Spark SQL 初探

首页 > 代码库 > Spark SQL 初探

2024-07-10 16:50:44 225人阅读

早在Spark Submit 2013里就有介绍到Spark SQL，不过更多的是介绍Catalyst查询优化框架。经过一年的开发，在今年Spark Submit 2014上，Databricks宣布放弃Shark 的开发，而转投Spark SQL，理由是Shark继承了Hive太多，优化出现了瓶颈，如图：

今天把Spark最新的代码签了下来，测试了一下：

1、编译SparkSQL

-bash-3.2$ git config --global http.sslVerify false
-bash-3.2$ git clone https://github.com/apache/spark.git
正克隆到 'spark'...
remote: Reusing existing pack: 107821, done.
remote: Counting objects: 103, done.
remote: Compressing objects: 100% (72/72), done.
remote: Total 107924 (delta 20), reused 64 (delta 16)
Receiving objects: 100% (107924/107924), 69.06 MiB | 3.39 MiB/s, done.
Resolving deltas: 100% (50174/50174), done.

这里还是需要先build一下的，sbt/sbt assembly（如何build匹配版本，请参考Spark编译及集群搭建）

运行 sbt/sbt hive/console也会进行编译。

最新的spark sql提供了一个console，在这里可以直接的运行交互式查下，也提供了几个例子。

2、执行Spark SQL

官方提供给我们了一个测试用例。通过查看log，find . -name TestHive* 找到了位于：

/app/hadoop/shengli/spark/sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveTestHive.scala 有兴趣可以自己打开编译调试下看看。

首先进入控制台：

sbt/sbt hive/console
[info] Starting scala interpreter...
[info] 
import org.apache.spark.sql.catalyst.analysis._
import org.apache.spark.sql.catalyst.dsl._
import org.apache.spark.sql.catalyst.errors._
import org.apache.spark.sql.catalyst.expressions._
import org.apache.spark.sql.catalyst.plans.logical._
import org.apache.spark.sql.catalyst.rules._
import org.apache.spark.sql.catalyst.types._
import org.apache.spark.sql.catalyst.util._
import org.apache.spark.sql.execution
import org.apache.spark.sql.hive._
import org.apache.spark.sql.hive.test.TestHive._
import org.apache.spark.sql.parquet.ParquetTestData
Welcome to Scala version 2.10.4 (Java HotSpot(TM) 64-Bit Server VM, Java 1.6.0_20).
Type in expressions to have them evaluated.
Type :help for more information.
scala>

查看一下当前RunTime下都提供了哪些方法：

scala> <init> ParquetTestData               SqlCmd booleanToLiteral              byteToLiteral classOf                       clear createParquetFile             createSchemaRDD doubleToLiteral               emptyResult executeSql                    execution getAll                        getClass hiveDevHome                   hiveFilesTemp hiveconf                      hiveql intToLiteral                  isCached loadTestTable                 logger ne                            notify originalUdfs                  outputBuffer planner                       prepareForExecution runHive                       runSqlHive sparkContext                  sql table                         testTables uncacheTable                  wait

DslAttribute                  DslExpression                 DslString                     DslSymbol analyzer                      autoConvertJoinSize           binaryToLiteral cacheTable                    cacheTables                   catalog clone                         configure                     contains createTable                   decimalToLiteral              describedTable eq                            equals                        executePlan finalize                      floatToLiteral                get getHiveFile                   getOption                     hashCode hiveHome                      hivePlanner                   hiveQTestUtilTables hql                           inRepoTests                   inferSchema joinBroadcastTables           jsonFile                      jsonRDD logicalPlanToSparkQuery       longToLiteral                 metastorePath notifyAll                     numShufflePartitions          optimizer parquetFile                   parseSql                      parser registerRDDAsTable            registerTestTable             reset sessionState                  set                           shortToLiteral stringToLiteral               symbolToUnresolvedAttribute   synchronized timestampToLiteral            toDebugString                 toString warehousePath

我们发现，这个测试用例里面有一个testTables，由于这些成员都是lazy的，所以一开始没有被加载：

查看测试用例要加载哪些表：

scala> testTables
14/07/02 18:45:59 INFO spark.SecurityManager: Changing view acls to: hadoop
14/07/02 18:45:59 INFO spark.SecurityManager: SecurityManager: authentication disabled; ui acls disabled; users with view permissions: Set(hadoop)
14/07/02 18:46:00 INFO slf4j.Slf4jLogger: Slf4jLogger started
14/07/02 18:46:00 INFO Remoting: Starting remoting
14/07/02 18:46:00 INFO Remoting: Remoting started; listening on addresses :[akka.tcp://spark@web02.dw:42984]
14/07/02 18:46:00 INFO Remoting: Remoting now listens on addresses: [akka.tcp://spark@web02.dw:42984]
14/07/02 18:46:00 INFO spark.SparkEnv: Registering MapOutputTracker
14/07/02 18:46:00 INFO spark.SparkEnv: Registering BlockManagerMaster
14/07/02 18:46:00 INFO storage.DiskBlockManager: Created local directory at /tmp/spark-local-20140702184600-9e16
14/07/02 18:46:00 INFO network.ConnectionManager: Bound socket to port 48348 with id = ConnectionManagerId(web02.dw,48348)
14/07/02 18:46:00 INFO storage.MemoryStore: MemoryStore started with capacity 1097.0 MB
14/07/02 18:46:00 INFO storage.BlockManagerMaster: Trying to register BlockManager
14/07/02 18:46:00 INFO storage.BlockManagerInfo: Registering block manager web02.dw:48348 with 1097.0 MB RAM
14/07/02 18:46:00 INFO storage.BlockManagerMaster: Registered BlockManager
14/07/02 18:46:00 INFO spark.HttpServer: Starting HTTP Server
14/07/02 18:46:01 INFO server.Server: jetty-8.1.14.v20131031
14/07/02 18:46:01 INFO server.AbstractConnector: Started SocketConnector@0.0.0.0:36260
14/07/02 18:46:01 INFO broadcast.HttpBroadcast: Broadcast server started at http://10.1.8.207:36260
14/07/02 18:46:01 INFO spark.HttpFileServer: HTTP File server directory is /tmp/spark-ca40f66c-edc3-484f-b317-d3f512aab244
14/07/02 18:46:01 INFO spark.HttpServer: Starting HTTP Server
14/07/02 18:46:01 INFO server.Server: jetty-8.1.14.v20131031
14/07/02 18:46:01 INFO server.AbstractConnector: Started SocketConnector@0.0.0.0:57821
14/07/02 18:46:01 INFO server.Server: jetty-8.1.14.v20131031
14/07/02 18:46:02 INFO server.AbstractConnector: Started SelectChannelConnector@0.0.0.0:4040
14/07/02 18:46:02 INFO ui.SparkUI: Started SparkUI at http://web02.dw:4040
metastore path is /tmp/sparkHiveMetastore8060064816530828092
warehousePath path is /tmp/sparkHiveWarehouse5366068035857129261
hiveHome path is Some(/home/hadoop/Java/lib/hive-0.6.0)
hiveDevHome path is None
res0: scala.collection.mutable.HashMap[String,org.apache.spark.sql.hive.test.TestHive.TestTable] = Map(sales -> TestTable(sales,WrappedArray(<function0>, <function0>)), src -> TestTable(src,WrappedArray(<function0>, <function0>)), src1 -> TestTable(src1,WrappedArray(<function0>, <function0>)), serdeins -> TestTable(serdeins,WrappedArray(<function0>, <function0>)), src_thrift -> TestTable(src_thrift,WrappedArray(<function0>)), srcpart -> TestTable(srcpart,WrappedArray(<function0>)), episodes -> TestTable(episodes,WrappedArray(<function0>, <function0>)), srcpart1 -> TestTable(srcpart1,WrappedArray(<function0>)))

测试select语句

1.首先声明一个sql

2.这是测试用例会用hive的metastore，创建一个derby的数据库

3.创建上述的所以表，并把数据加载进去。

4.Parse这条select * from sales 语句。

5. 生成SchemaRDD并产生查询计划。

6. 当对querySales这个RDD执行Action的时候，会计算这条sql的执行。

以下是执行的详细结果：（可以看到log打出的大概执行步骤）

scala> val querySales = sql("select * from sales")
14/07/02 18:51:19 INFO test.TestHive$: Loading test table sales
14/07/02 18:51:19 INFO parse.ParseDriver: Parsing command: CREATE TABLE IF NOT EXISTS sales (key STRING, value INT)
ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.RegexSerDe'
WITH SERDEPROPERTIES ("input.regex" = "([^ ]*)  ([^ ]*)")
       
14/07/02 18:51:19 INFO parse.ParseDriver: Parse Completed
14/07/02 18:51:19 INFO analysis.Analyzer: Max iterations (2) reached for batch MultiInstanceRelations
14/07/02 18:51:19 INFO analysis.Analyzer: Max iterations (2) reached for batch CaseInsensitiveAttributeReferences
14/07/02 18:51:19 INFO sql.SQLContext$$anon$1: Max iterations (2) reached for batch Add exchange
14/07/02 18:51:19 INFO sql.SQLContext$$anon$1: Max iterations (2) reached for batch Prepare Expressions
14/07/02 18:51:19 INFO ql.Driver: <PERFLOG method=Driver.run>
14/07/02 18:51:19 INFO ql.Driver: <PERFLOG method=TimeToSubmit>
14/07/02 18:51:19 INFO ql.Driver: <PERFLOG method=compile>
14/07/02 18:51:19 INFO ql.Driver: <PERFLOG method=parse>
14/07/02 18:51:19 INFO parse.ParseDriver: Parsing command: CREATE TABLE IF NOT EXISTS sales (key STRING, value INT)
ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.RegexSerDe'
WITH SERDEPROPERTIES ("input.regex" = "([^ ]*)  ([^ ]*)")
       
14/07/02 18:51:19 INFO parse.ParseDriver: Parse Completed
14/07/02 18:51:19 INFO ql.Driver: </PERFLOG method=parse start=1404298279883 end=1404298279885 duration=2>
14/07/02 18:51:19 INFO ql.Driver: <PERFLOG method=semanticAnalyze>
14/07/02 18:51:19 INFO parse.SemanticAnalyzer: Starting Semantic Analysis
14/07/02 18:51:19 INFO parse.SemanticAnalyzer: Creating table sales position=27
14/07/02 18:51:20 INFO metastore.HiveMetaStore: 0: Opening raw store with implemenation class:org.apache.hadoop.hive.metastore.ObjectStore
14/07/02 18:51:20 INFO metastore.ObjectStore: ObjectStore, initialize called
14/07/02 18:51:20 INFO DataNucleus.Persistence: Property datanucleus.cache.level2 unknown - will be ignored
14/07/02 18:51:21 WARN bonecp.BoneCPConfig: Max Connections < 1. Setting to 20
14/07/02 18:51:25 INFO metastore.ObjectStore: Setting MetaStore object pin classes with hive.metastore.cache.pinobjtypes="Table,StorageDescriptor,SerDeInfo,Partition,Database,Type,FieldSchema,Order"
14/07/02 18:51:25 INFO metastore.ObjectStore: Initialized ObjectStore
14/07/02 18:51:26 WARN bonecp.BoneCPConfig: Max Connections < 1. Setting to 20
14/07/02 18:51:26 WARN metastore.ObjectStore: Version information not found in metastore. hive.metastore.schema.verification is not enabled so recording the schema version 0.12.0
14/07/02 18:51:27 INFO metastore.HiveMetaStore: 0: get_table : db=default tbl=sales
14/07/02 18:51:27 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=get_table : db=default tbl=sales
14/07/02 18:51:27 INFO DataNucleus.Datastore: The class "org.apache.hadoop.hive.metastore.model.MFieldSchema" is tagged as "embedded-only" so does not have its own datastore table.
14/07/02 18:51:27 INFO DataNucleus.Datastore: The class "org.apache.hadoop.hive.metastore.model.MOrder" is tagged as "embedded-only" so does not have its own datastore table.
14/07/02 18:51:28 INFO ql.Driver: Semantic Analysis Completed
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=semanticAnalyze start=1404298279885 end=1404298288331 duration=8446>
14/07/02 18:51:28 INFO ql.Driver: Returning Hive schema: Schema(fieldSchemas:null, properties:null)
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=compile start=1404298279840 end=1404298288340 duration=8500>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=Driver.execute>
14/07/02 18:51:28 INFO ql.Driver: Starting command: CREATE TABLE IF NOT EXISTS sales (key STRING, value INT)
ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.RegexSerDe'
WITH SERDEPROPERTIES ("input.regex" = "([^ ]*)  ([^ ]*)")
       
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=TimeToSubmit start=1404298279840 end=1404298288351 duration=8511>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=runTasks>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=task.DDL.Stage-0>
14/07/02 18:51:28 INFO metastore.HiveMetaStore: 0: create_table: Table(tableName:sales, dbName:default, owner:hadoop, createTime:1404298288, lastAccessTime:0, retention:0, sd:StorageDescriptor(cols:[FieldSchema(name:key, type:string, comment:null), FieldSchema(name:value, type:int, comment:null)], location:null, inputFormat:org.apache.hadoop.mapred.TextInputFormat, outputFormat:org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat, compressed:false, numBuckets:-1, serdeInfo:SerDeInfo(name:null, serializationLib:org.apache.hadoop.hive.serde2.RegexSerDe, parameters:{serialization.format=1, input.regex=([^ ]*)        ([^ ]*)}), bucketCols:[], sortCols:[], parameters:{}, skewedInfo:SkewedInfo(skewedColNames:[], skewedColValues:[], skewedColValueLocationMaps:{}), storedAsSubDirectories:false), partitionKeys:[], parameters:{}, viewOriginalText:null, viewExpandedText:null, tableType:MANAGED_TABLE, privileges:PrincipalPrivilegeSet(userPrivileges:null, groupPrivileges:null, rolePrivileges:null))
14/07/02 18:51:28 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=create_table: Table(tableName:sales, dbName:default, owner:hadoop, createTime:1404298288, lastAccessTime:0, retention:0, sd:StorageDescriptor(cols:[FieldSchema(name:key, type:string, comment:null), FieldSchema(name:value, type:int, comment:null)], location:null, inputFormat:org.apache.hadoop.mapred.TextInputFormat, outputFormat:org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat, compressed:false, numBuckets:-1, serdeInfo:SerDeInfo(name:null, serializationLib:org.apache.hadoop.hive.serde2.RegexSerDe, parameters:{serialization.format=1, input.regex=([^ ]*)       ([^ ]*)}), bucketCols:[], sortCols:[], parameters:{}, skewedInfo:SkewedInfo(skewedColNames:[], skewedColValues:[], skewedColValueLocationMaps:{}), storedAsSubDirectories:false), partitionKeys:[], parameters:{}, viewOriginalText:null, viewExpandedText:null, tableType:MANAGED_TABLE, privileges:PrincipalPrivilegeSet(userPrivileges:null, groupPrivileges:null, rolePrivileges:null))
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=task.DDL.Stage-0 start=1404298288351 end=1404298288589 duration=238>
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=runTasks start=1404298288351 end=1404298288589 duration=238>
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=Driver.execute start=1404298288340 end=1404298288589 duration=249>
14/07/02 18:51:28 INFO ql.Driver: OK
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=releaseLocks>
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=releaseLocks start=1404298288590 end=1404298288590 duration=0>
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=Driver.run start=1404298279839 end=1404298288590 duration=8751>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=releaseLocks>
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=releaseLocks start=1404298288590 end=1404298288590 duration=0>
14/07/02 18:51:28 INFO parse.ParseDriver: Parsing command: LOAD DATA LOCAL INPATH 'sql/hive/src/test/resources/data/files/sales.txt' INTO TABLE sales
14/07/02 18:51:28 INFO parse.ParseDriver: Parse Completed
14/07/02 18:51:28 INFO analysis.Analyzer: Max iterations (2) reached for batch MultiInstanceRelations
14/07/02 18:51:28 INFO analysis.Analyzer: Max iterations (2) reached for batch CaseInsensitiveAttributeReferences
14/07/02 18:51:28 INFO sql.SQLContext$$anon$1: Max iterations (2) reached for batch Add exchange
14/07/02 18:51:28 INFO sql.SQLContext$$anon$1: Max iterations (2) reached for batch Prepare Expressions
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=Driver.run>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=TimeToSubmit>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=compile>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=parse>
14/07/02 18:51:28 INFO parse.ParseDriver: Parsing command: LOAD DATA LOCAL INPATH 'sql/hive/src/test/resources/data/files/sales.txt' INTO TABLE sales
14/07/02 18:51:28 INFO parse.ParseDriver: Parse Completed
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=parse start=1404298288629 end=1404298288629 duration=0>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=semanticAnalyze>
14/07/02 18:51:28 INFO metastore.HiveMetaStore: 0: get_table : db=default tbl=sales
14/07/02 18:51:28 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=get_table : db=default tbl=sales
14/07/02 18:51:28 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
14/07/02 18:51:28 INFO ql.Driver: Semantic Analysis Completed
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=semanticAnalyze start=1404298288630 end=1404298288942 duration=312>
14/07/02 18:51:28 INFO ql.Driver: Returning Hive schema: Schema(fieldSchemas:null, properties:null)
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=compile start=1404298288628 end=1404298288943 duration=315>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=Driver.execute>
14/07/02 18:51:28 INFO ql.Driver: Starting command: LOAD DATA LOCAL INPATH 'sql/hive/src/test/resources/data/files/sales.txt' INTO TABLE sales
14/07/02 18:51:28 INFO ql.Driver: </PERFLOG method=TimeToSubmit start=1404298288628 end=1404298288943 duration=315>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=runTasks>
14/07/02 18:51:28 INFO ql.Driver: <PERFLOG method=task.COPY.Stage-0>
14/07/02 18:51:28 INFO exec.Task: Copying data from file:/app/hadoop/spark/sql/hive/src/test/resources/data/files/sales.txt to file:/tmp/hive-hadoop/hive_2014-07-02_18-51-28_629_2309366591646930035-1/-ext-10000
14/07/02 18:51:28 INFO exec.Task: Copying file: file:/app/hadoop/spark/sql/hive/src/test/resources/data/files/sales.txt
14/07/02 18:51:29 INFO ql.Driver: </PERFLOG method=task.COPY.Stage-0 start=1404298288943 end=1404298289037 duration=94>
14/07/02 18:51:29 INFO ql.Driver: <PERFLOG method=task.MOVE.Stage-1>
14/07/02 18:51:29 INFO exec.Task: Loading data to table default.sales from file:/tmp/hive-hadoop/hive_2014-07-02_18-51-28_629_2309366591646930035-1/-ext-10000
14/07/02 18:51:29 INFO metastore.HiveMetaStore: 0: get_table : db=default tbl=sales
14/07/02 18:51:29 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=get_table : db=default tbl=sales
14/07/02 18:51:29 INFO metastore.HiveMetaStore: 0: get_table : db=default tbl=sales
14/07/02 18:51:29 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=get_table : db=default tbl=sales
14/07/02 18:51:29 INFO metastore.HiveMetaStore: 0: alter_table: db=default tbl=sales newtbl=sales
14/07/02 18:51:29 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=alter_table: db=default tbl=sales newtbl=sales
14/07/02 18:51:29 INFO metastore.HiveMetaStore: 0: get_table : db=default tbl=sales
14/07/02 18:51:29 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=get_table : db=default tbl=sales
14/07/02 18:51:29 INFO ql.Driver: </PERFLOG method=task.MOVE.Stage-1 start=1404298289037 end=1404298289196 duration=159>
14/07/02 18:51:29 INFO ql.Driver: <PERFLOG method=task.STATS.Stage-2>
14/07/02 18:51:29 INFO exec.StatsTask: Executing stats task
14/07/02 18:51:29 INFO metastore.HiveMetaStore: 0: get_table : db=default tbl=sales
14/07/02 18:51:29 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=get_table : db=default tbl=sales
14/07/02 18:51:29 INFO metastore.HiveMetaStore: 0: get_table : db=default tbl=sales
14/07/02 18:51:29 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=get_table : db=default tbl=sales
14/07/02 18:51:29 INFO metastore.HiveMetaStore: 0: alter_table: db=default tbl=sales newtbl=sales
14/07/02 18:51:29 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=alter_table: db=default tbl=sales newtbl=sales
14/07/02 18:51:29 INFO metastore.HiveMetaStore: 0: get_table : db=default tbl=sales
14/07/02 18:51:29 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=get_table : db=default tbl=sales
14/07/02 18:51:29 INFO exec.Task: Table default.sales stats: [num_partitions: 0, num_files: 1, num_rows: 0, total_size: 13, raw_data_size: 0]
14/07/02 18:51:29 INFO ql.Driver: </PERFLOG method=task.STATS.Stage-2 start=1404298289196 end=1404298289282 duration=86>
14/07/02 18:51:29 INFO ql.Driver: </PERFLOG method=runTasks start=1404298288943 end=1404298289282 duration=339>
14/07/02 18:51:29 INFO ql.Driver: </PERFLOG method=Driver.execute start=1404298288943 end=1404298289282 duration=339>
14/07/02 18:51:29 INFO ql.Driver: OK
14/07/02 18:51:29 INFO ql.Driver: <PERFLOG method=releaseLocks>
14/07/02 18:51:29 INFO ql.Driver: </PERFLOG method=releaseLocks start=1404298289282 end=1404298289282 duration=0>
14/07/02 18:51:29 INFO ql.Driver: </PERFLOG method=Driver.run start=1404298288628 end=1404298289282 duration=654>
14/07/02 18:51:29 INFO ql.Driver: <PERFLOG method=releaseLocks>
14/07/02 18:51:29 INFO ql.Driver: </PERFLOG method=releaseLocks start=1404298289282 end=1404298289282 duration=0>
14/07/02 18:51:29 INFO analysis.Analyzer: Max iterations (2) reached for batch MultiInstanceRelations
14/07/02 18:51:29 INFO analysis.Analyzer: Max iterations (2) reached for batch CaseInsensitiveAttributeReferences
14/07/02 18:51:29 INFO metastore.HiveMetaStore: 0: get_table : db=default tbl=sales
14/07/02 18:51:29 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=get_table : db=default tbl=sales
14/07/02 18:51:29 INFO storage.MemoryStore: ensureFreeSpace(355913) called with curMem=0, maxMem=1150314086
14/07/02 18:51:29 INFO storage.MemoryStore: Block broadcast_0 stored as values in memory (estimated size 347.6 KB, free 1096.7 MB)
14/07/02 18:51:29 INFO sql.SQLContext$$anon$1: Max iterations (2) reached for batch Add exchange
14/07/02 18:51:29 INFO sql.SQLContext$$anon$1: Max iterations (2) reached for batch Prepare Expressions
querySales: org.apache.spark.sql.SchemaRDD = 
SchemaRDD[0] at RDD at SchemaRDD.scala:100
== Query Plan ==
HiveTableScan [key#2,value#3], (MetastoreRelation default, sales, None), None

执行spark sql

scala> querySales.collect()
14/07/02 18:57:32 WARN snappy.LoadSnappy: Snappy native library is available
14/07/02 18:57:32 WARN snappy.LoadSnappy: Snappy native library not loaded
14/07/02 18:57:32 INFO mapred.FileInputFormat: Total input paths to process : 1
14/07/02 18:57:32 INFO spark.SparkContext: Starting job: collect at SparkPlan.scala:52
14/07/02 18:57:32 INFO scheduler.DAGScheduler: Got job 0 (collect at SparkPlan.scala:52) with 3 output partitions (allowLocal=false)
14/07/02 18:57:32 INFO scheduler.DAGScheduler: Final stage: Stage 0(collect at SparkPlan.scala:52)
14/07/02 18:57:32 INFO scheduler.DAGScheduler: Parents of final stage: List()
14/07/02 18:57:32 INFO scheduler.DAGScheduler: Missing parents: List()
14/07/02 18:57:32 INFO scheduler.DAGScheduler: Submitting Stage 0 (MappedRDD[5] at map at SparkPlan.scala:52), which has no missing parents
14/07/02 18:57:32 INFO scheduler.DAGScheduler: Submitting 3 missing tasks from Stage 0 (MappedRDD[5] at map at SparkPlan.scala:52)
14/07/02 18:57:32 INFO scheduler.TaskSchedulerImpl: Adding task set 0.0 with 3 tasks
14/07/02 18:57:32 INFO scheduler.TaskSetManager: Re-computing pending task lists.
14/07/02 18:57:32 INFO scheduler.TaskSetManager: Starting task 0.0:0 as TID 0 on executor localhost: localhost (PROCESS_LOCAL)
14/07/02 18:57:32 INFO scheduler.TaskSetManager: Serialized task 0.0:0 as 3606 bytes in 20 ms
14/07/02 18:57:32 INFO executor.Executor: Running task ID 0
14/07/02 18:57:32 INFO storage.BlockManager: Found block broadcast_0 locally
14/07/02 18:57:32 INFO rdd.HadoopRDD: Input split: file:/tmp/sparkHiveWarehouse5366068035857129261/sales/sales.txt:0+6
14/07/02 18:57:32 INFO executor.Executor: Serialized size of result for 0 is 1947
14/07/02 18:57:32 INFO executor.Executor: Sending result for 0 directly to driver
14/07/02 18:57:32 INFO executor.Executor: Finished task ID 0
14/07/02 18:57:32 INFO scheduler.TaskSetManager: Starting task 0.0:1 as TID 1 on executor localhost: localhost (PROCESS_LOCAL)
14/07/02 18:57:32 INFO scheduler.TaskSetManager: Serialized task 0.0:1 as 3606 bytes in 0 ms
14/07/02 18:57:32 INFO executor.Executor: Running task ID 1
14/07/02 18:57:32 INFO storage.BlockManager: Found block broadcast_0 locally
14/07/02 18:57:32 INFO scheduler.DAGScheduler: Completed ResultTask(0, 0)
14/07/02 18:57:32 INFO scheduler.TaskSetManager: Finished TID 0 in 243 ms on localhost (progress: 1/3)
14/07/02 18:57:32 INFO rdd.HadoopRDD: Input split: file:/tmp/sparkHiveWarehouse5366068035857129261/sales/sales.txt:6+6
14/07/02 18:57:32 INFO executor.Executor: Serialized size of result for 1 is 1948
14/07/02 18:57:32 INFO executor.Executor: Sending result for 1 directly to driver
14/07/02 18:57:32 INFO executor.Executor: Finished task ID 1
14/07/02 18:57:32 INFO scheduler.TaskSetManager: Starting task 0.0:2 as TID 2 on executor localhost: localhost (PROCESS_LOCAL)
14/07/02 18:57:32 INFO scheduler.TaskSetManager: Serialized task 0.0:2 as 3606 bytes in 1 ms
14/07/02 18:57:32 INFO executor.Executor: Running task ID 2
14/07/02 18:57:32 INFO scheduler.TaskSetManager: Finished TID 1 in 36 ms on localhost (progress: 2/3)
14/07/02 18:57:32 INFO scheduler.DAGScheduler: Completed ResultTask(0, 1)
14/07/02 18:57:32 INFO storage.BlockManager: Found block broadcast_0 locally
14/07/02 18:57:32 INFO rdd.HadoopRDD: Input split: file:/tmp/sparkHiveWarehouse5366068035857129261/sales/sales.txt:12+1
14/07/02 18:57:32 INFO executor.Executor: Serialized size of result for 2 is 1721
14/07/02 18:57:32 INFO executor.Executor: Sending result for 2 directly to driver
14/07/02 18:57:32 INFO executor.Executor: Finished task ID 2
14/07/02 18:57:32 INFO scheduler.TaskSetManager: Finished TID 2 in 96 ms on localhost (progress: 3/3)
14/07/02 18:57:32 INFO scheduler.DAGScheduler: Completed ResultTask(0, 2)
14/07/02 18:57:32 INFO scheduler.TaskSchedulerImpl: Removed TaskSet 0.0, whose tasks have all completed, from pool 
14/07/02 18:57:32 INFO scheduler.DAGScheduler: Stage 0 (collect at SparkPlan.scala:52) finished in 0.366 s
14/07/02 18:57:32 INFO spark.SparkContext: Job finished: collect at SparkPlan.scala:52, took 0.454512333 s
res1: Array[org.apache.spark.sql.Row] = Array([Joe,2], [Hank,2])

执行结果：

Array([Joe,2], [Hank,2])

查询计划优化：

<pre name="code" class="python">scala> val query = sql("SELECT * FROM (SELECT * FROM src) a")

query: org.apache.spark.sql.SchemaRDD = 
SchemaRDD[6] at RDD at SchemaRDD.scala:100
== Query Plan ==

<span style="font-family: Arial, Helvetica, sans-serif;">HiveTableScan [key#6,value#7], (MetastoreRelation default, src, None), None</span>

3、Spark SQL LINQ

在Spark相关的框架里，一切的核心体都是RDD，SchemaRDD提供类似LINQ的语法api：

such as take, where... etc

scala> query.
++                         aggregate                  as                         asInstanceOf               baseLogicalPlan            baseSchemaRDD              
cache                      cartesian                  checkpoint                 coalesce                   collect                    compute                    
context                    count                      countApprox                countApproxDistinct        countByValue               countByValueApprox         
dependencies               distinct                   filter                     filterWith                 first                      flatMap                    
flatMapWith                fold                       foreach                    foreachPartition           foreachWith                generate                   
getCheckpointFile          getPartitions              getStorageLevel            glom                       groupBy                    id                         
insertInto                 intersection               isCheckpointed             isInstanceOf               iterator                   join                       
keyBy                      limit                      map                        mapPartitions              mapPartitionsWithContext   mapPartitionsWithIndex     
mapPartitionsWithSplit     mapWith                    max                        min                        name                       name_=                     
orderBy                    partitioner                partitions                 persist                    pipe                       preferredLocations         
printSchema                queryExecution             randomSplit                reduce                     registerAsTable            repartition                
sample                     saveAsObjectFile           saveAsParquetFile          saveAsTable                saveAsTextFile             schemaString               
select                     setName                    sortBy                     sparkContext               sqlContext                 subtract                   
take                       takeOrdered                takeSample                 toArray                    toDebugString              toJavaRDD                  
toJavaSchemaRDD            toLocalIterator            toSchemaRDD                toString                   top                        union                      
unionAll                   unpersist                  where                      zip                        zipPartitions              zipWithIndex               
zipWithUniqueId

注意key前面带了一撇，这个是Catalyst的查下语法，以后我会写一篇详细介绍：

scala> query.where('key === 100).collect()
14/07/02 19:07:55 INFO analysis.Analyzer: Max iterations (2) reached for batch MultiInstanceRelations
14/07/02 19:07:55 INFO analysis.Analyzer: Max iterations (2) reached for batch CaseInsensitiveAttributeReferences
14/07/02 19:07:55 INFO metastore.HiveMetaStore: 0: get_table : db=default tbl=src
14/07/02 19:07:55 INFO HiveMetaStore.audit: ugi=hadoop  ip=unknown-ip-addr      cmd=get_table : db=default tbl=src
14/07/02 19:07:55 INFO storage.MemoryStore: ensureFreeSpace(358003) called with curMem=713876, maxMem=1150314086
14/07/02 19:07:55 INFO storage.MemoryStore: Block broadcast_3 stored as values in memory (estimated size 349.6 KB, free 1096.0 MB)
14/07/02 19:07:55 INFO sql.SQLContext$$anon$1: Max iterations (2) reached for batch Add exchange
14/07/02 19:07:55 INFO sql.SQLContext$$anon$1: Max iterations (2) reached for batch Prepare Expressions
14/07/02 19:07:55 INFO mapred.FileInputFormat: Total input paths to process : 1
14/07/02 19:07:55 INFO spark.SparkContext: Starting job: collect at SparkPlan.scala:52
14/07/02 19:07:55 INFO scheduler.DAGScheduler: Got job 2 (collect at SparkPlan.scala:52) with 2 output partitions (allowLocal=false)
14/07/02 19:07:55 INFO scheduler.DAGScheduler: Final stage: Stage 2(collect at SparkPlan.scala:52)
14/07/02 19:07:55 INFO scheduler.DAGScheduler: Parents of final stage: List()
14/07/02 19:07:55 INFO scheduler.DAGScheduler: Missing parents: List()
14/07/02 19:07:55 INFO scheduler.DAGScheduler: Submitting Stage 2 (MappedRDD[20] at map at SparkPlan.scala:52), which has no missing parents
14/07/02 19:07:55 INFO scheduler.DAGScheduler: Submitting 2 missing tasks from Stage 2 (MappedRDD[20] at map at SparkPlan.scala:52)
14/07/02 19:07:55 INFO scheduler.TaskSchedulerImpl: Adding task set 2.0 with 2 tasks
14/07/02 19:07:55 INFO scheduler.TaskSetManager: Starting task 2.0:0 as TID 5 on executor localhost: localhost (PROCESS_LOCAL)
14/07/02 19:07:55 INFO scheduler.TaskSetManager: Serialized task 2.0:0 as 3854 bytes in 0 ms
14/07/02 19:07:55 INFO executor.Executor: Running task ID 5
14/07/02 19:07:55 INFO storage.BlockManager: Found block broadcast_3 locally
14/07/02 19:07:55 INFO rdd.HadoopRDD: Input split: file:/tmp/sparkHiveWarehouse5366068035857129261/src/kv1.txt:0+2906
14/07/02 19:07:55 INFO executor.Executor: Serialized size of result for 5 is 1951
14/07/02 19:07:55 INFO executor.Executor: Sending result for 5 directly to driver
14/07/02 19:07:55 INFO scheduler.TaskSetManager: Starting task 2.0:1 as TID 6 on executor localhost: localhost (PROCESS_LOCAL)
14/07/02 19:07:55 INFO scheduler.TaskSetManager: Serialized task 2.0:1 as 3854 bytes in 0 ms
14/07/02 19:07:55 INFO executor.Executor: Finished task ID 5
14/07/02 19:07:55 INFO executor.Executor: Running task ID 6
14/07/02 19:07:55 INFO scheduler.TaskSetManager: Finished TID 5 in 44 ms on localhost (progress: 1/2)
14/07/02 19:07:55 INFO scheduler.DAGScheduler: Completed ResultTask(2, 0)
14/07/02 19:07:55 INFO storage.BlockManager: Found block broadcast_3 locally
14/07/02 19:07:55 INFO rdd.HadoopRDD: Input split: file:/tmp/sparkHiveWarehouse5366068035857129261/src/kv1.txt:2906+2906
14/07/02 19:07:55 INFO executor.Executor: Serialized size of result for 6 is 1951
14/07/02 19:07:55 INFO executor.Executor: Sending result for 6 directly to driver
14/07/02 19:07:55 INFO executor.Executor: Finished task ID 6
14/07/02 19:07:55 INFO scheduler.TaskSetManager: Finished TID 6 in 19 ms on localhost (progress: 2/2)
14/07/02 19:07:55 INFO scheduler.DAGScheduler: Completed ResultTask(2, 1)
14/07/02 19:07:55 INFO scheduler.TaskSchedulerImpl: Removed TaskSet 2.0, whose tasks have all completed, from pool 
14/07/02 19:07:55 INFO scheduler.DAGScheduler: Stage 2 (collect at SparkPlan.scala:52) finished in 0.062 s
14/07/02 19:07:55 INFO spark.SparkContext: Job finished: collect at SparkPlan.scala:52, took 0.06947625 s
res6: Array[org.apache.spark.sql.Row] = Array([100,val_100], [100,val_100])

查询出2个key为100的结果。

4、总结：

Spark SQL 提供了一种Catalyst查询优化框架，在把SQL解析成逻辑执行计划，对执行计划优化，最后变成RDD操作，多种框架一种API，简单，规范。

本文暂且为止，后续还会继续相关的深入研究。

——EOF——

原创文章，转载请注明出自http://blog.csdn.net/oopsoom/article/details/36440821

声明：以上内容来自用户投稿及互联网公开渠道收集整理发布，本网站不拥有所有权，未作人工编辑处理，也不承担相关法律责任，若内容有误或涉及侵权可进行投诉：投诉/举报工作人员会在5个工作日内联系你，一经查实，本站将立刻删除涉嫌侵权内容。

联系
我们