Mainly used to record Java, DBMS, HDFS... related learning notes
1、创建一个表，字段之间用 \t 分隔；
Hive>create table student (id int, name string) row format delimited fields terminated by '\t' ;
2、将本地一个数据提交到hive里去
hive>load data local inpath '/home/student.txt' into table student ;
2.1 增加分区表数据
alter table ZHIYOUBAO.CHECK_RECORD add partition (years='xxxx',months='xx',days='xx') location '/ZYB/CHECK_RECORD/yy=xxxx/mm=xx/dd=xx/';
3、查询表里的数据：
hive>select * from student ;
4、只查询前两条：
hive>select * from student limit 2 ;
5、统计一个表的行数：
hive>select count(*) from student ;
6、求一个表id字段的id 之和：
hive>select sum(id) from student ;
7、创建外部表：
hive>create external table ext_student (id int, name string) row format delimited fields terminated by ' \t ' location ' /data ' ; //这样就不必将文件放到hive里去 就可以对其进行操作了 ，只需要将文件放到hdfs上的/data目录下面。
8、内部表先有表后有数据；外部表先有数据后有表。
9、创建分区表：
hive>create external table beauties (id bigint, name string, size double) partitioned by (nation string) row format delimited fields terminated by '\t' location '\beauty' ;
hive>load data local inpath '/home/b.c' into table beauties partition(nation='China') ;
hive>alter table beauties add partition (nation='Japan') ;
hive>select * from beauties ;
hive>select * from beauties where nation='China' ; //查找某一分区的数据内容；
10、多表关联：
hive>select t . account , u . name , t . income , t . expenses , t . surplus from user_info u join (select account , sum(income) as income , sum(expenses) as expenses , sum(income-expenses) as surplus from
trade_detail group by account) t on u . account = t . account ;
11、存储过程没有返回值，函数有返回值
12、在linux环境下一次访问hive：
[hh@master ~]$ hive -e "selcte * from mytable limit 3" ;
13、[hh@master ~]$ hive -f 1.hql
14、打印表的字段信息：
hive>describe yourtable ;
15、创建数据库：
hive>create database financials ;
hive>create database if not exists financials ;
16、过滤数据库：
hive>show databases like " f . * " ;
17、添加描述信息：
hive> create database test092302 with dbproperties ('creator'='Mark', 'date'='2015-09-23');
hive> describe database extended test092302;
18、删除数据库：
hive> drop database if exists human_resources; 或者
hive> drop database human_resources;
19、删除存在表的数据库：
hive> drop database test0923 cascade; //在后面加上cascade关键字
20、创建数据库时添加描述信息：
hive> create database test092302 comment 'Holds all test tables'; //使用comment，创建表时也可以用
21、去重查询：group by的使用
hive>select * from mytable group by uid ;
22、独立UID总数：
hive>select count(distinct(uid)) from mytable ; （高效） 或者 hive>select count(*) from(select * from mytable group by uid) a ;
23、查询频度排名（频度最高的前50）：
hive> select keyword,count(*) as cnt from sogou_1w group by keyword order by cnt desc limit 50;
24、将查询的结果放入另一个表中：
hive> create table uid_cnt (uid string, cnt int) row format delimited fields terminated by '\t'; //先创建临时表 uid_cnt
hive> insert overwrite table sogou.uid_cnt select uid,count(*) from sogou_1w group by uid; //再将查询的数据结果放入临时表中
25 修改列名：
hive> alter table test > column ·stuname· name string;“ · ”右上角的~键 describe test;
26 增加列：
hive> alter table test add columns(
> height int);
hive>describe test;
27替换列：
hive> alter table test replace columns(
> id int,
> name string,
> age int);
28 为表添加属性：
hive> alter table test set tblproperties (
> 'note'='hello welcome');
show create table test; ========================================
29 创建带有分区的内部表：
hive> create table testpar(
> id int,
> name string,age int) PARTITIONED BY (day string)
> ROW FORMAT DELIMITED FIELDS TERMINATED BY '\t'
> location '/testpar';
30 为带有分区的内部表加载数据：
hive> load data local inpath '/home/test' into table testpar
> partition (day='0925');
31 添加防止删除的保护： hive> alter table testpar > partition (day=’0925’) enable no_drop;
32 测试：删除分区
hive> alter table testpar drop if exists partition (day='0925');
33 删除添加的”删除”保护：
hive> alter table testpar
> partition (day='0925') disable no_drop;
34 添加防止查询的保护：
hive> alter table testpar
> partition (day='0925') enable offline;
35 删除防止查询的保护:
hive> alter table testpar
> partition (day='0925') disable offline;
select * from testpar; ================================================
36 按条件向分区表插入数据
hive>from test_1 ts
insert into table testpart partition (day=’0920’) select * where ts.age>20
insert into table testpart partition (day=’0919’) select * where ts.name=’xiaofang’;
注释:
上面SQL语句分三部分
第一部分
from test_1 ts 从rest_1表中查询并为其添加ts别名
第二部分
insert into table testpart partition (day='0920') select * where ts.age>20
将test_1表中年龄大于20的数据添加到分区表testpart中新建的0920分区中.
第三部分
insert into table testpart partition (day='0919') select * where ts.name='xiaofang'
将test_1表中名字为xiaofang的数据添加到分区表testpart中新建的0919分区中
查询结果:
hive> select * from testpart;
37 向管理表中加载数据：
hive> load data local inpath '/home/test' overwrite into table testpar partition (day='0925');
38 通过查询语句向表中插入数据：
hive> insert into table testpar
> partition (day='0926')
> select * from test;
hive> select * from testpar;
hive> insert into table testpar
> partition (day='0922')
> select * from test
> where age >20;
hive> from test
> insert into table testpar
> partition (day='0921')
> select * where age>22;
hive> from test ts
> insert into table testpar
> partition (day='0920')
> select * where ts.age>20
> insert into table testpar
> partition (day='0919')
> select * where ts.name='张三';
=========================
动态分区插入
=========================
39 在test表中添加一列day
hive> alter table test add columns(day string);
[hh@master ~]$ vi test
[hh@master ~]$ cat test
1 张三 20 0921
2 李四 22 0922
3 Jarrey 25 0923 40 加载数据：
hive> load data local inpath '/home/test' overwrite into table test;
动态分区（下面两种方式实现的效果是一样的）：
hive> set hive.exec.dynamic.partition=true;
hive> set hive.exec.dynamic.partition.mode=nonstrict;
hive> set hive.exec.max.dynamic.partitions.pernode=1000;
hive> insert into table testpar
> partition(day)
> select * from test;
hive> insert into table testpar
> partition(day)
> select id,name,age,day from test;
41单个查询语句中创建表并加载数据:(注意关键字as)
hive> create table newtest
> as select id,name,age from test
> where name='李四';
hive> select * from newtest;
=========================
导出数据
=========================
42 Hadoop fs –cp source_path target_path cp scp -r /jdk slave://home/
注释:
scp =safety copy 即是安全模式下复制 r=recuresive 递归方式复制 即是从主目录到各个子目录依次复制
=========================
Sqoop工具（T15）
=========================
43 从hdfs集群中加载数据
hive>load data inpath 'hdfs目录文件' into table student;
44 按id降序排序
hive>select * from student order by id desc;
45 从hdfs集群中加载数据并为表设置指定分区
hive>load data input '本地文件路径' into table 表名 partition (分区字段=' ');
46 从本地内存中加载数据
hive>load data local inpath '本地目录文件' into table student;
47 按id降序排序
hive>select * from student order by id desc;
48 表联合查询
hive>select t.account u.name,t.income,t.expenses,t.surplus from user_info
u join (select account, sum(income) as income,sum(expenses) as expenses,sm(income_expenses)
as surplus from trade_detail group by account) on u.account=t.account;
=========================
数学函数
=========================
Hive语句运算:
49 int类型rank加运算
hive>select rank+1 from ext_sogou_20111230 limit 100;
50 对int字段平方
hive> select pow(rank,2) from ext_sogou_20111230;
51 取模:(如:2对三取模)
hive>select pmod(2,3) from ext_sogou_20111230 limit 10;
=========================
聚合函数
=========================
52
hive>select count(*) from ext_sogou_20111230 limit 10;
*表示表中所有字段也可以设置某些或某个字段 如
hive>select count(uid,ts) from ext_sogou_20111230 limit 10;
hive>select sum(uid) from ext_sogou_20111230;
54 最大值&最小值
hive>select max(rank), min(rank) from ext_sogou_20111230;
55 .独立uid(去重行数)
hive>select count(distinct uid) from ext_sogou_20111230;
56强转:
hive> select cast(rank as DOUBLE) from ext_sogou_20111230 limit 10;
57 拼接:
hive>select concat(uid,url) from ext_sogou_20111230 limit 10;
=========================
JSON
=========================
58 抽取JSON对象的某一属性值 hive>select get_json_object(‘{“name”:”xiaoming”,”age”:”15”}’,’$.age’) from ext_sogou_20111230 limit 5; 结果: 15
59
hive>select get_json_object(channel,'$.age') from ext_sogou_20111230 limit 3;
=============================================
60 查找url字符串中的5位置之后字符串baidu第一次出现的位置
hive> select locate("baidu",url,5) from ext_sogou_20111230 limit 100;
61 .抽取字符串baidu中符合正则表达式url的第5个部分的子字符串
hive> select regexp_extract("baidu",url,5) from ext_sogou_20111230 limit 100;
62 按照正则表达式”0”分割字符串uid,并将分割后的部分以字符串数组的方式返回
hive> select split(uid,"0") from ext_sogou_20111230 limit 100;
结果之一:["","875edc8a14a228","1bac1ddc","1fa18a1"]
63 对字符串url,从0处开截取长度为3的字符串,作为其子字符串
hive> select substr(url,0,3) from ext_sogou_20111230 limit 3;
64 .将字符串url中所有的字母转换成大写字母 、 hive> select upper(url) from ext_sogou_20111230 limit 3;
=====================
别名 嵌套SQL语句
=====================
65 复杂HQL 如别名、嵌套等
hive>select count(distinct e.uid) from (select * from ext_sogou_20111230 where
rank <=3 and order =1) e;
小括号中返回的也是一个表,它只是临时的 别名为e
66 where ..and 或者 where ….or where的 两种条件查询
hive> select * from ext_sogou_20111230 where rank<=3 and order =1 limit 3;
hive> select * from ext_sogou_20111230 where rank !=0 or order =1 limit 3;
where
1 出现在表后
2 可以有and or 表达式的操作符
3 表示格式
67 浮点类型的比较 一定要强转
68 like 过滤字符串
它是一个标准的SQL操作符
hive> select * from ext_sogou_20111230 where url like '%http%' limit 10;
'%http%'意为包含 http字符串
'%http' 以http开头的字符串
'http%'一http结束字符串
69 rlike 通过Java的正则表达式过滤 *与%功能一样
它是hive中扩展功能的操作符
hive> select * from ext_sogou_20111230 where url rlike ' .*http.* ' limit 3;
=======================
group by
=======================
70 Group by 语句通常会和聚合函数一起使用,按照一个或者多个对结果进行分组,然后对每个组执行聚合操作
hive>select year(ts), avg(rank) from ext_sogou_20111230 where ts like '%2011' group by year(ts);
71 对组过滤
hive> select rank ,count(*) from ext_sogou_20111230 group by rank ,order having rank >3 limit 10;
==========================
join
==========================
72 join 使用join时要选择具有独立的字段作为条件字段,否则会出现不必要的数据量
hive> select m.uid,m.keyword from ext_sogou_20111230 m join ext_sogou_20111230_limit3 n on m.uid =n.uid;
73 查搜索过”仙剑奇侠传” 的用户所搜过的关键字
hive>select m.uid,m.keyword from (select distinct n.uid from
ext_sogou_20111230 where keyword like '%仙剑奇侠传%' n ) m
where m.uid=n.uid;
74 查搜索过”仙剑奇侠传” 的用户所搜过的不包含”仙剑奇侠传”本身的关键字
hive>select m.uid,m.keyword from sogou_20111230 m join (select distinct uid from sogou_20111230 where keyword like '%仙剑奇侠传%') n on m.uid=n.uid where m.keyword not like '%仙剑奇侠传%';
75 left semi-join 左半表 semi 半挂的 半独立的
hive>select * from be where rank in(1,2,5);
hive>select * from ext_sogou_20111230 m left semi join ext_sogou_20111230_limit3 n on m.rank=n.rank;
76笛卡尔积
如5w 1w join 结果:5w*1w 一般不常用
77 map-side JOIN当两张表很小时使用(系统默认25MB)
功能:其中一张表为小表 即是将小表数据JOIN到大表中
hive>select /*+MAPJOIN(n)*/ m.uid,m.keyword,n.keyword
from ext_sogou_20111230 m join ext_sogou_20111230_limint3 n on m.uid=n.uid;
=====================
排序
=====================
78 全局排序(order by ) 和局部排序 (sort by)
hive>select * from ext_sogou_20111230 order by rank desc limit 100;
79 对sogou500w中降序排列uid次数
hive>select uid, count(*) as nct from ext_sogou_20111230 group by uid order by nct desc ;
80 cast()类型转换函数
hive>select cast(ts as bigint) from
ext_sogou_20111230_limit3;
81 UNION ALL可以将2个或多个表进行合并。
hive> select count(distinct e.uid)from(
select * from ext_sogou_20111230 where rank<11
union all
select * from ext_sogou_20111230_limit3 where rank < 11) e;
82
hive>select count(*) from ext_sogou_20111230_limit where keyword like '%www%';
83
hive> select e.url,e.keyword,count(*) from (
select * from ext_sogou_20111230 where keyword like '%www%'
)e group by e.url,e.keyword where instr(url,keyword) >0;
84搜索过’%仙剑奇侠传%’(模糊匹配),并且查询次数大于3的UID
hive>select uid, count(uid) as nct from
ext_sogou_20111230 where keyword like '%仙剑奇侠传%'
group by uid having nct>3 ;
================================
视图
================================
85视图 hive只支持逻辑视图 作用降低查询复杂度
创建视图
hive>create view sogou_view as
select * from ext_sogou_20111230 where rank <=3;
86 索引
Hive的索引需要单独创建表实现
创建索引
hive>CREATE INDEX employees_index ON TABLE employees (name) AS
'org.apache.hadoop.hive.ql.index.compact.CompactIndexHandler'
WITH DEFERRED REBUILD IDXPROPERTIES('creator' = 'me','
created_at '='some time') IN TABLE employees_index_table;
87 视图
hive>create view sogou_filter as select uid,count(*) from
ext_sogou_20111230 where keyword like '%仙剑奇侠传%'
复杂问题解题思路:
1)分步骤,使用临时表
2)分步骤,多个视图实现
create view
3)一个复杂的SQL
create table insert overwrite table...select * from ...
=======================================
Sogou 500w数据 88
搜索长度大于256（不区分中英文），并且点击次数<3的UID
老师:
hive>select m.uid,count(*) as cnt from(select * from sogou_view where length(
keyword) >256) m group by m.uid having cnt<3;
自己:
select uid from sogou_view where rank<3 and length(
keyword) >256;
hive> create view sogou_view as select * from
ext_sogou_20111230;
89
上午7-9点之间，搜索过“赶集网”的用户，哪些用户直接点击了赶集网的URL
老师:
hive> select distinct n.uid from (select * from sogou_view where keyword ='赶集网')
and substr(ts,9,2) in ('07','08','09')) n where n.url like '%ganjin.com%';
自己:
hive> select uid from sogou_view where (cast(substr(ts,9,2)
as int)>7 or cast(substr(ts,9,2) as int)<9) and url
like '%www.ganji.com%' or keyword like '%赶集网%' ;
或者
hive>select uid from sogou_view where substr(ts,9,2) in ('07','08','09') and url
like '%www.ganji.com%' and keyword like '%赶集网%' ;
90
rank<3的搜索中，多少用户的点击次数>2
老师:
hive>select a.uid from (select uid,count(*) as cnt from (select * from sogou_view where
rank<3) e group by e.uid having cnt>2) a;
自己:
hive>select uid,count(uid) as nct from sogou_view
where rank<3 group by uid having nct>2;
=======================
hive设计模式
=======================
1.表的划分方式:按天划分如table_2011_01_01
2.分区:hive中的分区功能很有用,
3 最原始的数据尽量少使用分区,
经过加工后的数据可以用分区.
4 表与分区的字段不能重复
5 分区有级别 根据实际的业务自定义分区
create table supply () partitioned by();
91 同一份数据多种处理
hive>insert overwrite table sogou_20111230_rank
select * from sogou_20111230 where rank=3;
92
hive>insert overwrite table sogou_20111230_order
select * from sogou_20111230 where order=3;
上面两句(91 92)合并成一句(93)如下
93
hive>from sogou_20111230
insert overwrite table sogou_20111230_rank
select * where rank =3
insert overwrite table sogou_20111230_order
select * where order=3;
94 为表增加列 (只能末尾追加)
ALTER TABLE sogou_20111230 ADD COLUMNS (user_id string) ;
列的存储有两种格式ORC和RCFile
========================================
Hive内置函数和UDF(用户自定义函数)
========================================
95 查看内置函数
hive> show functions;
96 查看某一函数具体描述
hive>describe function 函数名;
一般聚合函数与group by 组合使用
分3种:
1 UDF(标准函数):普通函数
2 UDAF(用户自定义聚合函数):多行多列变一行
3 UDTF(用户自定义表生成函数):多行多列变多行
==UDF操作过程==
91 在eclipse中创建java类 如UDFZodiacSign
92 添加UDFZodiacSign的jar包
hive>add jar /home/udf.jar
93 创建外部表如little_bigdata
hive>create external table if not exists
little_bigdata(name string,email string,bday
string,ip string, gender string, anum int)
row format delimited fields terminated by ',';
94 创建zodiac作为UDFZodiacSign类的临时函数 as’包名.类名’
hive>create temporary function zodiac as 'day1008.UDFZodiacSign';
95 查看zodiac是否OK
hive> describe function zodiac;
96 将little_bigdata表中name字段中数据传入临时函数zodiac中
hive> select zodiac(name) from little_bigdata;
============================================
97 统计没有农产品市场的省份有哪些
马:
hive> select e.name from (
select distinct prov from product
) a right outer join province e on a.prov = e.name
where a.prov is null
98统计排名前 3 的省份共同拥有的农产品类型
1计算前三省份的名称
2计算前三省份的所有去重产品名称
3计算共同拥有的产品
数据按A B C D E 步骤计算
hive>select c.name,count(*) as ct from
E 列出前三省相同的熟菜,并计数
(select a.prov,a.name from
D 从A数据中比较与B中前三个相同列 的省份及其熟菜
(select prov,name from product group by prov,name
A 分组列出所有省,及其所在省的熟菜(分组就是去重)
) a
left semi join
(select p.prov,count(*) as cnt from
C 对不同省份计数 省1 number1 省2 number2
并按降序排列列出前三个省
(select prov,name from product group by prov,name
B 分组列出所有省,及其所在省的熟菜(分组就是去重)
) p
group by p.prov order by cnt desc limit 3
) b
on a.prov = b.prov
) c group by c.name having ct > 2
hive> select (2015-age)as ag ,sex from car_1 where age !=null or sex !="";
hive> select m.ag,count(*) as nct from
(select (2015-age) as ag ,sex from car_1 where age !=null or sex !="")
m group by m.ag; --------------------------------------------------------------------------------------
=============================
自定义Hive文件和记录格式
=============================
hive三种文件格式:textfile sequencefile rcfile 前两种一行存储 rcfile以列存储 他们影响整个文件格式
sequencefile 与 textfile 文件格式在读取效率上 testfile更高些
默认分隔符格式/001 即是Ctr+A stored as textfile 表文件的存储格式
99 创建sequencefile格式的表
hive>create external table sogou_20111230_seq(ts string,
uid string,keyword string,rank int,order int
,url string) row format delimited fields
terminated by '\t' stored as sequencefile;
100 向该表中插入数据
hive>insert table sogou_20111230_seq select
ts,uid,keyword,rank,order,url from
sogou_20111230 limit 50000;
101 创建rcfile格式的表:基于列式存储
hive>create table sogou_20111230_rc(ts string,
uid string, keyword string,rank int, order
int, url string) row format delimited fields
terminated by '\t' stored as rcfile;
102 向该表中插入数据
hive>insert overwrite table sogou_20111230_rc
select ts, uid,keyword,rank,order,url
from ext_sogou_20111230 limit 50000;
103 记录格式 SerDe是序列化/反序列化的简写
104 CSV和TSV SerDe(csv内部实现各式逗号分割\n换行)
hive 记录格式:影响文件内部数据存储格式 105 XPath相关的函数
hive>SELECT xpath ('
<a><b id="foo">bl</b>
<b id="bar">b2</b></a>','//@id' )
FROM car_1 LIMIT 1;
106 计算北京市的每种农产品的价格波动趋势，即计算每天价格均值，并按照时间先后顺序排列该值。
某种农产品的价格均值计算公式：
PAVG = (PM1+PM2+...+PMn-max(P)-min(P))/(N-2)
其中， P 表示价格， Mn 表示 market，即农产品市场。 PM1 表示 M1 农产品市场的该产品价
格， max(P)表示价格最大值， min(P)价格最小值。
思路:
第一步:筛选出1-5天内 时间 熟菜名称 两个字段
第二步:用if三目运算,判断各种熟菜波动次数是否大于2次,
第三步:求平均值
hive>select m.date,m.name,if(count(*)>2,
round((sum(m.price)-max(m.price)-min(m.price))/(count(*)-2),2),
round(sum(m.price)/count(*),2))
from (
select * from product_20140101 where province='北京'
union all
select * from product_20140102 where province='北京'
union all
select * from product_20140103 where province='北京'
union all
select * from product_20140104 where province='北京'
union all
select * from product_20140105 where province='北京'
) m
group by m.date,m.name;
107 使用简单时间序列算法， 设置 N=3，预测 1.4、 1.5 日的平均价格
hive>create table price_hg_pre0104(ptime TIMESTAMP,name STRING,price FLOAT);
hive>insert overwrite table price_hg_pre0104
select * from price_hg where day(cast(ptime as string)) < 4
union all
select cast('2014-01-04 00:00:00' as timestamp) as ptime,'黄瓜' as name,sum(price)/3 as price from price_hg where day(cast(ptime as string)) < 4
108 并计算与实际数据的平方误差和
hive>create table price_hg_pre0105(ptime TIMESTAMP,name STRING,price FLOAT);
hive>insert overwrite table price_hg_pre0105
select cast('2014-01-05 00:00:00' as timestamp) as ptime,'黄瓜'
as name,sum(price)/3 as price from price_hg_pre where day(cast(ptime as string)) < 5
and day(cast(ptime as string)) > 1
109 表添加一列 ：
hive> ALTER TABLE pokes ADD COLUMNS (new_col INT);
110 添加一列并增加列字段注释
hive> ALTER TABLE invites ADD COLUMNS (new_col2 INT COMMENT 'a comment');
111 更改表名：
hive> ALTER TABLE events RENAME TO 3koobecaf; 112 删除列：hive> DROP TABLE pokes;
113增加、删除分区
•增加
ALTER TABLE table_name ADD [IF NOT EXISTS] partition_spec [ LOCATION 'location1' ] partition_spec [ LOCATION 'location2' ] ...
partition_spec:
: PARTITION (partition_col = partition_col_value, partition_col = partiton_col_value, ...)
•删除
ALTER TABLE table_name DROP partition_spec, partition_spec,...
REPLACE则是表示替换表中所有字段。
114 重命名表•
ALTER TABLE table_name RENAME TO new_table_name
115 修改列的名字、类型、位置、注释：
ALTER TABLE table_name CHANGE [COLUMN] col_old_name col_new_name column_type [COMMENT col_comment] [FIRST|AFTER column_name]
•这个命令可以允许改变列名、数据类型、注释、列位置或者它们的任意组合
116 表添加一列 ：
hive> ALTER TABLE pokes ADD COLUMNS (new_col INT);
117 添加一列并增加列字段注释
hive> ALTER TABLE invites ADD COLUMNS (new_col2 INT COMMENT 'a comment');
118增加/更新列
ALTER TABLE table_name ADD|REPLACE COLUMNS (col_name data_type [COMMENT col_comment], ...)
复制代码
• ADD是代表新增一字段，字段位置在所有列后面(partition列前)
119
create external table logs (ip string ,name1 string,name2 string,name3 string ,name4 string ,name5 string, name6 string, name7 string, name8 string,name9 string,name10 string,name11 string) row format delimited fields terminated by ' ';
select name1 from (select as c from logs where ip ='58.214.255.146';
数据格式：
183.166.128.178 - - [09/Apr/2016:07:58:33 +0800] "POST /boss/service/newCode.htm HTTP/1.1" 200 227 "-" "-" 120正序：
select ip ,sum(name9) as c from logs where name3 like '[09/Apr/2016:07:55%' group by ip order by c desc;
121 逆序：
select name6,count(1) as b from logs where name3 like '[09/Apr/2016:07:5%' group by name6 order by b asc; 122 逆序：
select name6,count(1) as b from logs where name3 like '[09/Apr/2016:07:5%' group by name6 sort by b asc;
123取前一千行放到一个新表里
hive> insert into table hivecontain_small
> select * from hivecontain limit 1000;
124 更新表字段
hive>insert overwrite table province_city_scenic_per_nums select spot_name, spot_city, substring(round(per,4),0,6) ,nums from province_city_scenic_per_nums ;
125 截取表字段部分值并插入新表
hive> insert table province_city_scenic_per_nums select spot_name, spot_city, substring(round(per,4),0,6) ,nums from province_city_scenic_per_nums ; 126.tourist_consume_details 用户消费信息（金额、订单数、游玩人次）
select link_name, sex, city ,tel, certificate_no ,sum(close_total_price) as total_price ,sum(popnum) as popnum,count(tourname) as tournum from order_raw_info
group by link_name,sex,city ,tel,certificate_no;
127.bucketed_user 分桶查询随机id
select * from bucketed_user TABLESAMPLE(BUCKET 1 OUT OF 4 ON rand())
128.bucketed_user 创建带桶的外部表
create external table if not exists bucketed_user2(id int,name string) clustered by (id) sorted by(name) into 4
buckets row format delimited fields terminated by ',' stored as textfile location '/kafka/' ;
129.province_city_scenic_per_nums 表字段的截取
select spot_name, spot_city, substring(round(per,4),0,6) ,nums from province_city_scenic_per_nums ;
130.province_city_scenic_per_nums 表字段更新(如0.001242更新为0.001)
insert overwrite table province_city_scenic_per_nums select spot_name, spot_city, substring(round(per,4),0,6) ,nums from province_city_scenic_per_nums ;
131.改表字段
alter table scenic_tour_info change `spotname` spot_name string;
132.split使用
select split("13901888346","1390188")[1] from quyu_visit_info limit 10;
133.quyu_visit_info 按条件插入表数据
insert into table quyu_visit_info select u.visitor_id, u.tel, u.city from solo_mobile_quyu u limit 100;
134.tourt_ype_date_total (new) 景区类型（按时间分组）游客量统计
select * from(
select p.date, p.tour_type,count(p.date) total
from (
select tour_type, substr(occ_date, 0,4) as date
from scenic_tour_info ) p
where p.date like '201%'
group by p.tour_type, p.date
order by p.date desc ) t
where t.date='2013' or t.date='2014' or t.date='2015' or t.date='2016';
135.date_spotprovince_type_total 按省份统计 景区类型（按时间分组）游客量统计
select * from(
select p.date, p.spot_province, p.tour_type as scenic_type,count(p.date) total
from (
select spot_province, tour_type, substr(occ_date, 0,4) as date
from scenic_tour_info ) p
where p.date like '201%'
group by p.spot_province,p.tour_type, p.date
order by p.date desc ) t
where t.date='2013' or t.date='2014' or t.date='2015' or t.date='2016' ; 136.scenic_city_province_per_nums 统计某省各景区客流量的比重（占该省比重）及其客流量
select distinct p.spot_name, p.spot_city, (p.nums/5358582) per ,p.nums
from province_city_scenic_nums p join province_tour_nums c
on p.spot_province='浙江省'
order by per desc 137.表重命名
ALTER TABLE tour_info_detail RENAME TO new_name; scenic_info_detail ; 138.rename_ziduan 重命名表字段名
alter table scenic_info_detail change `proname` spotprovince string;