sqoop源码编译

== 构建前提


如果要对Sqoop进行编译，需要依赖以下工具（将其加入环境变量）:


* Apache ant (1.7.1)
* Java JDK 1.6


另外，如果想生成帮助文档，还需要:


* asciidoc
* make
* python 2.5+
* xmlto
* tar
* gzip


并且, Sqoop还引入了以下工具:


* findbugs (1.3.9) 用于代码质量检查
* cobertura (1.9.4.1) 用于代码覆盖检查
* checkstyle (5.x) 用于代码风格检查


== 基础


Sqoop通过ant编译. 进入$SQOOP_HOME后运行“ant -p”查看所有可用targets.


Type +ant+ to compile all java sources. You can then run Sqoop with +bin/sqoop+.


If you want to build everything (including the documentation), type
+ant package+. This will appear in the
+build/sqoop-(version)/+ directory.




Sqoop版本以Hadoop 0.23作为默认依赖，可在maven仓库中查到. These dependencies are obtained via IVY which
downloads the necessary binaries.


== 测试Sqoop


Sqoop has several unit tests which can be run with +ant test+. This command
will run all the "basic" checks against an in-memory database, HSQLDB.


Sqoop also has compatibility tests that check its ability to work with
several third-party databases. To enable these tests, you will need to install
and configure the databases, and download the JDBC drivers for each one.


=== MySQL


Install MySQL server and client 5.0. Download MySQL Connector/J 5.0.8 for
JDBC. Instructions for configuring the MySQL database are in MySQLAuthTest
and DirectMySQLTest.


Use the system property sqoop.test.mysql.connectstring.host_url to specify the
URL for the MySQL host used for testing. Specify this property on the command
line or via the build.properties file. For example:


sqoop.test.mysql.connectstring.host_url=jdbc:mysql://host.example.com/


If not specified, the default value used for this property is:
jdbc:mysql://localhost/


=== Oracle


Install Oracle Enterprise Edition 10.2.0+. Instructions for configuring the
database are in OracleManagerTest. Download the ojdbc6_g jar.


If running the tests against Oracle XE (Express Edition) - a lot of them will
fail as it does not include the partitioning feature.


Use the system property sqoop.test.oracle.connectstring to specify the
connection string for Oracle host used for testing. Specify this property on the
command line or via the build.properties file. For example:


sqoop.test.oracle.connectstring=jdbc:oracle:thin:@//host.example.com/xe


If not specified, the default value used for this property is:
jdbc:oracle:thin:@//localhost/xe


Users sqooptest and sqooptest2 should be created prior to running the tests.
SQL script is available in src/test/oraoop/create_users.sql


=== PostgreSQL


Install PostgreSQL 8.3.9. Download the postgresql 8.4 jdbc driver. Instructions
for configuring the database are in PostgresqlTest.


Use the system property sqoop.test.postgresql.connectstring.host_url to specify
the URL for the PostgreSQL host used for testing. Specify this property on the
command line or via the build.properties file. For example:


sqoop.test.postgresql.connectstring.host_url=jdbc:postgresql://sqoop-dbs.sf.cloudera.com/


If not specified, the default value used for this property is:
jdbc:postgresql://localhost/


=== SQL Server


Install SQL Server Express 2008 R2 and create a database instance and
download the appropriate JDBC driver. Instructions for configuring the
database can be found in SQLServerManagerImportManualTest.


Use the system property sqoop.test.sqlserver.connectstring.host_url to specify
the URL for the SQL Server host used for testing. Specify this property on the
command line or via the build.properties file. For example:


sqoop.test.sqlserver.connectstring.host_url=jdbc:sqlserver://sqlserverhost:1433


If not specified, the default value used for this property is:
jdbc:sqlserver://sqlserverhost:1433


This can be useful if you have the hostname sqlserverhost mapped to the IP
address of the SQL Server instance.


=== Cubrid


Install Cubrid 9.2.2.0003 and create a database instance and download the
appropriate JDBC driver. Instructions for configuring the database are in
CubridAuthTest, CubridCompatTest, CubridManagerImportTest
and CubridManagerExportTest.


Use the system property sqoop.test.cubrid.connectstring.host_url to specify the
URL for the Cubrid host used for testing. Specify this property on the command
line or via the build.properties file. For example:


sqoop.test.cubrid.connectstring.host_url=jdbc:cubrid:localhost


If not specified, the default value used for this property is:
jdbc:cubrid:localhost


=== DB2


Install DB2 9.74 Express C and download the appropriate JDBC driver.
Instructions for configuring the server can be found in
DB2ManagerImportManualTest.


Use the system property sqoop.test.db2.connectstring.host_url to specify
the URL for the DB Server host used for testing. Specify this property on
the command line or via build.properties file. For example:


sqoop.test.db2.connectstring.host_url=jdbc:db2://db2host:50000


If not specified, the default value used for this property is:
jdbc:db2://db2host:50000


This can be useful if you have the hostname db2host mapped to the IP
address of the DB2 Server instance.


=== 运行第三方测试


After the third-party databases are installed and configured, run:


++++
ant test -Dthirdparty=true -Dsqoop.thirdparty.lib.dir=/path/to/jdbc/drivers/
++++


This command will run all thirdparty tests except for the SQL Server test.
To run the SQL Server test, specify the property "manual" instead of "thirdparty"
as follows:


++++
ant test -Dmanual=true -Dsqoop.thirdparty.lib.dir=/path/to/jdbc/drivers/
++++


Note that +sqoop.thirdparty.lib.dir+ can also be specified in
+build.properties+.


== 代码质量分析


We have three tools which can be used to analyze Sqoop's code quality.


=== Findbugs


Findbugs detects common errors in programming. New patches should not
trigger additional warnings in Findbugs.


Install findbugs (1.3.9) according to its instructions. To use it,
run:


++++
ant findbugs -Dfindbugs.home=/path/to/findbugs/
++++


A report will be generated in +build/findbugs/+


=== Cobertura


Cobertura runs code coverage checks. It instruments the build and
checks that each line and conditional expression is evaluated along
all possible paths.


Install Cobertura according to its instructions. Then run a test with:


++++
ant clean
ant cobertura -Dcobertura.home=/path/to/cobertura
ant cobertura -Dcobertura.home=/path/to/cobertura \
    -Dthirdparty=true -Dsqoop.thirdparty.lib.dir=/path/to/thirdparty
++++


(You'll need to run the cobertura target twice; once against the regular
test targets, and once against the thirdparty targets.)


When complete, the report will be placed in +build/cobertura/+


New patches should come with sufficient tests for their functionality
as well as their error recovery code paths. Cobertura can help assess
whether your tests are thorough enough, or where gaps lie.


=== Checkstyle


Checkstyle enforces our style guide. There are currently a very small
number of violations of this style in the source code, but hopefully this
will remain the case. New code should not trigger additional checkstyle
warnings.


Checkstyle does not need to be installed manually; it will be retrieved via
Ivy when necessary.


To run checkstyle, execute:


++++
ant checkstyle
++++


A report will be generated as +build/checkstyle-errors.html+




== Deploying to Maven


To use Sqoop as a dependency in other projects, you can pull Sqoop into your
dependency management system through Maven.


To install Sqoop in your local +.m2+ cache, run:


++++
ant mvn-install
++++


This will install a pom and the Sqoop jar.


To deploy Sqoop to a public repository, use:


++++
ant mvn-deploy
++++


By default, this deploys to repository.cloudera.com. You can choose
the complete URL to deploy to with the +mvn.deploy.url+ property.
By default, this deploys to the "snapshots" repository. To deploy to
"staging" or "releases" on repository.cloudera.com, set the
+mvn.repo+ property accordingly.（在building.xml中搜索到对应属性，修改值即可）


== Releasing Sqoop


To build a full release of Sqoop, run +ant release -Dversion=(somever)+.
This will build a binary release tarball and the web-based documentation
as well as run a release audit which flags any source files which may
be missing license headers.


(The release audit can be run standalone with the +ant releaseaudit+
target.)


You must set the +version+ property explicitly; you cannot release a
snapshot. To simultaneously deploy this to a maven repository, include
the +mvn-install+ or +mvn-deploy+ targets as well.




== Using Eclipse


Running +ant eclipse+ will generate +.project+ and +.classpath+ files that
will allow you to edit Sqoop sources in Eclipse with all the library
dependencies correctly resolved. To compile the jars, you should still
use ant.




== 定义依赖特定版本Hadoop


Now Sqoop defaults to use Hadoop 0.23 available from Apache maven repository.
To switch back to the previous version of Hadoop 0.20, for example, run:


++++
ant test -Dhadoopversion=20
++++