【Spark】Spark五： Spark RDD API初步第三部分

本文对Sogou的日志进行分析，Sogou日志下载地址.

http://download.labs.sogou.com/dl/sogoulabdown/SogouQ/SogouQ2012.mini.tar.gz

 

Sogou日志说明：

解压查看Sougou日志的内容和格式，

 

文件格式：访问时间\t加密的用户ID\t搜索关键词\t该URL的搜索结果排名\t该URL用户点击排序结果\t用户点击的URL

日志上传至HDFS

hdfs dfs -put /home/hadoop/software/spark-1.2.0-bin-hadoop2.4/data/sogou/SogouQ.mini /user/hadoop/

每行日志格式正确性检查

scala> val rdd = sc.textFile("SogouQ.mini");
scala> rdd.count
///结果
2000
scala> val validRDD = rdd.map(_.split("\t")).filter(_.length == 6)
///结果
2000 

可见，文件的总行数为2000，同时，对2000文本进行分割过滤长度为6的总行数也是2000，即每行日志的格式都是是正确的

统计搜索结果和点击结果都是第一的记录

1. 计算结果

scala> val result = validRDD.filter(_(3).toInt == 1).filter(_(4).toInt == 1)
scala> result.count
///结果
res9: Long = 794

///注意，这里如果使用下面的语句则得不到正确的结果，结果为0
scala> val result = rdd.filter(_(3).toInt == 1).filter(_(4).toInt == 1)
scala> result.count
res8: Long = 0

2. 查看RDD转换

scala> result.toDebugString
res10: String = 
(1) FilteredRDD[32] at filter at <console>:16 []
 |  FilteredRDD[31] at filter at <console>:16 []
 |  FilteredRDD[28] at filter at <console>:14 []
 |  MappedRDD[27] at map at <console>:14 []
 |  SogouQ.mini MappedRDD[26] at textFile at <console>:12 []
 |  SogouQ.mini HadoopRDD[25] at textFile at <console>:12 []

 

对用户ID的查询次数进行排序

1.

scala> val result = validRDD.map(x => (x(1),1)).reduceByKey(_+_).map(x => (x._2,x._1)).sortByKey(false).map(x => (x._2,x._1))
scala> result.count
///结果
res11: Long = 1186

///保存到HDFS中
scala> result.saveAsTextFile("SortedResultByUserID"); 

 2. 查看HDFS上的排序结果

[hadoop@hadoop sogou]$ hdfs dfs -cat /user/hadoop/SortedResultByUserID/part-00000| head -20
(f6492a1da9875f20e01ff8b5804dcc35,14)
(d3034ac9911c30d7cf9312591ecf990e,11)
(e7579c6b6b9c0ea40ecfa0f425fc765a,11)
(ec0363079f36254b12a5e30bdc070125,10)
(5c853e91940c5eade7455e4a289722d6,10)
(2a36742c996300d664652d9092e8a554,9)
(439fa809ba818cee624cc8b6e883913a,9)
(828f91e6717213a65c97b694e6279201,9)
(45c304b5f2dd99182451a02685252312,8)
(5ea391fd07dbb616e9857a7d95f460e0,8)
(596444b8c02b7b30c11273d5bbb88741,8)
(a06830724b809c0db56263124b2bd142,8)
(41389fb54f9b3bec766c5006d7bce6a2,7)
(6056710d9eafa569ddc800fe24643051,7)
(8897bbb7bdff69e80f7fb2041d83b17d,7)
(bc8cc0577bb80fafd6fad1ed67d3698e,7)
(29ede0f2544d28b714810965400ab912,6)
(b6d439f7b71172e8416a28ea356c06ef,6)
(6da1dcbaeab299deffe5932d902e775d,6)
(091a8d6c26aa8e95185383f45624227c,6)


[hadoop@hadoop sogou]$ hdfs dfs -cat /user/hadoop/SortedResultByUserID/part-00000| tail -10
(9e76272ecf1e526834eb7c9609f03fc0,1)
(57375476989eea12893c0c3811607bcf,1)
(7de315ffe0730e2275ae6146854237cc,1)
(ede950da3337974f2b49bf0c97ca5664,1)
(3d29715c7ef674b1223b2814efb87d20,1)
(e0259578a1754f9f2221f4367bbc8c84,1)
(bb4023be70aef58d29aac9102284705f,1)
(1e0b9bab1eaf1255f11140c5f6fc816d,1)
(14a23e70a246a8ffabfd6f97711aedc2,1)
(c491b1f1d01ec698a69123d4fdff44cc,1)

 结果正确

参考：Spark实战高手之路系列文章