10、Hive数据仓库——函数
作者：互联网
Hive数据仓库——函数
文章目录
- Hive数据仓库——函数
- Hive 常用函数
- Hive 中的wordCount
- Hive 开窗函数
- 测试数据
- 建表语句
- row_number：无并列排名
- dense_rank：有并列排名，并且依次递增
- rank：有并列排名，不依次递增
- percent_rank：(rank的结果-1)/(分区内数据的个数-1)
- cume_dist：计算某个窗口或分区中某个值的累积分布。
- NTILE(n)：对分区内数据再分成n组，然后打上组号
- max、min、avg、count、sum：基于每个partition分区内的数据做对应的计算
- 窗口帧：用于从分区中选择指定的多条记录，供窗口函数处理
- LAG(col,n)：往前第n行数据
- LEAD(col,n)：往后第n行数据
- FIRST_VALUE：取分组内排序后，截止到当前行，第一个值
- LAST_VALUE：取分组内排序后，截止到当前行，最后一个值，对于并列的排名，取最后一个
- Hive 行转列
- Hive 列转行
- Hive自定义函数UserDefineFunction
- Hive Shell
- 连续登陆问题
Hive 常用函数
关系运算
// 等值比较 = == <=>
// 不等值比较 != <>
// 区间比较： select * from default.students where id between 1500100001 and 1500100010;
// 空值/非空值判断：is null、is not null、nvl()、isnull()
// like、rlike、regexp用法
数值计算
取整函数(四舍五入)：round
向上取整：ceil
向下取整：floor
条件函数
- if： if(表达式,如果表达式成立的返回值,如果表达式不成立的返回值)
select if(1>0,1,0);
select if(1>0,if(-1>0,-1,1),0);
select score,if(score>120,'优秀',if(score>100,'良好',if(score>90,'及格','不及格'))) as pingfen from score limit 20;
- COALESCE
select COALESCE(null,'1','2'); // 1 从左往右 一次匹配 直到非空为止
select COALESCE('1',null,'2'); // 1
- case when
select score
,case when score>120 then '优秀'
when score>100 then '良好'
when score>90 then '及格'
else '不及格'
end as pingfen
from default.score limit 20;
select name
,case name when "施笑槐" then "槐ge"
when "吕金鹏" then "鹏ge"
when "单乐蕊" then "蕊jie"
else "算了不叫了"
end as nickname
from default.students limit 10;
注意条件的顺序
日期函数
select from_unixtime(1610611142,'YYYY/MM/dd HH:mm:ss');
select from_unixtime(unix_timestamp(),'YYYY/MM/dd HH:mm:ss');
// '2021年01月14日' -> '2021-01-14'
select from_unixtime(unix_timestamp('2021年01月14日','yyyy年MM月dd日'),'yyyy-MM-dd');
// "04牛2021数加16逼" -> "2021/04/16"
select from_unixtime(unix_timestamp("04牛2021数加16逼","MM牛yyyy数加dd逼"),"yyyy/MM/dd");
字符串函数
concat('123','456'); // 123456
concat('123','456',null); // NULL
select concat_ws('#','a','b','c'); // a#b#c
select concat_ws('#','a','b','c',NULL); // a#b#c 可以指定分隔符，并且会自动忽略NULL
select concat_ws("|",cast(id as string),name,cast(age as string),gender,clazz) from students limit 10;
select substring("abcdefg",1); // abcdefg HQL中涉及到位置的时候 是从1开始计数
// '2021/01/14' -> '2021-01-14'
select concat_ws("-",substring('2021/01/14',1,4),substring('2021/01/14',6,2),substring('2021/01/14',9,2));
// 建议使用日期函数去做日期
select from_unixtime(unix_timestamp('2021/01/14','yyyy/MM/dd'),'yyyy-MM-dd');
select split("abcde,fgh",","); // ["abcde","fgh"]
select split("a,b,c,d,e,f",",")[2]; // c
select explode(split("abcde,fgh",",")); // abcde
// fgh
// 解析json格式的数据
select get_json_object('{"name":"zhangsan","age":18,"score":[{"course_name":"math","score":100},{"course_name":"english","score":60}]}',"$.score[0].score"); // 100
Hive 中的wordCount
create table words(
words string
)row format delimited fields terminated by '|';
// 数据
hello,java,hello,java,scala,python
hbase,hadoop,hadoop,hdfs,hive,hive
hbase,hadoop,hadoop,hdfs,hive,hive
select word,count(*) from (select explode(split(words,',')) word from words) a group by a.word;
// 结果
hadoop 4
hbase 2
hdfs 2
hello 2
hive 4
java 2
python 1
scala 1
Hive 开窗函数
好像给每一份数据 开一扇窗户 所以叫开窗函数
在sql中有一类函数叫做聚合函数,例如sum()、avg()、max()等等,这类函数可以将多行数据按照规则聚集为一行,一般来讲聚集后的行数是要少于聚集前的行数的.但是有时我们想要既显示聚集前的数据,又要显示聚集后的数据,这时我们便引入了窗口函数.
测试数据
111,69,class1,department1
112,80,class1,department1
113,74,class1,department1
114,94,class1,department1
115,93,class1,department1
121,74,class2,department1
122,86,class2,department1
123,78,class2,department1
124,70,class2,department1
211,93,class1,department2
212,83,class1,department2
213,94,class1,department2
214,94,class1,department2
215,82,class1,department2
216,74,class1,department2
221,99,class2,department2
222,78,class2,department2
223,74,class2,department2
224,80,class2,department2
225,85,class2,department2
建表语句
create table new_score(
id int
,score int
,clazz string
,department string
) row format delimited fields terminated by ",";
row_number：无并列排名
- 用法： select xxxx, row_number() over(partition by 分组字段 order by 排序字段 desc) as rn from tb group by xxxx
dense_rank：有并列排名，并且依次递增
rank：有并列排名，不依次递增
percent_rank：(rank的结果-1)/(分区内数据的个数-1)
cume_dist：计算某个窗口或分区中某个值的累积分布。
假定升序排序，则使用以下公式确定累积分布： 小于等于当前值x的行数 / 窗口或partition分区内的总行数。其中，x 等于 order by 子句中指定的列的当前行中的值。
NTILE(n)：对分区内数据再分成n组，然后打上组号
max、min、avg、count、sum：基于每个partition分区内的数据做对应的计算
窗口帧：用于从分区中选择指定的多条记录，供窗口函数处理
Hive 提供了两种定义窗口帧的形式：
ROWS和RANGE。两种类型都需要配置上界和下界。例如，ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW表示选择分区起始记录到当前记录的所有行；SUM(close) RANGE BETWEEN 100 PRECEDING AND 200 FOLLOWING则通过 字段差值 来进行选择。如当前行的close字段值是200，那么这个窗口帧的定义就会选择分区中close字段值落在100至400区间的记录。以下是所有可能的窗口帧定义组合。如果没有定义窗口帧，则默认为RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW。只能运用在max、min、avg、count、sum、FIRST_VALUE、LAST_VALUE这几个窗口函数上
(ROWS | RANGE) BETWEEN (UNBOUNDED | [num]) PRECEDING AND ([num] PRECEDING | CURRENT ROW | (UNBOUNDED | [num]) FOLLOWING) (ROWS | RANGE) BETWEEN CURRENT ROW AND (CURRENT ROW | (UNBOUNDED | [num]) FOLLOWING) (ROWS | RANGE) BETWEEN [num] FOLLOWING AND (UNBOUNDED | [num]) FOLLOWING range between 3 PRECEDING and 11 FOLLOWINGSELECT id ,score ,clazz ,SUM(score) OVER w as sum_w ,round(avg(score) OVER w,3) as avg_w ,count(score) OVER w as cnt_w FROM new_score WINDOW w AS (PARTITION BY clazz ORDER BY score rows between 2 PRECEDING and 2 FOLLOWING);111 69 class1 217 72.333 3 113 74 class1 297 74.25 4 216 74 class1 379 75.8 5 112 80 class1 393 78.6 5 215 82 class1 412 82.4 5 212 83 class1 431 86.2 5 211 93 class1 445 89.0 5 115 93 class1 457 91.4 5 213 94 class1 468 93.6 5 114 94 class1 375 93.75 4 214 94 class1 282 94.0 3 124 70 class2 218 72.667 3 121 74 class2 296 74.0 4 223 74 class2 374 74.8 5 222 78 class2 384 76.8 5 123 78 class2 395 79.0 5 224 80 class2 407 81.4 5 225 85 class2 428 85.6 5 122 86 class2 350 87.5 4 221 99 class2 270 90.0 3
select id
,score
,clazz
,department
,row_number() over (partition by clazz order by score desc) as rn_rk
,dense_rank() over (partition by clazz order by score desc) as dense_rk
,rank() over (partition by clazz order by score desc) as rk
,percent_rank() over (partition by clazz order by score desc) as percent_rk
,round(cume_dist() over (partition by clazz order by score desc),3) as cume_rk
,NTILE(3) over (partition by clazz order by score desc) as ntile_num
,max(score) over (partition by clazz order by score desc range between 3 PRECEDING and 11 FOLLOWING) as max_p
from new_score;
id score clazz department rn_rk ds_rk rk percent_rk cume_rk ntile_num max_p
114 94 class1 department1 1 1 1 0.0 0.273 1 94
214 94 class1 department2 2 1 1 0.0 0.273 1 94
213 94 class1 department2 3 1 1 0.0 0.273 1 94
211 93 class1 department2 4 2 4 0.3 0.455 1 94
115 93 class1 department1 5 2 4 0.3 0.455 2 94
212 83 class1 department2 6 3 6 0.5 0.545 2 83
215 82 class1 department2 7 4 7 0.6 0.636 2 83
112 80 class1 department1 8 5 8 0.7 0.727 2 83
113 74 class1 department1 9 6 9 0.8 0.909 3 74
216 74 class1 department2 10 6 9 0.8 0.909 3 74
111 69 class1 department1 11 7 11 1.0 1.0 3 69
221 99 class2 department2 1 1 1 0.0 0.111 1 99
122 86 class2 department1 2 2 2 0.125 0.222 1 86
225 85 class2 department2 3 3 3 0.25 0.333 1 86
224 80 class2 department2 4 4 4 0.375 0.444 2 80
123 78 class2 department1 5 5 5 0.5 0.667 2 80
222 78 class2 department2 6 5 5 0.5 0.667 2 80
121 74 class2 department1 7 6 7 0.75 0.889 3 74
223 74 class2 department2 8 6 7 0.75 0.889 3 74
124 70 class2 department1 9 7 9 1.0 1.0 3 70
LAG(col,n)：往前第n行数据
LEAD(col,n)：往后第n行数据
FIRST_VALUE：取分组内排序后，截止到当前行，第一个值
LAST_VALUE：取分组内排序后，截止到当前行，最后一个值，对于并列的排名，取最后一个
select id
,score
,clazz
,department
,lag(id,2) over (partition by clazz order by score desc) as lag_num
,LEAD(id,2) over (partition by clazz order by score desc) as lead_num
,FIRST_VALUE(id) over (partition by clazz order by score desc) as first_v_num
,LAST_VALUE(id) over (partition by clazz order by score desc) as last_v_num
,NTILE(3) over (partition by clazz order by score desc) as ntile_num
from new_score;
id score clazz department lag_num lead_num first_v_num last_v_num ntile_num
114 94 class1 department1 NULL 213 114 213 1
214 94 class1 department2 NULL 211 114 213 1
213 94 class1 department2 114 115 114 213 1
211 93 class1 department2 214 212 114 115 1
115 93 class1 department1 213 215 114 115 2
212 83 class1 department2 211 112 114 212 2
215 82 class1 department2 115 113 114 215 2
112 80 class1 department1 212 216 114 112 2
113 74 class1 department1 215 111 114 216 3
216 74 class1 department2 112 NULL 114 216 3
111 69 class1 department1 113 NULL 114 111 3
221 99 class2 department2 NULL 225 221 221 1
122 86 class2 department1 NULL 224 221 122 1
225 85 class2 department2 221 123 221 225 1
224 80 class2 department2 122 222 221 224 2
123 78 class2 department1 225 121 221 222 2
222 78 class2 department2 224 223 221 222 2
121 74 class2 department1 123 124 221 223 3
223 74 class2 department2 222 NULL 221 223 3
124 70 class2 department1 121 NULL 221 124 3
https://blog.csdn.net/qq_26937525/article/details/54925827
Hive 行转列
lateral view explode
create table testArray2(
name string,
weight array<string>
)row format delimited
fields terminated by '\t'
COLLECTION ITEMS terminated by ',';
志凯 "150","170","180"
上单 "150","180","190"
select name,col1 from testarray2 lateral view explode(weight) t1 as col1;
志凯 150
志凯 170
志凯 180
上单 150
上单 180
上单 190
select key from (select explode(map('key1',1,'key2',2,'key3',3)) as (key,value)) t;
key1
key2
key3
select name,col1,col2 from testarray2 lateral view explode(map('key1',1,'key2',2,'key3',3)) t1 as col1,col2;
志凯 key1 1
志凯 key2 2
志凯 key3 3
上单 key1 1
上单 key2 2
上单 key3 3
select name,pos,col1 from testarray2 lateral view posexplode(weight) t1 as pos,col1;
志凯 0 150
志凯 1 170
志凯 2 180
上单 0 150
上单 1 180
上单 2 190
Hive 列转行
// testLieToLine
name col1
志凯 150
志凯 170
志凯 180
上单 150
上单 180
上单 190
create table testLieToLine(
name string,
col1 int
)row format delimited
fields terminated by '\t';
select name,collect_list(col1) from testLieToLine group by name;
// 结果
上单 ["150","180","190"]
志凯 ["150","170","180"]
select t1.name
,collect_list(t1.col1)
from (
select name
,col1
from testarray2
lateral view explode(weight) t1 as col1
) t1 group by t1.name;
Hive自定义函数UserDefineFunction
UDF：一进一出
- 创建maven项目，并加入依赖
<dependency>
<groupId>org.apache.hive</groupId>
<artifactId>hive-exec</artifactId>
<version>1.2.1</version>
</dependency>
- 编写代码，继承org.apache.hadoop.hive.ql.exec.UDF，实现evaluate方法，在evaluate方法中实现自己的逻辑
import org.apache.hadoop.hive.ql.exec.UDF;
public class HiveUDF extends UDF {
// hadoop => #hadoop$
public String evaluate(String col1) {
// 给传进来的数据 左边加上 # 号 右边加上 $
String result = "#" + col1 + "$";
return result;
}
}
- 打成jar包并上传至Linux虚拟机
- 在hive shell中，使用
add jar 路径将jar包作为资源添加到hive环境中
add jar /usr/local/soft/jars/HiveUDF2-1.0.jar;
- 使用jar包资源注册一个临时函数，fxxx1是你的函数名，'MyUDF’是主类名
create temporary function fxxx1 as 'MyUDF';
- 使用函数名处理数据
select fxx1(name) as fxx_name from students limit 10;
#施笑槐$
#吕金鹏$
#单乐蕊$
#葛德曜$
#宣谷芹$
#边昂雄$
#尚孤风$
#符半双$
#沈德昌$
#羿彦昌$
UDTF：一进多出
“key1:value1,key2:value2,key3:value3”
key1 value1
key2 value2
key3 value3
方法一：使用 explode+split
select split(t.col1,":")[0],split(t.col1,":")[1]
from (select explode(split("key1:value1,key2:value2,key3:value3",",")) as col1) t;
方法二：自定UDTF
- 代码
import org.apache.hadoop.hive.ql.exec.UDFArgumentException;
import org.apache.hadoop.hive.ql.metadata.HiveException;
import org.apache.hadoop.hive.ql.udf.generic.GenericUDTF;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorFactory;
import org.apache.hadoop.hive.serde2.objectinspector.StructObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.primitive.PrimitiveObjectInspectorFactory;
import java.util.ArrayList;
public class HiveUDTF extends GenericUDTF {
// 指定输出的列名 及 类型
@Override
public StructObjectInspector initialize(StructObjectInspector argOIs) throws UDFArgumentException {
ArrayList<String> filedNames = new ArrayList<String>();
ArrayList<ObjectInspector> filedObj = new ArrayList<ObjectInspector>();
filedNames.add("col1");
filedObj.add(PrimitiveObjectInspectorFactory.javaStringObjectInspector);
filedNames.add("col2");
filedObj.add(PrimitiveObjectInspectorFactory.javaStringObjectInspector);
return ObjectInspectorFactory.getStandardStructObjectInspector(filedNames, filedObj);
}
// 处理逻辑 my_udtf(col1,col2,col3)
// "key1:value1,key2:value2,key3:value3"
// my_udtf("key1:value1,key2:value2,key3:value3")
public void process(Object[] objects) throws HiveException {
// objects 表示传入的N列
String col = objects[0].toString();
// key1:value1 key2:value2 key3:value3
String[] splits = col.split(",");
for (String str : splits) {
String[] cols = str.split(":");
// 将数据输出
forward(cols);
}
}
// 在UDTF结束时调用
public void close() throws HiveException {
}
}
- SQL
select my_udtf("key1:value1,key2:value2,key3:value3");
字段：id,col1,col2,col3,col4,col5,col6,col7,col8,col9,col10,col11,col12 共13列
数据：
a,1,2,3,4,5,6,7,8,9,10,11,12
b,11,12,13,14,15,16,17,18,19,20,21,22
c,21,22,23,24,25,26,27,28,29,30,31,32
转成3列：id,hours,value
例如：
a,1,2,3,4,5,6,7,8,9,10,11,12
a,0时,1
a,2时,2
a,4时,3
a,6时,4
…
create table udtfData(
id string
,col1 string
,col2 string
,col3 string
,col4 string
,col5 string
,col6 string
,col7 string
,col8 string
,col9 string
,col10 string
,col11 string
,col12 string
)row format delimited fields terminated by ',';
代码：
import org.apache.hadoop.hive.ql.exec.UDFArgumentException;
import org.apache.hadoop.hive.ql.metadata.HiveException;
import org.apache.hadoop.hive.ql.udf.generic.GenericUDTF;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorFactory;
import org.apache.hadoop.hive.serde2.objectinspector.StructObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.primitive.PrimitiveObjectInspectorFactory;
import java.util.ArrayList;
public class HiveUDTF2 extends GenericUDTF {
@Override
public StructObjectInspector initialize(StructObjectInspector argOIs) throws UDFArgumentException {
ArrayList<String> filedNames = new ArrayList<String>();
ArrayList<ObjectInspector> fieldObj = new ArrayList<ObjectInspector>();
filedNames.add("col1");
fieldObj.add(PrimitiveObjectInspectorFactory.javaStringObjectInspector);
filedNames.add("col2");
fieldObj.add(PrimitiveObjectInspectorFactory.javaStringObjectInspector);
return ObjectInspectorFactory.getStandardStructObjectInspector(filedNames, fieldObj);
}
public void process(Object[] objects) throws HiveException {
int hours = 0;
for (Object obj : objects) {
hours = hours + 1;
String col = obj.toString();
ArrayList<String> cols = new ArrayList<String>();
cols.add(hours + "时");
cols.add(col);
forward(cols);
}
}
public void close() throws HiveException {
}
}
添加jar资源:
add jar /usr/local/soft/HiveUDF2-1.0.jar;
注册udtf函数：
create temporary function my_udtf as 'MyUDTF';
SQL:
select id,hours,value from udtfData lateral view my_udtf(col1,col2,col3,col4,col5,col6,col7,col8,col9,col10,col11,col12) t as hours,value ;
UDAF：多进一出
Hive Shell
第一种：
hive -e "select * from test1.students limit 10"
第二种：
hive -f hql文件路径
将HQL写在一个文件里，再使用 -f 参数指定该文件
连续登陆问题
在电商、物流和银行可能经常会遇到这样的需求：统计用户连续交易的总额、连续登陆天数、连续登陆开始和结束时间、间隔天数等
数据：
注意：每个用户每天可能会有多条记录
id datestr amount
1,2019-02-08,6214.23
1,2019-02-08,6247.32
1,2019-02-09,85.63
1,2019-02-09,967.36
1,2019-02-10,85.69
1,2019-02-12,769.85
1,2019-02-13,943.86
1,2019-02-14,538.42
1,2019-02-15,369.76
1,2019-02-16,369.76
1,2019-02-18,795.15
1,2019-02-19,715.65
1,2019-02-21,537.71
2,2019-02-08,6214.23
2,2019-02-08,6247.32
2,2019-02-09,85.63
2,2019-02-09,967.36
2,2019-02-10,85.69
2,2019-02-12,769.85
2,2019-02-13,943.86
2,2019-02-14,943.18
2,2019-02-15,369.76
2,2019-02-18,795.15
2,2019-02-19,715.65
2,2019-02-21,537.71
3,2019-02-08,6214.23
3,2019-02-08,6247.32
3,2019-02-09,85.63
3,2019-02-09,967.36
3,2019-02-10,85.69
3,2019-02-12,769.85
3,2019-02-13,943.86
3,2019-02-14,276.81
3,2019-02-15,369.76
3,2019-02-16,369.76
3,2019-02-18,795.15
3,2019-02-19,715.65
3,2019-02-21,537.71
建表语句
create table deal_tb(
id string
,datestr string
,amount string
)row format delimited fields terminated by ',';
计算逻辑
- 先按用户和日期分组求和，使每个用户每天只有一条数据
select id
,datestr
,sum(amount) as sum_amount
from deal_tb
group by id,datestr
- 根据用户ID分组按日期排序，将日期和分组序号相减得到连续登陆的开始日期，如果开始日期相同说明连续登陆
select tt1.id
,tt1.datestr
,tt1.sum_amount
,date_sub(tt1.datestr,rn) as grp
from(
select t1.id
,t1.datestr
,t1.sum_amount
,row_number() over(partition by id order by datestr) as rn
from(
select id
,datestr
,sum(amount) as sum_amount
from deal_tb
group by id,datestr
) t1
) tt1
- 统计用户连续交易的总额、连续登陆天数、连续登陆开始和结束时间、间隔天数
select ttt1.id
,ttt1.grp
,round(sum(ttt1.sum_amount),2) as sc_sum_amount
,count(1) as sc_days
,min(ttt1.datestr) as sc_start_date
,max(ttt1.datestr) as sc_end_date
,datediff(ttt1.grp,lag(ttt1.grp,1) over(partition by ttt1.id order by ttt1.grp)) as iv_days
from(
select tt1.id
,tt1.datestr
,tt1.sum_amount
,date_sub(tt1.datestr,rn) as grp
from(
select t1.id
,t1.datestr
,t1.sum_amount
,row_number() over(partition by id order by datestr) as rn
from(
select id
,datestr
,sum(amount) as sum_amount
from deal_tb
group by id,datestr
) t1
) tt1
) ttt1
group by ttt1.id,ttt1.grp;
- 结果
1 2019-02-07 13600.23 3 2019-02-08 2019-02-10 NULL
1 2019-02-08 2991.650 5 2019-02-12 2019-02-16 1
1 2019-02-09 1510.8 2 2019-02-18 2019-02-19 1
1 2019-02-10 537.71 1 2019-02-21 2019-02-21 1
2 2019-02-07 13600.23 3 2019-02-08 2019-02-10 NULL
2 2019-02-08 3026.649 4 2019-02-12 2019-02-15 1
2 2019-02-10 1510.8 2 2019-02-18 2019-02-19 2
2 2019-02-11 537.71 1 2019-02-21 2019-02-21 1
3 2019-02-07 13600.23 3 2019-02-08 2019-02-10 NULL
3 2019-02-08 2730.04 5 2019-02-12 2019-02-16 1
3 2019-02-09 1510.8 2 2019-02-18 2019-02-19 1
3 2019-02-10 537.71 1 2019-02-21 2019-02-21 1
标签：02,10,数据仓库,Hive,score,2019,class2,select,class1 来源： https://blog.csdn.net/hujieliang123/article/details/123074895