问题导读
1.sqlContext有哪些关键的概念和组件?
2.LogicalPlan里维护着什么方面的数据?
3.LogicalPlan有哪三种类型?
4.对于SQL语句解析时,会调用和SQL匹配的操作方法来进行解析;这些操作分哪四大类?
5.SqlParser的功能是什么?
6.SQL语法表达式支持哪3种操作?
7.Optimizer的功能是什么?
上篇:
sparkSQL1.1入门之二:sparkSQL运行架构
上篇在总体上介绍了sparkSQL的运行架构及其基本实现方法(Tree和Rule的配合),也大致介绍了sparkSQL中涉及到的各个概念和组件。本篇将详细地介绍一下关键的一些概念和组件,由于hiveContext继承自sqlContext,关键的概念和组件类似,只不过后者针对hive的特性做了一些修正和重写,所以本篇就只介绍sqlContext的关键的概念和组件。概念:
组件:
SqlParser Analyzer Optimizer Planner
1:LogicalPlan
在sparkSQL的运行架构中,LogicalPlan贯穿了大部分的过程,其中catalyst中的SqlParser、Analyzer、Optimizer都要对LogicalPlan进行操作。LogicalPlan的定义如下:
abstract class LogicalPlan extends QueryPlan[LogicalPlan] {
self: Product =>
case class Statistics(
sizeInBytes: BigInt
)
lazy val statistics: Statistics = {
if (children.size == 0) {
throw new UnsupportedOperationException(s"LeafNode $nodeName must implement statistics.")
}
Statistics(
sizeInBytes = children.map(_.statistics).map(_.sizeInBytes).product)
}
/**
* Returns the set of attributes that this node takes as
* input from its children.
*/
lazy val inputSet: AttributeSet = AttributeSet(children.flatMap(_.output))
/**
* Returns true if this expression and all its children have been resolved to a specific schema
* and false if it is still contains any unresolved placeholders. Implementations of LogicalPlan
* can override this (e.g.
* [[org.apache.spark.sql.catalyst.analysis.UnresolvedRelation UnresolvedRelation]]
* should return `false`).
*/
lazy val resolved: Boolean = !expressions.exists(!_.resolved) && childrenResolved
/**
* Returns true if all its children of this query plan have been resolved.
*/
def childrenResolved: Boolean = !children.exists(!_.resolved)
/**
* Optionally resolves the given string to a [[NamedExpression]] using the input from all child
* nodes of this LogicalPlan. The attribute is expressed as
* as string in the following form: `[scope].AttributeName.[nested].[fields]...`.
*/
def resolveChildren(name: String): Option[NamedExpression] =
resolve(name, children.flatMap(_.output))
/**
* Optionally resolves the given string to a [[NamedExpression]] based on the output of this
* LogicalPlan. The attribute is expressed as string in the following form:
* `[scope].AttributeName.[nested].[fields]...`.
*/
def resolve(name: String): Option[NamedExpression] =
resolve(name, output)
/** Performs attribute resolution given a name and a sequence of possible attributes. */
protected def resolve(name: String, input: Seq[Attribute]): Option[NamedExpression] = {
val parts = name.split("\\.")
val options = input.flatMap { option =>
val remainingParts =
if (option.qualifiers.contains(parts.head) && parts.size > 1) parts.drop(1) else parts
if (option.name == remainingParts.head) (option, remainingParts.tail.toList) :: Nil else Nil
}
options.distinct match {
case Seq((a, Nil)) => Some(a) // One match, no nested fields, use it.
// One match, but we also need to extract the requested nested field.
case Seq((a, nestedFields)) =>
a.dataType match {
case StructType(fields) =>
Some(Alias(nestedFields.foldLeft(a: Expression)(GetField), nestedFields.last)())
case _ => None // Don't know how to resolve these field references
}
case Seq() => None // No matches.
case ambiguousReferences =>
throw new TreeNodeException(
this, s"Ambiguous references to $name: ${ambiguousReferences.mkString(",")}")
}
}
} 复制代码
在LogicalPlan里维护着一套统计数据和属性数据,也提供了解析方法。同时延伸了三种类型的LogicalPlan:
LeafNode:对应于trees.LeafNode的LogicalPlan UnaryNode:对应于trees.UnaryNode的LogicalPlan BinaryNode:对应于trees.BinaryNode的LogicalPlan
而对于SQL语句解析时,会调用和SQL匹配的操作方法来进行解析;这些操作分四大类,最终生成LeafNode、UnaryNode、BinaryNode中的一种:
basicOperators:一些数据基本操作,如Ioin、Union、Filter、Project、Sort commands:一些命令操作,如SetCommand、CacheCommand partitioning:一些分区操作,如RedistributeData ScriptTransformation:对脚本的处理,如ScriptTransformation LogicalPlan类的总体架构如下所示
'
2:SqlParser
SqlParser的功能就是将SQL语句解析成Unresolved LogicalPlan。现阶段的SqlParser语法解析功能比较简单,支持的语法比较有限。其解析过程中有两个关键组件和一个关键函数:
词法读入器SqlLexical,其作用就是将输入的SQL语句进行扫描、去空、去注释、校验、分词等动作。 SQL语法表达式query,其作用定义SQL语法表达式,同时也定义了SQL语法表达式的具体实现,即将不同的表达式生成不同sparkSQL的Unresolved LogicalPlan。 函数phrase(),上面个两个组件通过调用phrase(query)(new lexical.Scanner(input)),完成对SQL语句的解析;在解析过程中,SqlLexical一边读入,一边解析,如果碰上生成符合SQL语法的表达式时,就调用相应SQL语法表达式的具体实现函数,将SQL语句解析成Unresolved LogicalPlan。
下面看看sparkSQL的整个解析过程和相关组件:
A:解析过程
首先,在sqlContext中使用下面代码调用catalyst.SqlParser:
/*源自 sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala */
protected[sql] val parser = new catalyst.SqlParser
protected[sql] def parseSql(sql: String): LogicalPlan = parser(sql) 复制代码
然后,直接在SqlParser的apply方法中对输入的SQL语句进行解析,解析功能的核心代码就是:
phrase(query)(new lexical.Scanner(input))
/*源自 src/main/scala/org/apache/spark/sql/catalyst/SqlParser.scala */
class SqlParser extends StandardTokenParsers with PackratParsers {
def apply(input: String): LogicalPlan = {
if (input.trim.toLowerCase.startsWith("set")) {
//set设置项的处理
......
} else {
phrase(query)(new lexical.Scanner(input)) match {
case Success(r, x) => r
case x => sys.error(x.toString)
}
}
}
...... 复制代码
可以看得出来,该语句就是调用phrase()函数,使用SQL语法表达式query,对词法读入器lexical读入的SQL语句进行解析,其中词法读入器lexical通过重写语句:override val lexical = new SqlLexical(reservedWords) 调用扩展了功能的SqlLexical。其定义:
/*源自 src/main/scala/org/apache/spark/sql/catalyst/SqlParser.scala */
// Use reflection to find the reserved words defined in this class.
protected val reservedWords =
this.getClass
.getMethods
.filter(_.getReturnType == classOf[Keyword])
.map(_.invoke(this).asInstanceOf[Keyword].str)
override val lexical = new SqlLexical(reservedWords) 复制代码
为了加深对SQL语句解析过程的理解,让我们看看下面这个简单数字表达式解析过程来说明:
import scala.util.parsing.combinator.PackratParsers
import scala.util.parsing.combinator.syntactical._
object mylexical extends StandardTokenParsers with PackratParsers {
//定义分割符
lexical.delimiters ++= List(".", ";", "+", "-", "*")
//定义表达式,支持加,减,乘
lazy val expr: PackratParser[Int] = plus | minus | multi
//加法表示式的实现
lazy val plus: PackratParser[Int] = num ~ "+" ~ num ^^ { case n1 ~ "+" ~ n2 => n1.toInt + n2.toInt}
//减法表达式的实现
lazy val minus: PackratParser[Int] = num ~ "-" ~ num ^^ { case n1 ~ "-" ~ n2 => n1.toInt - n2.toInt}
//乘法表达式的实现
lazy val multi: PackratParser[Int] = num ~ "*" ~ num ^^ { case n1 ~ "*" ~ n2 => n1.toInt * n2.toInt}
lazy val num = numericLit
def parse(input: String) = {
//定义词法读入器myread,并将扫描头放置在input的首位
val myread = new PackratReader(new lexical.Scanner(input))
print("处理表达式 " + input)
phrase(expr)(myread) match {
case Success(result, _) => println(" Success!"); println(result); Some(result)
case n => println(n); println("Err!"); None
}
}
def main(args: Array[String]) {
val prg = "6 * 3" :: "24-/*aaa*/4" :: "a+5" :: "21/3" :: Nil
prg.map(parse)
}
} 复制代码
运行结果:
处理表达式 6 * 3 Success! //lexical对空格进行了处理,得到6*3
18 //6*3符合乘法表达式,调用n1.toInt * n2.toInt,得到结果并返回
处理表达式 24-/*aaa*/4 Success! //lexical对注释进行了处理,得到20-4
20 //20-4符合减法表达式,调用n1.toInt - n2.toInt,得到结果并返回
处理表达式 a+5[1.1] failure: number expected
//lexical在解析到a,发现不是整数型,故报错误位置和内容
a+5
^
Err!
处理表达式 21/3[1.3] failure: ``*'' expected but ErrorToken(illegal character) found
//lexical在解析到/,发现不是分割符,故报错误位置和内容
21/3
^
Err! 复制代码
在运行的时候,首先对表达式 6 * 3 进行解析,词法读入器myread将扫描头置于6的位置;当phrase()函数使用定义好的数字表达式expr处理6 * 3的时候,6 * 3每读入一个词法,就和expr进行匹配,如读入6*和expr进行匹配,先匹配表达式plus,*和+匹配不上;就继续匹配表达式minus,*和-匹配不上;就继续匹配表达式multi,这次匹配上了,等读入3的时候,因为3是num类型,就调用调用n1.toInt * n2.toInt进行计算。
注意,这里的expr、plus、minus、multi、num都是表达式,|、~、^^是复合因子,表达式和复合因子可以组成一个新的表达式,如plus(num ~ "+" ~ num ^^ { case n1 ~ "+" ~ n2 => n1.toInt + n2.toInt})就是一个由num、+、num、函数构成的复合表达式;而expr(plus | minus | multi)是由plus、minus、multi构成的复合表达式;复合因子的含义定义在类scala/util/parsing/combinator/Parsers.scala,下面是几个常用的复合因子:
p ~ q p成功,才会q;放回p,q的结果 p ~> q p成功,才会q,返回q的结果 p <~ q p成功,才会q,返回p的结果 p | q p失败则q,返回第一个成功的结果 p ^^ f 如果p成功,将函数f应用到p的结果上 p ^? f 如果p成功,如果函数f可以应用到p的结果上的话,就将p的结果用f进行转换
针对上面的6 * 3使用的是multi表达式(num ~ "*" ~ num ^^ { case n1 ~ "*" ~ n2 => n1.toInt * n2.toInt}),其含义就是:num后跟*再跟num,如果满足就将使用函数n1.toInt * n2.toInt。
到这里为止,大家应该明白整个解析过程了吧, 。SqlParser的原理和这个表达式解析器使用了一样的原理,只不过是定义的SQL语法表达式query复杂一些,使用的词法读入器更丰富一些而已。下面分别介绍一下相关组件SqlParser、SqlLexical、query。
B:SqlParser
首先,看看SqlParser的UML图:
其次,看看SqlParser的定义,SqlParser继承自类StandardTokenParsers和特质PackratParsers:
其中,PackratParsers:
扩展了scala.util.parsing.combinator.Parsers所提供的parser,做了内存化处理; Packrat解析器实现了回溯解析和递归下降解析,具有无限先行和线性分析时的优势。同时,也支持左递归词法解析。 从Parsers中继承出来的class或trait都可以使用PackratParsers,如:object MyGrammar extends StandardTokenParsers with PackratParsers; PackratParsers将分析结果进行缓存,因此,PackratsParsers需要PackratReader(内存化处理的Reader)作为输入,程序员可以手工创建PackratReader,如production(new PackratReader(new lexical.Scanner(input))),更多的细节参见scala库中/scala/util/parsing/combinator/PackratParsers.scala文件。
StandardTokenParsers是最终继承自Parsers
增加了词法的处理能力(Parsers是字符处理),在StdTokenParsers中定义了四种基本词法:
keyword tokens numeric literal tokens string literal tokens identifier tokens
定义了一个词法读入器lexical,可以进行词法读入
SqlParser在进行解析SQL语句的时候是调用了PackratParsers中phrase():
/*源自 scala/util/parsing/combinator/PackratParsers.scala */
/**
* A parser generator delimiting whole phrases (i.e. programs).
*
* Overridden to make sure any input passed to the argument parser
* is wrapped in a `PackratReader`.
*/
override def phrase[T](p: Parser[T]) = {
val q = super.phrase(p)
new PackratParser[T] {
def apply(in: Input) = in match {
case in: PackratReader[_] => q(in)
case in => q(new PackratReader(in))
}
}
} 复制代码
在解析过程中,一般会定义多个表达式,如上面例子中的plus | minus | multi,一旦前一个表达式不能解析的话,就会调用下一个表达式进行解析:
/*源自 scala/util/parsing/combinator/Parsers.scala */
def append[U >: T](p0: => Parser[U]): Parser[U] = { lazy val p = p0 // lazy argument
Parser{ in => this(in) append p(in)}
} 复制代码
表达式解析正确后,具体的实现函数是在PackratParsers中完成:
/*源自 scala/util/parsing/combinator/PackratParsers.scala */
def memo[T](p: super.Parser[T]): PackratParser[T] = {
new PackratParser[T] {
def apply(in: Input) = {
val inMem = in.asInstanceOf[PackratReader[Elem]]
//look in the global cache if in a recursion
val m = recall(p, inMem)
m match {
//nothing has been done due to recall
case None =>
val base = LR(Failure("Base Failure",in), p, None)
inMem.lrStack = base::inMem.lrStack
//cache base result
inMem.updateCacheAndGet(p,MemoEntry(Left(base)))
//parse the input
val tempRes = p(in)
//the base variable has passed equality tests with the cache
inMem.lrStack = inMem.lrStack.tail
//check whether base has changed, if yes, we will have a head
base.head match {
case None =>
/*simple result*/
inMem.updateCacheAndGet(p,MemoEntry(Right(tempRes)))
tempRes
case s@Some(_) =>
/*non simple result*/
base.seed = tempRes
//the base variable has passed equality tests with the cache
val res = lrAnswer(p, inMem, base)
res
}
case Some(mEntry) => {
//entry found in cache
mEntry match {
case MemoEntry(Left(recDetect)) => {
setupLR(p, inMem, recDetect)
//all setupLR does is change the heads of the recursions, so the seed will stay the same
recDetect match {case LR(seed, _, _) => seed.asInstanceOf[ParseResult[T]]}
}
case MemoEntry(Right(res: ParseResult[_])) => res.asInstanceOf[ParseResult[T]]
}
}
}
}
}
} 复制代码
StandardTokenParsers增加了词法处理能力,SqlParers定义了大量的关键字,重写了词法读入器,将这些关键字应用于词法读入器。
C:SqlLexical
词法读入器SqlLexical扩展了StdLexical的功能,首先增加了大量的关键字:
/*源自 src/main/scala/org/apache/spark/sql/catalyst/SqlParser.scala */
protected val ALL = Keyword("ALL")
protected val AND = Keyword("AND")
protected val AS = Keyword("AS")
protected val ASC = Keyword("ASC")
......
protected val SUBSTR = Keyword("SUBSTR")
protected val SUBSTRING = Keyword("SUBSTRING") 复制代码
其次丰富了分隔符、词法处理、空格注释处理:
/*源自 src/main/scala/org/apache/spark/sql/catalyst/SqlParser.scala */
delimiters += (
"@", "*", "+", "-", "<", "=", "<>", "!=", "<=", ">=", ">", "/", "(", ")",
",", ";", "%", "{", "}", ":", "[", "]"
)
override lazy val token: Parser[Token] = (
identChar ~ rep( identChar | digit ) ^^
{ case first ~ rest => processIdent(first :: rest mkString "") }
| rep1(digit) ~ opt('.' ~> rep(digit)) ^^ {
case i ~ None => NumericLit(i mkString "")
case i ~ Some(d) => FloatLit(i.mkString("") + "." + d.mkString(""))
}
| '\'' ~ rep( chrExcept('\'', '\n', EofCh) ) ~ '\'' ^^
{ case '\'' ~ chars ~ '\'' => StringLit(chars mkString "") }
| '"' ~ rep( chrExcept('"', '\n', EofCh) ) ~ '"' ^^
{ case '"' ~ chars ~ '"' => StringLit(chars mkString "") }
| EofCh ^^^ EOF
| '\'' ~> failure("unclosed string literal")
| '"' ~> failure("unclosed string literal")
| delim
| failure("illegal character")
)
override def identChar = letter | elem('_') | elem('.')
override def whitespace: Parser[Any] = rep(
whitespaceChar
| '/' ~ '*' ~ comment
| '/' ~ '/' ~ rep( chrExcept(EofCh, '\n') )
| '#' ~ rep( chrExcept(EofCh, '\n') )
| '-' ~ '-' ~ rep( chrExcept(EofCh, '\n') )
| '/' ~ '*' ~ failure("unclosed comment")
) 复制代码
最后看看SQL语法表达式query。
D:query
SQL语法表达式支持3种操作:select、insert、cache
/*源自 src/main/scala/org/apache/spark/sql/catalyst/SqlParser.scala */
protected lazy val query: Parser[LogicalPlan] = (
select * (
UNION ~ ALL ^^^ { (q1: LogicalPlan, q2: LogicalPlan) => Union(q1, q2) } |
INTERSECT ^^^ { (q1: LogicalPlan, q2: LogicalPlan) => Intersect(q1, q2) } |
EXCEPT ^^^ { (q1: LogicalPlan, q2: LogicalPlan) => Except(q1, q2)} |
UNION ~ opt(DISTINCT) ^^^ { (q1: LogicalPlan, q2: LogicalPlan) => Distinct(Union(q1, q2)) }
)
| insert | cache
) 复制代码
而这些操作还有具体的定义,如select,这里开始定义了具体的函数,将SQL语句转换成构成Unresolved LogicalPlan的一些Node:
/*源自 src/main/scala/org/apache/spark/sql/catalyst/SqlParser.scala */
protected lazy val select: Parser[LogicalPlan] =
SELECT ~> opt(DISTINCT) ~ projections ~
opt(from) ~ opt(filter) ~
opt(grouping) ~
opt(having) ~
opt(orderBy) ~
opt(limit) <~ opt(";") ^^ {
case d ~ p ~ r ~ f ~ g ~ h ~ o ~ l =>
val base = r.getOrElse(NoRelation)
val withFilter = f.map(f => Filter(f, base)).getOrElse(base)
val withProjection =
g.map {g =>
Aggregate(assignAliases(g), assignAliases(p), withFilter)
}.getOrElse(Project(assignAliases(p), withFilter))
val withDistinct = d.map(_ => Distinct(withProjection)).getOrElse(withProjection)
val withHaving = h.map(h => Filter(h, withDistinct)).getOrElse(withDistinct)
val withOrder = o.map(o => Sort(o, withHaving)).getOrElse(withHaving)
val withLimit = l.map { l => Limit(l, withOrder) }.getOrElse(withOrder)
withLimit
} 复制代码
3:Analyzer
Analyzer的功能就是对来自SqlParser的Unresolved LogicalPlan中的UnresolvedAttribute项和UnresolvedRelation项,对照catalog和FunctionRegistry生成Analyzed LogicalPlan。Analyzer定义了5大类14小类的rule:
/*源自 sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala */
val batches: Seq[Batch] = Seq(
Batch("MultiInstanceRelations", Once,
NewRelationInstances),
Batch("CaseInsensitiveAttributeReferences", Once,
(if (caseSensitive) Nil else LowercaseAttributeReferences :: Nil) : _*),
Batch("Resolution", fixedPoint,
ResolveReferences ::
ResolveRelations ::
ResolveSortReferences ::
NewRelationInstances ::
ImplicitGenerate ::
StarExpansion ::
ResolveFunctions ::
GlobalAggregates ::
UnresolvedHavingClauseAttributes ::
typeCoercionRules :_*),
Batch("Check Analysis", Once,
CheckResolution),
Batch("AnalysisOperators", fixedPoint,
EliminateAnalysisOperators)
) 复制代码
MultiInstanceRelations
NewRelationInstances
CaseInsensitiveAttributeReferences
LowercaseAttributeReferences
Resolution
ResolveReferences
ResolveRelations
ResolveSortReferences
NewRelationInstances
ImplicitGenerate
StarExpansion
ResolveFunctions
GlobalAggregates
UnresolvedHavingClauseAttributes
typeCoercionRules
Check Analysis
CheckResolution
AnalysisOperators
EliminateAnalysisOperators
这些rule都是使用transform对Unresolved
LogicalPlan进行操作,其中typeCoercionRules是对HiveQL语义进行处理,在其下面又定义了多个rule:PropagateTypes、ConvertNaNs、WidenTypes、PromoteStrings、BooleanComparisons、BooleanCasts、StringToIntegralCasts、FunctionArgumentConversion、CaseWhenCoercion、Division,同样了这些rule也是使用transform对Unresolved
LogicalPlan进行操作。这些rule操作后,使得LogicalPlan的信息变得丰满和易懂。下面拿其中的两个rule来简单介绍一下:
比如rule之ResolveReferences,最终调用LogicalPlan的resolveChildren对列名给一名字和序号,如name#67之列的,这样保持列的唯一性:
/*源自 sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala */
object ResolveReferences extends Rule[LogicalPlan] {
def apply(plan: LogicalPlan): LogicalPlan = plan transformUp {
case q: LogicalPlan if q.childrenResolved =>
logTrace(s"Attempting to resolve ${q.simpleString}")
q transformExpressions {
case u @ UnresolvedAttribute(name) =>
// Leave unchanged if resolution fails. Hopefully will be resolved next round.
val result = q.resolveChildren(name).getOrElse(u)
logDebug(s"Resolving $u to $result")
result
}
}
} 复制代码
又比如rule之StarExpansion,其作用就是将Select * Fom tbl中的*展开,赋予列名:
/*源自 sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala */
object StarExpansion extends Rule[LogicalPlan] {
def apply(plan: LogicalPlan): LogicalPlan = plan transform {
// Wait until children are resolved
case p: LogicalPlan if !p.childrenResolved => p
// If the projection list contains Stars, expand it.
case p @ Project(projectList, child) if containsStar(projectList) =>
Project(
projectList.flatMap {
case s: Star => s.expand(child.output)
case o => o :: Nil
},
child)
case t: ScriptTransformation if containsStar(t.input) =>
t.copy(
input = t.input.flatMap {
case s: Star => s.expand(t.child.output)
case o => o :: Nil
}
)
// If the aggregate function argument contains Stars, expand it.
case a: Aggregate if containsStar(a.aggregateExpressions) =>
a.copy(
aggregateExpressions = a.aggregateExpressions.flatMap {
case s: Star => s.expand(a.child.output)
case o => o :: Nil
}
)
}
/**
* Returns true if `exprs` contains a [[Star]].
*/
protected def containsStar(exprs: Seq[Expression]): Boolean =
exprs.collect { case _: Star => true }.nonEmpty
}
} 复制代码
4:Optimizer
Optimizer的功能就是将来自Analyzer的Analyzed LogicalPlan进行多种rule优化,生成Optimized LogicalPlan。Optimizer定义了3大类12个小类的优化rule:
/*源自 sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala */
object Optimizer extends RuleExecutor[LogicalPlan] {
val batches =
Batch("Combine Limits", FixedPoint(100),
CombineLimits) ::
Batch("ConstantFolding", FixedPoint(100),
NullPropagation,
ConstantFolding,
LikeSimplification,
BooleanSimplification,
SimplifyFilters,
SimplifyCasts,
SimplifyCaseConversionExpressions) ::
Batch("Filter Pushdown", FixedPoint(100),
CombineFilters,
PushPredicateThroughProject,
PushPredicateThroughJoin,
ColumnPruning) :: Nil
} 复制代码
Combine Limits 合并Limit
CombineLimits:将两个相邻的limit合为一个
ConstantFolding 常量叠加
NullPropagation 空格处理 ConstantFolding:常量叠加 LikeSimplification:like表达式简化 BooleanSimplification:布尔表达式简化 SimplifyFilters:Filter简化 SimplifyCasts:Cast简化 SimplifyCaseConversionExpressions:CASE大小写转化表达式简化
Filter Pushdown Filter下推
CombineFilters Filter合并 PushPredicateThroughProject 通过Project谓词下推 PushPredicateThroughJoin 通过Join谓词下推 ColumnPruning 列剪枝
这些优化rule都是使用transform对LogicalPlan进行操作,如合并、删除冗余、简化、剪枝等,是整个LogicalPlan变得更简洁更高效。
比如将两个相邻的limit进行合并,可以使用CombineLimits。象sql("select * from (select * from src limit 5)a limit 3 ") 这样一个SQL语句,会将limit 5和limit 3进行合并,只剩一个一个limit 3。
/*源自 sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala */
object CombineLimits extends Rule[LogicalPlan] {
def apply(plan: LogicalPlan): LogicalPlan = plan transform {
case ll @ Limit(le, nl @ Limit(ne, grandChild)) =>
Limit(If(LessThan(ne, le), ne, le), grandChild)
}
} 复制代码
又比如Null值的处理,可以使用NullPropagation处理。象sql("select count(null) from src where key is not null")这样一个SQL语句会转换成sql("select count(0) from src where key is not null")来处理。
/*源自 sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala */
object NullPropagation extends Rule[LogicalPlan] {
def apply(plan: LogicalPlan): LogicalPlan = plan transform {
case q: LogicalPlan => q transformExpressionsUp {
case e @ Count(Literal(null, _)) => Cast(Literal(0L), e.dataType)
case e @ Sum(Literal(c, _)) if c == 0 => Cast(Literal(0L), e.dataType)
case e @ Average(Literal(c, _)) if c == 0 => Literal(0.0, e.dataType)
case e @ IsNull(c) if !c.nullable => Literal(false, BooleanType)
case e @ IsNotNull(c) if !c.nullable => Literal(true, BooleanType)
case e @ GetItem(Literal(null, _), _) => Literal(null, e.dataType)
case e @ GetItem(_, Literal(null, _)) => Literal(null, e.dataType)
case e @ GetField(Literal(null, _), _) => Literal(null, e.dataType)
case e @ EqualNullSafe(Literal(null, _), r) => IsNull(r)
case e @ EqualNullSafe(l, Literal(null, _)) => IsNull(l)
......
}
}
} 复制代码
对于具体的优化方法可以使用下一章所介绍的hive/console调试方法进行调试,用户可以使用自定义的优化函数,也可以使用sparkSQL提供的优化函数。使用前先定义一个要优化查询,然后查看一下该查询的Analyzed LogicalPlan,再使用优化函数去优化,将生成的Optimized LogicalPlan和Analyzed LogicalPlan进行比较,就可以看到优化的效果。