1. hbase连接

首先用hbase shell 命令来进入到hbase数据库，然后用list命令来查看hbase下所有表，以其中表“DB_level0”为例，可以看到库名“baotouyiqi”是拼接的，python代码访问时先连接：

def hbase_connection(hbase_master, hbase_port, table_prefix=None):
    connection = happybase.Connection(host=hbase_master, port=hbase_port, table_prefix=table_prefix)
    return connection
connection = hbase_connection(hbase_master, hbase_port, table_prefix)  # 在连接的时候创建项目空间
table = connection.table(tablename)  # 获取表连接

备注：完整代码在最后，想运行的直接滑倒最后复制即可

2. 按条件读取hbase数据

然后按照条件来查询表中想要的数据集，这里只列举两个条件：时间区间和指定列。同样，我们在shell下用scan命令来查看表中的数据结构：

可以看到第一列是ROW，第二列是COLUMN+CELL，python代码取数据方法差不多：

date_prex_start = bytes('dt_' + starttime, encoding='utf-8')  # row_start
date_prex_end = bytes('dt_' + endtime, encoding='utf-8')  # row_stop
# 通过设置row key的前缀row_prefix参数来进行局部扫描
outdata = dict(table.scan(row_start=date_prex_start, row_stop=date_prex_end,
                          columns=[onecolumn]))

得到的结果如下，是个字典格式：

3. 按格式输出hbase数据结果

我们希望输出的结果是dataframe的，而且第一列是time，第二列是value，所以就做个简单格式处理：

timesep = list(map(lambda x: x.decode('utf-8').replace('dt_', ''), outdata.keys()))
tempdata = list(outdata.values())
valuelist = list(map(lambda x: float(list(x.values())[0]), tempdata))
if len(timesep) > 0:
    db_data2 = pd.DataFrame({'时间': timesep, onecolumn: valuelist})
    db_data2.loc[:, '时间2'] = [i[:16] for i in db_data2['时间']]
    db_data2 = db_data2.drop_duplicates(subset=['时间2'], keep='last')  # 一分钟内多次数值取一个即可
else:
    db_data2 = pd.DataFrame()
if len(db_data2) < 1:
    return pd.DataFrame()
db_data2.loc[:, '时间戳'] = [time.mktime(time.strptime(i, "%Y-%m-%d %H:%M:%S")) for i in db_data2['时间']]
db_data2 = db_data2.sort_values(by=['时间戳'], ascending=False)  # 将最新的数值放最前面
db_data3 = db_data2.drop(columns=['时间2', '时间戳'])
db_data3.columns = ['time', 'value']

4. 完整代码（code）

import happybase
import time
import pandas as pd
from pathlib import Path

os_file_name = Path(__file__).name


def hbase_connection(hbase_master, hbase_port, table_prefix=None):
    connection = happybase.Connection(host=hbase_master, port=hbase_port, table_prefix=table_prefix)
    return connection


def get_data_by_tum(hbase_master, hbase_port, table_prefix, tablename, columnslist, starttime, endtime):
    columnsid = '$'.join(columnslist)
    onecolumn = 'TimeSe:dt_' + columnsid  # column

    connection = hbase_connection(hbase_master, hbase_port, table_prefix)  # 在连接的时候创建项目空间
    table = connection.table(tablename)  # 获取表连接
    date_prex_start = bytes('dt_' + starttime, encoding='utf-8')  # row_start
    date_prex_end = bytes('dt_' + endtime, encoding='utf-8')  # row_stop
    # 通过设置row key的前缀row_prefix参数来进行局部扫描
    outdata = dict(table.scan(row_start=date_prex_start, row_stop=date_prex_end,
                              columns=[onecolumn]))

    timesep = list(map(lambda x: x.decode('utf-8').replace('dt_', ''), outdata.keys()))
    tempdata = list(outdata.values())
    valuelist = list(map(lambda x: float(list(x.values())[0]), tempdata))
    if len(timesep) > 0:
        db_data2 = pd.DataFrame({'时间': timesep, onecolumn: valuelist})
        db_data2.loc[:, '时间2'] = [i[:16] for i in db_data2['时间']]
        db_data2 = db_data2.drop_duplicates(subset=['时间2'], keep='last')  # 一分钟内多次数值取一个即可
    else:
        db_data2 = pd.DataFrame()
    if len(db_data2) < 1:
        return pd.DataFrame()
    db_data2.loc[:, '时间戳'] = [time.mktime(time.strptime(i, "%Y-%m-%d %H:%M:%S")) for i in db_data2['时间']]
    db_data2 = db_data2.sort_values(by=['时间戳'], ascending=False)  # 将最新的数值放最前面
    db_data3 = db_data2.drop(columns=['时间2', '时间戳'])
    db_data3.columns = ['time', 'value']
    return db_data3


if __name__ == '__main__':
    begin_time = '2023-08-22 00:00:00'
    end_time = '2023-08-23 00:00:00'
    hbase_master = "142.21.8.22"
    hbase_port = 9097
    table_prefix = "baotouyiqi"
    table_name = "DB_level0"
    onedata = ["62340", "20", "204"]
    dataget = get_data_by_tum(hbase_master, hbase_port, table_prefix, table_name,
                              onedata, begin_time, end_time)
    print(dataget)