HBase Python API

HBase Python API

HBase通过thrift机制可以实现多语言编程，信息通过端口传递，因此Python是个不错的选择

吐槽

博主在Mac上配置HBase，奈何Zoomkeeper一直报错，结果Ubuntu虚拟机上10min解决……但是虚拟机里没有IDE写Java代码还是不方便，因此用Mac主机连接虚拟机的想法孕育而生，这样又可以愉快地使用主机的IDE了~

一、服务端启动Hbase Thrift RPC

HBase的启动方式有很多，这里不再赘述，Ubuntu启动HBase之后，启动thrift

hbase-daemon.sh start thrift

默认的服务端口是9090

二、客户端安装依赖包

sudo pip install thrift
sudo pip install hbase-thrift

三、编写客户端代码

# coding=utf-8
from thrift.transport import TSocket
from thrift.transport.TTransport import TBufferedTransport
from thrift.protocol import TBinaryProtocol

from hbase import Hbase
from hbase.ttypes import ColumnDescriptor
from hbase.ttypes import Mutation

class HBaseClient(object):
    def __init__(self, ip, port=9090):
        """
        建立与thrift server端的连接
        """
        # server端地址和端口设定
        self.__transport = TBufferedTransport(TSocket.TSocket(ip, port))
        # 设置传输协议
        protocol = TBinaryProtocol.TBinaryProtocol(self.__transport)
        # 客户端
        self.__client = Hbase.Client(protocol)
        # 打开连接
        self.__transport.open()

    def __del__(self):
        self.__transport.close()

    def get_tables(self):
        """
        获得所有表
        :return:表名列表
        """
        return self.__client.getTableNames()

    def create_table(self, table, *columns):
        """
        创建表格
        :param table:表名
        :param columns:列族名
        """
        func = lambda col: ColumnDescriptor(col)
        column_families = map(func, columns)
        self.__client.createTable(table, column_families)

    def put(self, table, row, columns):
        """
        添加记录
        :param table:表名
        :param row:行键
        :param columns:列名
        :return:
        """
        func = lambda (k, v): Mutation(column=k, value=v)
        mutations = map(func, columns.items())
        self.__client.mutateRow(table, row, mutations)

    def delete(self, table, row, column):
        """
        删除记录
        :param table:表名
        :param row:行键
        """
        self.__client.deleteAll(table, row, column)

    def scan(self, table, start_row="", columns=None):
        """
        获得记录
        :param table: 表名
        :param start_row: 起始行
        :param columns: 列族
        :param attributes:
        """
        scanner = self.__client.scannerOpen(table, start_row, columns)
        func = lambda (k, v): (k, v.value)
        while True:
            r = self.__client.scannerGet(scanner)
            if not r:
                break
            yield dict(map(func, r[0].columns.items()))

if __name__ == '__main__':
    client = HBaseClient("10.211.55.7")

    # client.create_table('student', 'name', 'course')
    client.put("student", "1",
               {"name:": "Jack",
                "course:art": "88",
                "course:math": "12"})

    client.put("student", "2",
               {"name:": "Tom", "course:art": "90",
                "course:math": "100"})

    client.put("student", "3",
               {"name:": "Jerry"})
    client.delete('student', '1', 'course:math')
    for v in client.scan('student'):
        print v

四、测试结果

{'course:art': '88', 'name:': 'Jack'}
{'course:art': '90', 'name:': 'Tom', 'course:math': '100'}
{'name:': 'Jerry'}

五、小结

有了Python接口后，编写简单任务脚本变得非常方便，这大大得益于RPC机制，很好地解耦了Client和Server，方便开发人员合作。