Mailing-List: contact dev-help@drill.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@drill.apache.org
From: jacques-n <git@git.apache.org>
To: dev@drill.apache.org
Reply-To: dev@drill.apache.org
References: <git-pr-397-drill@git.apache.org>
In-Reply-To: <git-pr-397-drill@git.apache.org>
Subject: [GitHub] drill pull request: Drill 4372 review
Content-Type: text/plain
Message-Id: <20160302181905.9B6B7E05E1@git1-us-west.apache.org>
Date: Wed,  2 Mar 2016 18:19:05 +0000 (UTC)

Github user jacques-n commented on a diff in the pull request:

    https://github.com/apache/drill/pull/397#discussion_r54765921
  
    --- Diff: exec/java-exec/src/main/java/org/apache/drill/exec/expr/fn/DrillFunctionRegistry.java ---
    @@ -92,38 +94,110 @@ public DrillFunctionRegistry(ScanResult classpathScan) {
       }
     
       public int size(){
    -    return methods.size();
    +    return registeredFunctions.size();
       }
     
       /** Returns functions with given name. Function name is case insensitive. */
       public List<DrillFuncHolder> getMethods(String name) {
    -    return this.methods.get(name.toLowerCase());
    +    return this.registeredFunctions.get(name.toLowerCase());
    +  }
    +
    +  public Collection<DrillFuncHolder> getAllMethods() {
    +    return Collections.unmodifiableCollection(registeredFunctions.values());
       }
     
       public void register(DrillOperatorTable operatorTable) {
    -    SqlOperator op;
    -    for (Entry<String, Collection<DrillFuncHolder>> function : methods.asMap().entrySet()) {
    -      Set<Integer> argCounts = Sets.newHashSet();
    -      String name = function.getKey().toUpperCase();
    +    for (Entry<String, Collection<DrillFuncHolder>> function : registeredFunctions.asMap().entrySet()) {
    +      final ArrayListMultimap<Pair<Integer, Integer>, DrillFuncHolder> functions = ArrayListMultimap.create();
    +      final ArrayListMultimap<Integer, DrillFuncHolder> aggregateFunctions = ArrayListMultimap.create();
    +      final String name = function.getKey().toUpperCase();
    +      boolean isDeterministic = true;
           for (DrillFuncHolder func : function.getValue()) {
    -        if (argCounts.add(func.getParamCount())) {
    -          if (func.isAggregating()) {
    -            op = new DrillSqlAggOperator(name, func.getParamCount());
    -          } else {
    -            boolean isDeterministic;
    -            // prevent Drill from folding constant functions with types that cannot be materialized
    -            // into literals
    -            if (DrillConstExecutor.NON_REDUCIBLE_TYPES.contains(func.getReturnType().getMinorType())) {
    -              isDeterministic = false;
    -            } else {
    -              isDeterministic = func.isDeterministic();
    -            }
    -            op = new DrillSqlOperator(name, func.getParamCount(), func.getReturnType(), isDeterministic);
    -          }
    -          operatorTable.add(function.getKey(), op);
    +        final int paramCount = func.getParamCount();
    +        if(func.isAggregating()) {
    +          aggregateFunctions.put(paramCount, func);
    +        } else {
    +          final Pair<Integer, Integer> argNumerRange = getArgNumerRange(name, func);
    +          functions.put(argNumerRange, func);
             }
    +
    +        if(!func.isDeterministic()) {
    +          isDeterministic = false;
    +        }
    +      }
    +      for (Entry<Pair<Integer, Integer>, Collection<DrillFuncHolder>> entry : functions.asMap().entrySet()) {
    +        final DrillSqlOperator drillSqlOperator;
    +        final Pair<Integer, Integer> range = entry.getKey();
    +        final int max = range.getRight();
    +        final int min = range.getLeft();
    +        drillSqlOperator = new DrillSqlOperator(
    +            name,
    +            Lists.newArrayList(entry.getValue()),
    +            min,
    +            max,
    +            isDeterministic);
    +        operatorTable.add(name, drillSqlOperator);
    +      }
    +      for (Entry<Integer, Collection<DrillFuncHolder>> entry : aggregateFunctions.asMap().entrySet()) {
    +        operatorTable.add(name, new DrillSqlAggOperator(name, Lists.newArrayList(entry.getValue()), entry.getKey()));
           }
         }
    +
    +    registerCalcitePlaceHolderFunction(operatorTable);
    +  }
    +
    +  /**
    +   * These {@link DrillSqlOperator} merely act as a placeholder so that Calcite
    +   * allows convert_to(), convert_from(), flatten(), date_part() functions in SQL.
    +   */
    +  private void registerCalcitePlaceHolderFunction(DrillOperatorTable operatorTable) {
    +    final String convert_to = "CONVERT_TO";
    +    final String convert_from = "CONVERT_FROM";
    +    final String flatten = "FLATTEN";
    +    final String date_part = "DATE_PART";
    +
    +    operatorTable.add(convert_to,
    +        new DrillSqlOperator(convert_to,
    +            2,
    +            true));
    +    operatorTable.add(convert_from,
    +        new DrillSqlOperator(convert_from,
    +            2,
    +            true));
    +    operatorTable.add(flatten,
    +        new DrillSqlOperator(flatten,
    +            1,
    +            true));
    +    operatorTable.add(date_part,
    +        new DrillSqlOperator(date_part,
    +            2,
    +            true));
       }
     
    +  private Pair<Integer, Integer> getArgNumerRange(final String name, final DrillFuncHolder func) {
    +    switch(name.toUpperCase()) {
    +      case "CONCAT":
    +        return Pair.of(1, Integer.MAX_VALUE);
    +
    +      // Drill does not have a FunctionTemplate for the lpad/rpad with two arguments.
    +      // It relies on DrillOptiq.java to add a third dummy argument to be acceptable
    +      // by the FunctionTemplate in StringFunctions.java
    +      case "LPAD":
    +      case "RPAD":
    +        return Pair.of(2, 3);
    +
    +      // Similar to the reason above, DrillOptiq.java is used for rewritting
    +      case "LTRIM":
    +      case "RTRIM":
    +      case "BTRIM":
    +        return Pair.of(1, 2);
    +
    +      // Similar to the reason above, DrillOptiq.java is used for rewritting
    +      case "LENGTH":
    +        return Pair.of(1, 2);
    +
    +      default:
    +        return Pair.of(func.getParamCount(), func.getParamCount());
    +    }
    +  }
    --- End diff --
    
    Your last suggestion is what I'm saying. Just implement a 2 argument implementation then everything works the same as everything else. The 2 argument version can share code/delegate the three argument version. 


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---