hadoop pig入门总结

在这里贴一个pig源码的分析，做pig很长时间没做笔记，不包含任何细节，以后有机会再说吧

http://blackproof.iteye.com/blog/1769219

hadoop pig入门总结

pig简介
pig数据类型
pig latin语法
pig udf自定义
pig derived衍生
推荐书籍 programming pig
推荐网站 http://pig.apache.org/docs/r0.10.0/basic.html

pig简介

pig是hadoop上层的衍生架构，与hive类似。对比hive（hive类似sql，是一种声明式的语言），pig是一种过程语言，类似于存储过程一步一步得进行数据转化。

pig数据类型

double > float > long > int > bytearray
tuple|bag|map|chararray > bytearray

double float long int chararray bytearray都相当于pig的基本类型

tuple相当于数组，但是可以类型不一，举例('dirkzhang','dallas',41)

Bag相当于tuple的一个集合，举例{('dirk',41),('kedde',2),('terre',31)}，在group的时候会生成bag

Map相当于哈希表，key为chararray，value为任意类型，例如['name'#dirk,'age'#36,'num'#41

nulls 表示的不只是数据不存在，他更表示数据是unkown

pig latin语法

1：load

LOAD 'data' [USING function] [AS schema];

例如：

load = LOAD 'sql://{SELECT MONTH_ID,DAY_ID,PROV_ID FROM zb_d_bidwmb05009_010}' USING com.xxxx.dataplatform.bbdp.geniuspig.VerticaLoader('oracle','192.168.6.5','dev','1522','vbap','vbap','1') AS (MONTH_ID:chararray,DAY_ID:chararray,PROV_ID:chararray);

Table = load ‘url’ as (id,name…..); //table和load之间除了等号外还必须有个空格不然会出错，url一定要带引号，且只能是单引号。

2：filter

alias = FILTER alias BY expression;

Table = filter Table1 by + A； //A可以是 id > 10;not name matches ‘’，is not null 等，可以用and 和or连接各条件

例如：

filter = filter load20 by ( MONTH_ID == '1210' and DAY_ID == '18' and PROV_ID == '010' );

3：group

alias = GROUP alias { ALL | BY expression} [, alias ALL | BY expression …] [USING 'collected' | 'merge'] [PARTITION BY partitioner] [PARALLEL n];

pig的分组，不仅是数据上的分组，在数据的schema形式上也进行分组为groupcolumn:bag

Table3 = group Table2 by id；也可以Table3 = group Table2 by （id，name）；括号必须加

可以使用ALL实现对所有字段的分组

4：foreach

alias = FOREACH alias GENERATE expression [AS schema] [expression [AS schema]….];

alias = FOREACH nested_alias {

alias = {nested_op | nested_exp}; [{alias = {nested_op | nested_exp}; …]

GENERATE expression [AS schema] [expression [AS schema]….]

};

一般跟generate一块使用

Table = foreach Table generate （id，name）；括号可加可不加。

avg = foreach Table generate group, AVG(age); MAX ,MIN..

在进行数据过滤时，建议尽早使用foreach generate将多余的数据过滤掉，减少数据交换

5：join

Inner join Syntax

alias = JOIN alias BY {expression|'('expression [, expression …]')'} (, alias BY {expression|'('expression [, expression …]')'} …) [USING 'replicated' | 'skewed' | 'merge' | 'merge-sparse'] [PARTITION BY partitioner] [PARALLEL n];

Outer join Syntax

alias = JOIN left-alias BY left-alias-column [LEFT|RIGHT|FULL] [OUTER], right-alias BY right-alias-column [USING 'replicated' | 'skewed' | 'merge'] [PARTITION BY partitioner] [PARALLEL n];

join/left join / right join

daily = load 'A' as (id，name, sex);

divs = load 'B' as (id，name, sex);

join

jnd = join daily by (id, name), divs by (id, name);

left join

jnd = join daily by (id, name) left outer, divs by (id, name);

也可以同时多个变量，但只用于inner join

A = load 'input1' as (x, y);

B = load 'input2' as (u, v);

C = load 'input3' as (e, f);

alpha = join A by x, B by u, C by e;

6: union

alias = UNION [ONSCHEMA] alias, alias [, alias …];

union 相当与sql中的union，但与sql不通的是pig中的union可以针对两个不同模式的变量：如果两个变量模式相同，那么union后的变量模式与变量的模式一样；如果一个变量的模式可以由另一各变量的模式强制类型转换，那么union后的变量模式与转换后的变量模式相同；否则，union后的变量没有模式。

A = load 'input1' as (x:int, y:float);

B = load 'input2' as (x:int, y:float);

C = union A, B;

describe C;

C: {x: int,y: float}

A = load 'input1' as (x:double, y:float);

B = load 'input2' as (x:int, y:double);

C = union A, B;

describe C;

C: {x: double,y: double}

A = load 'input1' as (x:int, y:float);

B = load 'input2' as (x:int, y:chararray);

C = union A, B;

describe C;

Schema for C unknown.

注意：在pig 1.0中执行不了最后一种union。

如果需要对两个具有不通列名的变量union的话，可以使用onschema关键字

A = load 'input1' as (w: chararray, x:int, y:float);

B = load 'input2' as (x:int, y:double, z:chararray);

C = union onschema A, B;

describe C;

C: {w: chararray,x: int,y: double,z: chararray}

join和union之后alias的别名会变

7：Dump

dump alias

用于在屏幕上显示数据。

8：Order by

alias = ORDER alias BY { * [ASC|DESC] | field_alias [ASC|DESC] [, field_alias [ASC|DESC] …] } [PARALLEL n];

A = order Table by id desc;

9：distinct

A = distinct alias;

10：limit

A = limit alias 10;

11：sample

SAMPLE alias size;

随机抽取指定比例（0到1）的数据。

some = sample divs 0.1;

13：cross

alias = CROSS alias, alias [, alias …] [PARTITION BY partitioner] [PARALLEL n];

将多个数据集中的数据按照字段名进行同值组合，形成笛卡尔积。

--cross.pig

daily = load 'NYSE_daily' as (exchange:chararray, symbol:chararray,date:chararray, open:float, high:float, low:float,

close:float, volume:int, adj_close:float);

divs = load 'NYSE_dividends' as (exchange:chararray, symbol:chararray,date:chararray, dividends:float);

tonsodata = cross daily, divs parallel 10;

15：split

Syntax

SPLIT alias INTO alias IF expression, alias IF expression [, alias IF expression …] [, alias OTHERWISE];

A = LOAD 'data' AS (f1:int,f2:int,f3:int);

DUMP A;

(1,2,3)

(4,5,6)

(7,8,9)

SPLIT A INTO X IF f1<7, Y IF f2==5, Z IF (f3<6 OR f3>6);

DUMP X;

(1,2,3)

(4,5,6)

DUMP Y;

(4,5,6)

DUMP Z;

(1,2,3)

(7,8,9)

16：store

Store … into … Using…

pig在别名维护上：

1、join

如e = join d by name,b by name;

g = foreach e generate $0 as one:chararray, $1 as two:int, $2 as three:chararray,$3 asfour:int;

他生成的schemal：

e: {d::name: chararray,d::position: int,b::name: chararray,b::age: int}

g: {one: chararray,two: int,three: chararray,four: int}

2、group

B = GROUP A BY age;

----------------------------------------------------------------------

| B     | group: int | A: bag({name: chararray,age: int,gpa: float}) |

----------------------------------------------------------------------

|       | 18         | {(John, 18, 4.0), (Joe, 18, 3.8)}             |

|       | 20         | {(Bill, 20, 3.9)}                             |

----------------------------------------------------------------------

(18,{(John,18,4.0F),(Joe,18,3.8F)})

pig udf自定义

pig支持嵌入user defined function，一个简单的udf 继承于evalFunc，通常用在filter，foreach中

public class MyUDF extends EvalFunc<String> {
@Override
public String exec(Tuple input) throws IOException {
if(input == null || input.size() ==0)
return null;
try {
String val = (String) input.get(0);
return new StringBuffer(val).append(" pig").toString();
} catch (Exception e) {
throw new IOException(e.getMessage());
}
}
}

pig支持udf in loader and store

udf loader 需要继承于LoadFunc

udf storer 需要继承于StoreFunc

这类似于hadoop中写inputformat和outputformat

其中vertica就是写了一个DB版本的

这里贴一个简单的loader的例子：

public class MyLoader extends LoadFunc{
protected RecordReader recordReader = null;
private PreparedStatement ps;
private Connection conn;
private final String jdbcURL;
private final String user;
private final String pwd;
private final String querySql;
private ResultSet rs;
public MyLoader(String driver,String jdbcURL,String user,String pwd,String querySql){
try {
Class.forName(driver);
} catch (Exception e) {
// TODO: handle exception
}
this.jdbcURL = jdbcURL;
this.user = user;
this.pwd = pwd;
this.querySql = querySql;
}
@Override
public InputFormat getInputFormat() throws IOException {
return new PigTextInputFormat();
}
@Override
public Tuple getNext() throws IOException {
// TODO 重要的读取过程
Text val = null;
boolean next = false;
try {
next = rs.next();
} catch (Exception e) {
// TODO: handle exception
}
if(!next)
return null;
ResultSetMetaData rsmd;
try {
// rsmd = result
} catch (Exception e) {
// TODO: handle exception
}
return null;
}
@Override
public void prepareToRead(RecordReader arg0, PigSplit arg1)
throws IOException {
this.recordReader = arg0;
}
@Override
public void setLocation(String arg0, Job arg1) throws IOException {
//no idea
}
public ResourceSchema getSchema(String location,Job job) throws IOException{
Configuration conf = job.getConfiguration();
Schema schema = new Schema();
try {
//TODO:reader from database table
// Connection conn = DriverManager.getConnection(this.jdbcURL, this.user, this.pwd);
FieldSchema fieldName = new FieldSchema("name", DataType.CHARARRAY);
FieldSchema fieldPosition = new FieldSchema("position", DataType.INTEGER);
schema.add(fieldName);
schema.add(fieldPosition);
} catch (Exception e) {
//TODO log exception
}
return null;
}
public void prepareToRead(){
}
}

其中getNext方法就是如何处理reader读取出的数据

getSchema可以固定读取数据的schema

setLocation可以处理输入的数据源

prepareToRead是读取数据之前，可以在此做标识，等等

pig 衍生

1.penny：

1. Penny的描述

Penny是pig的贡献项目，是pig的调试和监控工具，而且支持根据API自定义penny的监视器和协作器，已实现不同的功能；

2. Penny的总架构

Penny将监视器插入到pig的工作操作中，主要用于监视pig数据流的变化，监视器可以调用协作器，完成各种功能。

3. Penny的总类图关系

ParsePigScript负责根据用户监视器生成新计划newPlan，在ToolsPigServer中根据以前的脚本执行新计划。在执行新计划时，当监视器监视对象数据发生变化，出发监视器，运行自定义的业务，也可以将数据流变化传回协作器里处理，总类图如下：

4. Penny的使用

Penny的使用需要自定义两个类，一个类继承于监视器基类MonitorAgent，另一个继承于协作器基类Coordinator。然后根据上边类图，就可以使用PennyServer和ParsePigScript进行监控和调试

5.在pig中就可以找到penny这个贡献的源码

Vertica：

vertica是pig loader和storer的udf

附件里是vertica，来自github，和vertica的介绍使用文档

贴一篇将vertica的帖子 http://blackproof.iteye.com/blog/1791995

推荐书籍

programming pig

推荐网址

http://pig.apache.org/docs/r0.10.0/basic.html 官网

pig pen开发工具，这个我现在玩得还不熟，就不介绍了，有兴趣的可以去搜搜玩玩

我在工作中pig的使用，主要是数据的ETL，所以比较适合。在选择pig hive还是其他非hadoop架构，如redis，这还是一个需要继续尝试探索的问题。