Hive Query Language

Ahmad Alkilani
www.pluralsight.com
 Outline
 Data Types

 Load and Organize Data

 Managed/External Partitioned Tables
 Dynamic Partition Inserts

 Single Scan-Multiple Inserts

 Hive Functions, Aggregates, Group By, Cube, Rollup, Having

 Sorting and Clustering Results

 Using the CLI in the real world

 Batch mode
 Variable Substitution
 Primitive Data Types

Numeric

• TINYINT, SMALLINT, INT, BIGINT

• FLOAT
• DOUBLE
• DECIMAL – Starting Hive 0.11

Date/Time

• TIMESTAMP starting Hive 0.8

• Strings must be in format "YYYY-MM-DD HH:MM:SS.fffffffff"
• Integer types as UNIX timestamp in seconds from UNIX epoch
• Floating point types same as Integer with decimal precision
• DATE starting Hive 0.12

Misc.

• BOOLEAN
• STRING
• BINARY
 Complex/Collection Types

Type Syntax

Arrays ARRAY<data_type>
Maps MAP<primitive_type, data_type>
Struct STRUCT<col_name : data_type [COMMENT col_comment],
…>
Union Type UNIONTYPE<data_type, data_type, …>

CREATE TABLE movies (

movie_name string,
participants ARRAY<string>,
release_dates MAP<string, timestamp>,
studio_addr STRUCT<state:string, city:string, zip:string, streetnbr:int, streetname:string,
unit:string>,
complex_participants MAP<string, STRUCT<address:string, attributes MAP<string,
string>>>
misc UNIONTYPE<int, string, ARRAY<double>>
);
 Complex/Collection Types
CREATE TABLE movies (
movie_name string,
participants ARRAY<string>,
release_dates MAP<string, timestamp>,
studio_addr STRUCT<state:string, city:string, zip:string, streetnbr:int, streetname:string,
unit:string>,
complex_participants MAP<string, STRUCT<address:string, attributes MAP<string,
string>>>
misc UNIONTYPE<int, string, ARRAY<double>>
“Inception” 2010-07-16 00:00:00 91505 “Dark Green”
); {0:800}
“Planes” 2013-08-09 00:00:00 91505 “Green” {3:[1.0, 2.3, 5.6]}
SELECT movie_name,
participants[0],
release_dates[“USA”],
studio_addr.zip,
complex_participants[“Leonardo
DiCaprio”].attributes[“fav_color”],
misc
FROM movies;
 Type Conversions

Implicit
Conversions

DOUBLE FLOAT BOOLEAN STRING

BIGINT SMALLINT TINYINT

FLOAT STRING INT TIMESTAMP
16L 16S 16Y

Explicit Conversions
 CAST(‘13’ AS INT)
 CAST(‘This results in NULL’ AS INT)
 CAST(‘2.0’ AS FLOAT)
 CAST(CAST(binary_data AS STRING) AS DOUBLE)
Loading and organizing data in Hive

Hive Query Language

 Table Partitions
 Managed Partitioned Tables
CREATE TABLE page_views ( eventTime STRING, userid STRING, page STRING)
PARTITIONED BY(dt STRING, applicationtype STRING)
STORED AS TEXTFILE;

/apps/hive/warehouse/page_views
/apps/hive/warehouse/page_views/dt=2013-08-
10/application=android
LOAD DATA INPATH
page_views ‘/mydata/android/Aug_10_2013/pageviews/’
INTO TABLE page_views
PARTITION (dt = ‘2013-08-10’, applicationtype = ‘android’);
dt=2013-08-10 LOAD DATA INPATH
‘/sample/android/Aug_10_2013/pageviews/’
OVERWRITE INTO TABLE page_views
application=androi PARTITION (dt = ‘2013-08-10’, applicationtype = ‘android’);
d
 Table Partitions
 Virtual Partition Columns
CREATE TABLE page_views ( eventTime STRING, userid STRING, page STRING)
PARTITIONED BY(dt STRING, applicationtype STRING)
STORED AS TEXTFILE;

eventTime STRING
userid STRING
page STRING
dt STRING
applicationtype STRING

SELECT dt as eventDate, page, count(*) as pviewCount FROM page_views

WHERE applicationtype = ‘iPhone’;
 Table Partitions
 External Partitioned Tables
CREATE EXTERNAL TABLE page_views ( eventTime STRING, userid STRING,
page STRING)
PARTITIONED BY(dt STRING, applicationtype STRING)
STORED AS TEXTFILE;
eventTime STRING
userid STRING
page STRING
dt STRING
applicationtype STRING
ALTER TABLE page_views ADD PARTITION (dt=‘2013-09-09’, applicationtype=‘Windows Phone 8’)
LOCATION ‘/somewhere/on/hdfs/data/2013-09-09/wp8’;

ALTER TABLE page_views ADD PARTITION (dt=‘2013-09-09’, applicationtype=‘iPhone’)

LOCATION ‘hdfs://NameNode/somewhere/on/hdfs/data/iphone/current’;

ALTER TABLE page_views ADD IF NOT EXISTS

PARTITION (dt=‘2013-09-09’, applicationtype=‘iPhone’) LOCATION
‘/somewhere/on/hdfs/data/iphone/current’
PARTITION (dt=‘2013-09-08’, applicationtype=‘iPhone’) LOCATION
‘/somewhere/on/hdfs/data/prev1/iphone’
PARTITION (dt=‘2013-09-07’, applicationtype=‘iPhone’) LOCATION
‘/somewhere/on/hdfs/data/iphone/prev2’;
Demo
 Multiple Inserts
 Interchangeability of blocks
FROM movies
SELECT *;

 Syntax
FROM from_statement
INSERT OVERWRITE TABLE table1 [PARTITION (partcol1=val1, partcol2=val2)] select_statement1
INSERT INTO TABLE table2 [PARTITION (partcol1=val1, partcol2=val2) [IF NOT EXISTS]]
select_statement2
INSERT OVERWRITE DIRECTORY ‘path’ select_statement3;

 Extract action and horror movies into tables for further processing
FROM movies
INSERT OVERWRITE TABLE horror_movies SELECT * WHERE horror = 1 AND release_date = ‘8/23/2013’
INSERT INTO action_movies SELECT * WHERE action = 1 AND release_date = ‘8/23/2013’;

FROM (SELECT * FROM movies WHERE release_date = ‘8/23/2013’) src

INSERT OVERWRITE TABLE horror_movies SELECT * WHERE horror = 1
INSERT INTO action_movies SELECT * WHERE action = 1;
 Dynamic Partition Inserts
CREATE TABLE views_stg (eventTime STRING, userid STRING)
PARTITIONED BY(dt STRING, applicationtype STRING, page STRING);

FROM page_views src

INSERT OVERWRITE TABLE views_stg PARTITION (dt=‘2013-09-13’, applicationtype=‘Web’, page=‘Home’)
SELECT src.eventTime, src.userid WHERE dt=‘2013-09-13’ AND applicationtype=‘Web’, page=‘Home’
INSERT OVERWRITE TABLE views_stg PARTITION (dt=‘2013-09-14’, applicationtype=‘Web’, page=‘Cart’)
SELECT src.eventTime, src.userid WHERE dt=‘2013-09-14’ AND applicationtype=‘Web’, page=‘Cart’
INSERT OVERWRITE TABLE views_stg PARTITION (dt=‘2013-09-15’, applicationtype=‘Web’,
page=‘Checkout’)
SELECT src.eventTime, src.userid WHERE dt=‘2013-09-15 AND applicationtype=‘Web’, page=‘Checkout’
FROM page_views src
INSERT OVERWRITE TABLE views_stg PARTITION (applicationtype=‘Web’, dt, page)
SELECT src.eventTime, src.userid, src.dt, src.page WHERE applicationtype=‘Web’

 Dynamically determine partitions to create and populate

 Use input data to determine partitions
 Dynamic Partition Inserts
 Default maximum dynamic partitions = 1000
 hive.exec.max.dynamic.partitions
 hive.exec.max.dynamic.partitions.pernode

 Enable/Disable dynamic partition inserts

 hive.exec.dynamic.partition=true

 Use strict mode when in doubt

 hive.exec.dynamic.partition.mode=strict

 Increase max number of files a data node can service in (hdfs-site.xml)

 dfs.datanode.max.xcievers=4096
 Table Partitions
 Partitions for managed tables created by loading data into table
 LOCATION for EXTERNAL partitioned tables is optional
 Advantages to using same directory structure of managed tables
 Apache Hive
 MSCK REPAIR TABLE table_name;
 Amazon's Elastic Map Reduce
 ALTER TABLE table_name RECOVER PARTITIONS;
 Virtual columns and column name collision
 ALTER TABLE ADD PARTITION isn’t restricted to managed tables
 ALTER TABLE table_name [PARTITION spec] SET LOCATION "new
location“
 Not everything results in partition pruning
 Data is in lowest level, leaf, directory
 When filter doesn’t show in explain plan that means partition pruning
was used to service the predicate.
Data Retrieval

Hive Query Language

 Group By

SELECT a b _c0
a, b, SUM(c) 1 B 10
FROM 1 H 30
t1 a b c
1 S 10
GROUP BY 1 H 10
2 A 10
a, b 2 A 10
1 H 20
1 B 10
SELECT a _c0
a, SUM(c) 1 S 10
1 50
FROM 2 10
t1
GROUP BY
a
 Grouping Sets, Cube, Rollup

SELECT a, b, SUM(c) FROM t1 GROUP BY a, b GROUPING SETS ((a,b),a)

SELECT a, b, SUM(c) FROM t1 GROUP BY a, b

UNION ALL
SELECT a, NULL, SUM(c) FROM t1 GROUP BY a

SELECT a, b, SUM(c) FROM t1 GROUP BY a, b GROUPING SETS (a,b,())

SELECT a, NULL, SUM(c) FROM t1 GROUP BY a

UNION ALL
SELECT NULL, b, SUM(c) FROM t1 GROUP BY b
UNION ALL
SELECT NULL, NULL, SUM(c) FROM t1
 Grouping Sets, Cube, Rollup

Cube

SELECT a, b, c, SUM(d) FROM t1 GROUP BY a, b WITH CUBE

SELECT a, b, c, SUM(d) FROM t1 GROUP BY a, b, c GROUPING SETS

((a,b,c),(a,b),(b,c),(a,c),a,b,c,())

Rollup
SELECT a, b, c, SUM(d) FROM t1 GROUP BY a, b WITH ROLLUP

SELECT a, b, c, SUM(d) FROM t1 GROUP BY a, b, c GROUPING SETS

((a,b,c),(a,b),a,())
 Functions in Hive

 Built -in Functions

 Mathematical
 Collection
 Type conversion
 Date
 Conditional
 String
 Misc.
 xPath

 UDAFs

 UDTFs
 Built-in Functions

 Mathematical
SELECT rand(), a FROM t1; SELECT rand(3), rand(a) FROM t1;
SELECT pow(a, b) FROM t2; SELECT tan(a) FROM t3;

abs(double a)
round(double a, int d)
floor(double a)

 Collection
size(Map<K.V>)
map_keys(Map<K.V>)
map_values(Map<K.V>)

SELECT array_contains(a, ‘test’) FROM t1;

 Built-in Functions

 Date
unix_timestamp()
year(string d), month(string d), day(string d), hour, second
datediff(string enddate, string startdate)
date_add(string startdate, int days)
date_sub(string startdate, int days)
to_date(string timestamp)

 Conditional
SELECT IF(a = b, ‘true result’, ‘false result’) FROM t1;
SELECT COALESCE(a, b, c) FROM t1;
SELECT CASE a WHEN 123 THEN ‘first’ WHEN 456 THEN ‘second’
ELSE ‘none’ END FROM t1;
SELECT CASE WHEN a = 13 THEN c ELSE d END FROM t1;
 Built-in Functions

 String
SELECT concat(a, b) FROM t1; SELECT concat_ws(sep, a, b) FROM t1;
SELECT regex_replace(“Hive Rocks”, “ive”, “adoop”) FROM dummy;

substr(string|binary A, int start)

substring(string|binary A, int start, int length)

sentences(string str, string lang, string locale)

SELECT sentences(“Loving this course! Hive is awesome.”) FROM dummy;

((“Loving”, “this”, “course”), (“Hive”, “is”,

“awesome”))
 Built-in Aggregate Functions (UDAFs)
COUNT(*), COUNT(expr), COUNT(DISTINCT expr)
SUM(col), SUM(DISTINCT col)

AVG, MIN, MAX, VARIANCE, STDDEV_POP

HISTOGRAM_NUMERIC(col, b)
returns array<struct {‘x’, ‘y’}>

array[0].y
array[2].y
array[1].y

array[0].x array[1].x array[2].x

 HAVING & GROUP BY

 Having Syntax

SELECT
a, b, SUM(c)
FROM  Group By on Function
t1
GROUP BY SELECT
a, b CONCAT(a,b) as r
HAVING , SUM(c)
SUM(c) > 2 FROM
t1
GROUP BY
CONCAT(a,b)
HAVING
SUM(c) > 2
 Sorting in Hive
ORDER BY
SELECT x, y, z FROM t1 ORDER BY x ASC

part-00000

A A
Map
B A
D Reducer B
C C
Map
A D
 Sorting in Hive
SORT BY
SELECT x, y, z FROM t1 SORT BY x
part-00000
A
Reducer
A
D
Map
B
part-00001
D Reducer A
C C
Map
A
part-00002
Reducer B
 Controlling Data Flow
DISTIRBUTE BY
SELECT x, y, z FROM t1 DISTRIBUTE BY y

key y z
Reducer
x1 1 A
Map x1 2 B
x1 3 D Reducer
x1 4 C
Map
x2 5 A

Reducer
 Controlling Data Flow
DISTIRBUTE BY
SELECT x, y, z FROM t1 DISTRIBUTE BY y

key y z
Reducer
x1 1 A
Map x1 2 B
x1 1 D Reducer
x1 4 C
Map
x2 5 A
Reducer
 Controlling Data Flow
DISTIRBUTE BY with SORT BY
SELECT x, y, z FROM t1 DISTRIBUTE BY y SORT BY z

key y z
Reducer
x1 1 A
Map x1 2 B
x1 1 D Reducer
x1 4 C
Map
x2 5 A
Reducer

CLUSTER BY
SELECT x, y, z FROM t1 CLUSTER BY y
Command line options and variable substitution

Hive CLI
 The CLI
 hive
 hive -e ‘select a, b, from t1 where c = 15’
 hive -S -e ‘select a, b from t1’ > results.txt
-e and -f run hive in batch mode
 hive -f /my/local/file/system/get-data.sql

 Variable Substitution
$ hive -d srctable=movies
4 namespaces
hive> set hivevar:cond=123;
 hivevar hive> select a,b,c from pluralsight.${hivevar:srctable}
 -d, --define , --hivevar where a = ${hivevar:cond};
 set hivevar:name=value
$ hive -v -d src=movies -d db=pluralsight -e 'select * from
 hiveconf ${hivevar:db}.${hivevar:src} LIMIT 100;‘
 --hiveconf
 set hiveconf:property=value
 system
 set system:property=value
 env
 set env:property=value
 Summary
 Data Types
 Primitive and Complex

 Table Partitioning
 Managed tables by loading data
 Alter Table for External tables
 Dynamic partition inserts

 Multi Inserts

 Functions

 Order By, Sort By, Distribute By, Cluster By

 The Hive CLI