count(*)与count(1)非常常用,可是你真的了解它嘛?
本文基于spark 3.2
先给结论,在spark sql中count(*)不管在运行效率方面,还是在最终展示结果方面 都等同于count(1)。
虽然是我在给大家出作业题,这篇也算是我自己给自己交的一个作业吧
本文大纲
1、案例展示
2、为啥count(*)==count(1) 源码分析
3、spark sql count函数详解(count函数的代码虽然简单,但是开发该代码的同学还是做了一些优化在里面)
案例展示
sql:
代码语言:javascript复制SELECT A,COUNT(*) AS CNT FROM TESTDATA2 GROUP BY A
从上面打印可以看到,count(*)在生成未解析的逻辑执行计划时就被转换成了count(1)。
也就是说,在spark sql中count(*)==count(1)。
源码分析
从sql生成的AstTree上来看,处理函数的节点是functioncall:
在AstBuilder类中找到visitFunctionCall方法:
spark sql count函数详解
count函数的代码虽然简单,但是开发该代码的同学还是做了一些优化在里面。
count函数功能:
1、count(*) 返回检索到的行的总数,包括包含null的行。
2、count(expr[, expr...]) 返回提供的表达式均为非空的行数。
3、count(DISTINCT expr[, expr...]) 返回提供的表达式唯一且非空的行数。
一个点:
判断expression是不是为null,先从expression的nullable的属性里拿,如果nullable的属性不能判断了,再用isnull来判断具体的值是不是为null。
这样在一定程度上能节省计算资源。比如count(1)这样的常量,1一定是不为null的,属性里就可以确定了,不用再走一篇eval取值。
代码语言:javascript复制
package org.apache.spark.sql.catalyst.expressions.aggregate
import org.apache.spark.sql.catalyst.analysis.TypeCheckResult
import org.apache.spark.sql.catalyst.dsl.expressions._
import org.apache.spark.sql.catalyst.expressions._
import org.apache.spark.sql.catalyst.trees.TreePattern.{COUNT, TreePattern}
import org.apache.spark.sql.internal.SQLConf
import org.apache.spark.sql.types._
// scalastyle:off line.size.limit
@ExpressionDescription(
usage = """
_FUNC_(*) - Returns the total number of retrieved rows, including rows containing null.
_FUNC_(expr[, expr...]) - Returns the number of rows for which the supplied expression(s) are all non-null.
_FUNC_(DISTINCT expr[, expr...]) - Returns the number of rows for which the supplied expression(s) are unique and non-null.
""",
examples = """
Examples:
> SELECT _FUNC_(*) FROM VALUES (NULL), (5), (5), (20) AS tab(col);
4
> SELECT _FUNC_(col) FROM VALUES (NULL), (5), (5), (20) AS tab(col);
3
> SELECT _FUNC_(DISTINCT col) FROM VALUES (NULL), (5), (5), (10) AS tab(col);
2
""",
group = "agg_funcs",
since = "1.0.0")
// scalastyle:on line.size.limit
case class Count(children: Seq[Expression]) extends DeclarativeAggregate {
override def nullable: Boolean = false
final override val nodePatterns: Seq[TreePattern] = Seq(COUNT)
// Return data type.
override def dataType: DataType = LongType
override def checkInputDataTypes(): TypeCheckResult = {
if (children.isEmpty && !SQLConf.get.getConf(SQLConf.ALLOW_PARAMETERLESS_COUNT)) {
TypeCheckResult.TypeCheckFailure(s"$prettyName requires at least one argument. "
s"If you have to call the function $prettyName without arguments, set the legacy "
s"configuration `${SQLConf.ALLOW_PARAMETERLESS_COUNT.key}` as true")
} else {
TypeCheckResult.TypeCheckSuccess
}
}
protected lazy val count = AttributeReference("count", LongType, nullable = false)()
override lazy val aggBufferAttributes = count :: Nil
override lazy val initialValues = Seq(
/* count = */ Literal(0L)
)
override lazy val mergeExpressions = Seq(
/* count = */ count.left count.right
)
override lazy val evaluateExpression = count
override def defaultResult: Option[Literal] = Option(Literal(0L))
override lazy val updateExpressions = {
//去掉为空的expression(这一层的判断是走expression里的nullable的属性)
val nullableChildren = children.filter(_.nullable)
if (nullableChildren.isEmpty) {
Seq(
/* count = */ count 1L
)
} else {
// expression的nullable的属性只能依赖拿到具体值判断时,用isnull方法来判断是不是空值
Seq(
/* count = */ If(nullableChildren.map(IsNull).reduce(Or), count, count 1L)
)
}
}
override protected def withNewChildrenInternal(newChildren: IndexedSeq[Expression]): Count =
copy(children = newChildren)
}
object Count {
def apply(child: Expression): Count = Count(child :: Nil)
}