }
        groupByOutputRowResolver.putExpression(colName, colInfo2);
        // establish mapping from the output column to the input column
        colExprMap.put(field, grpByExprNode);
    }
    // Generate group-by operator
    float groupByMemoryUsage = HiveConf.getFloatVar(conf, HiveConf.ConfVars.HIVEMAPAGGRHASHMEMORY);
    float memoryThreshold = HiveConf.getFloatVar(conf, HiveConf.ConfVars.HIVEMAPAGGRMEMORYTHRESHOLD);
    Operator op = putOpInsertMap(OperatorFactory.getAndMakeChild(new GroupByDesc(mode, outputColumnNames, groupByKeys, aggregations, false, groupByMemoryUsage, memoryThreshold, null, false, -1, false), new RowSchema(groupByOutputRowResolver.getColumnInfos()), input), groupByOutputRowResolver);
    op.setColumnExprMap(colExprMap);
    return op;
}