hive个人学习总结---【基于微博数据的HQL练习】

练习：

1. 统计用户所有微博被转发的总次数，并输出TOP-3 用户

hive> select t.userId,count(*) count  from 
(select get_json_object(w.json,'$.beForwardWeiboId') beForwardWeiboId,get_json_object(w.json,'$.userId') userId from 
(select substring(json,2,length(json)-2) as json from weibo) w) t
 where t.beForwardWeiboId !='' group by t.userId order by `count` desc limit 3;

说明：

• get_json_object ：函数可将json解析成对象，get_json_object函数第一个参数填写json对象变量，第二个参数使用$表示json变量标识，然后用 . 或 [] 读取对象或数组；

• substring(json,2,length(json)-2) as json *** ：是将前两个字符和最后两个字符*去掉。数据格式是：

[{"beCommentWeiboId":"","beForwardWeiboId":"","catchTime":"1387159495","commentCount":"1419","content":"分享图片","createTime":"1386981067","info1":"","info2":"","info3":"","mlevel":"","musicurl":[],"pic_list":["http://ww3.sinaimg.cn/thumbnail/40d61044jw1ebixhnsiknj20qo0qognx.jpg"],"praiseCount":"5265","reportCount":"1285","source":"iPad客户端","userId":"1087770692","videourl":[],"weiboId":"3655325888057474","weiboUrl":"http://weibo.com/1087770692/AndhixO7g"}]

结果：

结果

---

2. 统计每个用户的发送微博总数，并存储到临时表

• 先建立临时表user_count

CREATE TABLE user_count (
userId STRING,
`count` INT
);

• 插入数据

insert into table user_count
select userId,count(distinct(weiboId)) `count` from 
(select get_json_object(w.json,'$.userId') userId,get_json_object(w.json,'$.weiboId') weiboId from
 (select subString(json,2,length(json)-2) as json from weibo) w) w1 group by userId order by `count` desc 

• 去临时表user_count查询前5条，查看是否插入成功

结果：

结果

---

通过自定义udf来执行HQL

Q：将微博的点赞人数与转发人数相加求和，并将相加之和降序排列，取前 10

• 自定义udf函数代码

package udf;

import org.apache.hadoop.hive.ql.exec.UDF;

public final class addCount extends UDF {

    public long evaluate(final Double column1, final Double column2) {
        if (column1 != null && column2 != null) {
            Double result = column1 + column2;
            return result.longValue();  
        }
        if(column1 == null && column2 != null){
            return column2.longValue();
        }
        if(column2 == null && column1 != null){
            return column1.longValue();
        } else {
            return 0;
        }
    }
    
}

• 生成jar文件：

** 1.**

eclipse下将udf函数打包

** 2.**

在linux中找到刚刚打好的jar包

• 执行hql语句

引入udf的jar：

hive> add jar /home/xc2046/workspace/jar/addCount001.jar;
Added /home/xc2046/workspace/jar/addCount001.jar to class path
Added resource: /home/xc2046/workspace/jar/addCount001.jar

创建临时udf函数：

hive> create temporary function addcount as 'udf.addCount';
OK
Time taken: 0.048 seconds

执行带有自定义udf的查询hql：

hive> SELECT weiboId,addcount(SUM(praiseCount),SUM(reportCount)) `count` FROM(
SELECT 
get_json_object(w.json,'$.weiboId') weiboId,
get_json_object(w.json,'$.praiseCount') praiseCount,
get_json_object(w.json,'$.reportCount') reportCount 
FROM (SELECT SUBSTRING(json,2,LENGTH(json)-2) AS json FROM weibo) w
)w1 GROUP BY weiboId ORDER BY `count` DESC LIMIT 10;

• 最终结果：

Paste_Image.png