[SPARK-8770][SQL] Create BinaryOperator abstract class.

Our current BinaryExpression abstract class is not for generic binary expressions, i.e. it requires left/right children to have the same type. However, due to its name, contributors build new binary expressions that don't have that assumption (e.g. Sha) and still extend BinaryExpression.

This patch creates a new BinaryOperator abstract class, and update the analyzer o only apply type casting rule there. This patch also adds the notion of "prettyName" to expressions, which defines the user-facing name for the expression.

Author: Reynold Xin <[email protected]>

Closes #7174 from rxin/binary-opterator and squashes the following commits:

f31900d [Reynold Xin] [SPARK-8770][SQL] Create BinaryOperator abstract class.
fceb216 [Reynold Xin] Merge branch 'master' of github.com:apache/spark into binary-opterator
d8518cf [Reynold Xin] Updated Python tests.

Author: rxin

python/pyspark/sql/dataframe.py

@@ -802,11 +802,11 @@ def groupBy(self, *cols):
             Each element should be a column name (string) or an expression (:class:`Column`).
 
         >>> df.groupBy().avg().collect()
-        [Row(AVG(age)=3.5)]
+        [Row(avg(age)=3.5)]
         >>> df.groupBy('name').agg({'age': 'mean'}).collect()
-        [Row(name=u'Alice', AVG(age)=2.0), Row(name=u'Bob', AVG(age)=5.0)]
+        [Row(name=u'Alice', avg(age)=2.0), Row(name=u'Bob', avg(age)=5.0)]
         >>> df.groupBy(df.name).avg().collect()
-        [Row(name=u'Alice', AVG(age)=2.0), Row(name=u'Bob', AVG(age)=5.0)]
+        [Row(name=u'Alice', avg(age)=2.0), Row(name=u'Bob', avg(age)=5.0)]
         >>> df.groupBy(['name', df.age]).count().collect()
         [Row(name=u'Bob', age=5, count=1), Row(name=u'Alice', age=2, count=1)]
         """
@@ -864,10 +864,10 @@ def agg(self, *exprs):
         (shorthand for ``df.groupBy.agg()``).
 
         >>> df.agg({"age": "max"}).collect()
-        [Row(MAX(age)=5)]
+        [Row(max(age)=5)]
         >>> from pyspark.sql import functions as F
         >>> df.agg(F.min(df.age)).collect()
-        [Row(MIN(age)=2)]
+        [Row(min(age)=2)]
         """
         return self.groupBy().agg(*exprs)
 

python/pyspark/sql/functions.py

@@ -266,7 +266,7 @@ def coalesce(*cols):
 
     >>> cDf.select(coalesce(cDf["a"], cDf["b"])).show()
     +-------------+
-    |Coalesce(a,b)|
+    |coalesce(a,b)|
     +-------------+
     |         null|
     |            1|
@@ -275,7 +275,7 @@ def coalesce(*cols):
 
     >>> cDf.select('*', coalesce(cDf["a"], lit(0.0))).show()
     +----+----+---------------+
-    |   a|   b|Coalesce(a,0.0)|
+    |   a|   b|coalesce(a,0.0)|
     +----+----+---------------+
     |null|null|            0.0|
     |   1|null|            1.0|

python/pyspark/sql/group.py

@@ -75,11 +75,11 @@ def agg(self, *exprs):
 
         >>> gdf = df.groupBy(df.name)
         >>> gdf.agg({"*": "count"}).collect()
-        [Row(name=u'Alice', COUNT(1)=1), Row(name=u'Bob', COUNT(1)=1)]
+        [Row(name=u'Alice', count(1)=1), Row(name=u'Bob', count(1)=1)]
 
         >>> from pyspark.sql import functions as F
         >>> gdf.agg(F.min(df.age)).collect()
-        [Row(name=u'Alice', MIN(age)=2), Row(name=u'Bob', MIN(age)=5)]
+        [Row(name=u'Alice', min(age)=2), Row(name=u'Bob', min(age)=5)]
         """
         assert exprs, "exprs should not be empty"
         if len(exprs) == 1 and isinstance(exprs[0], dict):
@@ -110,9 +110,9 @@ def mean(self, *cols):
         :param cols: list of column names (string). Non-numeric columns are ignored.
 
         >>> df.groupBy().mean('age').collect()
-        [Row(AVG(age)=3.5)]
+        [Row(avg(age)=3.5)]
         >>> df3.groupBy().mean('age', 'height').collect()
-        [Row(AVG(age)=3.5, AVG(height)=82.5)]
+        [Row(avg(age)=3.5, avg(height)=82.5)]
         """
 
     @df_varargs_api
@@ -125,9 +125,9 @@ def avg(self, *cols):
         :param cols: list of column names (string). Non-numeric columns are ignored.
 
         >>> df.groupBy().avg('age').collect()
-        [Row(AVG(age)=3.5)]
+        [Row(avg(age)=3.5)]
         >>> df3.groupBy().avg('age', 'height').collect()
-        [Row(AVG(age)=3.5, AVG(height)=82.5)]
+        [Row(avg(age)=3.5, avg(height)=82.5)]
         """
 
     @df_varargs_api
@@ -136,9 +136,9 @@ def max(self, *cols):
         """Computes the max value for each numeric columns for each group.
 
         >>> df.groupBy().max('age').collect()
-        [Row(MAX(age)=5)]
+        [Row(max(age)=5)]
         >>> df3.groupBy().max('age', 'height').collect()
-        [Row(MAX(age)=5, MAX(height)=85)]
+        [Row(max(age)=5, max(height)=85)]
         """
 
     @df_varargs_api
@@ -149,9 +149,9 @@ def min(self, *cols):
         :param cols: list of column names (string). Non-numeric columns are ignored.
 
         >>> df.groupBy().min('age').collect()
-        [Row(MIN(age)=2)]
+        [Row(min(age)=2)]
         >>> df3.groupBy().min('age', 'height').collect()
-        [Row(MIN(age)=2, MIN(height)=80)]
+        [Row(min(age)=2, min(height)=80)]
         """
 
     @df_varargs_api
@@ -162,9 +162,9 @@ def sum(self, *cols):
         :param cols: list of column names (string). Non-numeric columns are ignored.
 
         >>> df.groupBy().sum('age').collect()
-        [Row(SUM(age)=7)]
+        [Row(sum(age)=7)]
         >>> df3.groupBy().sum('age', 'height').collect()
-        [Row(SUM(age)=7, SUM(height)=165)]
+        [Row(sum(age)=7, sum(height)=165)]
         """
 
 

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/HiveTypeCoercion.scala

@@ -150,6 +150,7 @@ object HiveTypeCoercion {
    * Converts string "NaN"s that are in binary operators with a NaN-able types (Float / Double) to
    * the appropriate numeric equivalent.
    */
+  // TODO: remove this rule and make Cast handle Nan.
   object ConvertNaNs extends Rule[LogicalPlan] {
     private val StringNaN = Literal("NaN")
 
@@ -159,19 +160,19 @@ object HiveTypeCoercion {
         case e if !e.childrenResolved => e
 
         /* Double Conversions */
-        case b @ BinaryExpression(StringNaN, right @ DoubleType()) =>
+        case b @ BinaryOperator(StringNaN, right @ DoubleType()) =>
           b.makeCopy(Array(Literal(Double.NaN), right))
-        case b @ BinaryExpression(left @ DoubleType(), StringNaN) =>
+        case b @ BinaryOperator(left @ DoubleType(), StringNaN) =>
           b.makeCopy(Array(left, Literal(Double.NaN)))
 
         /* Float Conversions */
-        case b @ BinaryExpression(StringNaN, right @ FloatType()) =>
+        case b @ BinaryOperator(StringNaN, right @ FloatType()) =>
           b.makeCopy(Array(Literal(Float.NaN), right))
-        case b @ BinaryExpression(left @ FloatType(), StringNaN) =>
+        case b @ BinaryOperator(left @ FloatType(), StringNaN) =>
           b.makeCopy(Array(left, Literal(Float.NaN)))
 
         /* Use float NaN by default to avoid unnecessary type widening */
-        case b @ BinaryExpression(left @ StringNaN, StringNaN) =>
+        case b @ BinaryOperator(left @ StringNaN, StringNaN) =>
           b.makeCopy(Array(left, Literal(Float.NaN)))
       }
     }
@@ -245,12 +246,12 @@ object HiveTypeCoercion {
 
         Union(newLeft, newRight)
 
-      // Also widen types for BinaryExpressions.
+      // Also widen types for BinaryOperator.
       case q: LogicalPlan => q transformExpressions {
         // Skip nodes who's children have not been resolved yet.
         case e if !e.childrenResolved => e
 
-        case b @ BinaryExpression(left, right) if left.dataType != right.dataType =>
+        case b @ BinaryOperator(left, right) if left.dataType != right.dataType =>
           findTightestCommonTypeOfTwo(left.dataType, right.dataType).map { widestType =>
             val newLeft = if (left.dataType == widestType) left else Cast(left, widestType)
             val newRight = if (right.dataType == widestType) right else Cast(right, widestType)
@@ -478,7 +479,7 @@ object HiveTypeCoercion {
 
         // Promote integers inside a binary expression with fixed-precision decimals to decimals,
         // and fixed-precision decimals in an expression with floats / doubles to doubles
-        case b @ BinaryExpression(left, right) if left.dataType != right.dataType =>
+        case b @ BinaryOperator(left, right) if left.dataType != right.dataType =>
           (left.dataType, right.dataType) match {
             case (t, DecimalType.Fixed(p, s)) if intTypeToFixed.contains(t) =>
               b.makeCopy(Array(Cast(left, intTypeToFixed(t)), right))

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/ExpectsInputTypes.scala

@@ -0,0 +1,59 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License.  You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.spark.sql.catalyst.expressions
+
+import org.apache.spark.sql.catalyst.analysis.TypeCheckResult
+import org.apache.spark.sql.types.DataType
+
+
+/**
+ * An trait that gets mixin to define the expected input types of an expression.
+ */
+trait ExpectsInputTypes { self: Expression =>
+
+  /**
+   * Expected input types from child expressions. The i-th position in the returned seq indicates
+   * the type requirement for the i-th child.
+   *
+   * The possible values at each position are:
+   * 1. a specific data type, e.g. LongType, StringType.
+   * 2. a non-leaf data type, e.g. NumericType, IntegralType, FractionalType.
+   * 3. a list of specific data types, e.g. Seq(StringType, BinaryType).
+   */
+  def inputTypes: Seq[Any]
+
+  override def checkInputDataTypes(): TypeCheckResult = {
+    // We will do the type checking in `HiveTypeCoercion`, so always returning success here.
+    TypeCheckResult.TypeCheckSuccess
+  }
+}
+
+/**
+ * Expressions that require a specific `DataType` as input should implement this trait
+ * so that the proper type conversions can be performed in the analyzer.
+ */
+trait AutoCastInputTypes { self: Expression =>
+
+  def inputTypes: Seq[DataType]
+
+  override def checkInputDataTypes(): TypeCheckResult = {
+    // We will always do type casting for `AutoCastInputTypes` in `HiveTypeCoercion`,
+    // so type mismatch error won't be reported here, but for underling `Cast`s.
+    TypeCheckResult.TypeCheckSuccess
+  }
+}