redis 的bitmap 開源包 bitmapist的應用

原理：http://my.oschina.net/u/1458120/blog/545224

bitmap包：https://github.com/Doist/bitmapist

要求redis>= 2.6.0 

bitmapist-master.zip  安裝本身下載解壓 python setup.py install

（因爲包內默認鏈接本機redis 6379 端口 ，若是想要修改 把下載的包放到本身的項目中 修改__init__.py中

SYSTEMS = {
    'default': redis.Redis(host='192.168.1.3', port=6379)
}

）

不能運行在集羣模式下，使用時最好keys *查看存過哪些key  設置過時時間
 

事件標記默認天周月

只標記天可修改__init__.py 中_mark()函數

# obj_classes = [MonthEvents, WeekEvents, DayEvents]

# obj_classes = [DayEvents]

推薦閱讀：

http://www.zhihu.com/question/21581696 

（cohort analysis 同期羣分析）感受bitmapist的 cohort 講的就是這 

聲明：本文的測試大部分都是 開源中的原測試加上本身的一些理解，讀者能夠直接看源文

# -*- coding: utf-8 -*-
# from builtins import range
import os
import traceback
import subprocess
import atexit
import socket
import time
import pytest
from bitmapist.cohort import get_dates_data

from bitmapist import setup_redis, delete_all_events

from datetime import datetime, timedelta

from bitmapist import mark_event, unmark_event,\
                      MonthEvents, WeekEvents, DayEvents, HourEvents,\
                      BitOpAnd, BitOpOr, get_event_names
from bitmapist.cohort import get_dates_data
##################################################################
#下面是啓動redis服務 若是已經啓動可忽略

def redis_server():
    """
    Fixture starting the Redis server
    """
    redis_host = '192.168.15.100'
    redis_port = 6379
    if is_socket_open(redis_host, redis_port):
        yield None
    else:
        proc = start_redis_server(redis_port)
        wait_for_socket(redis_host, redis_port)
        yield proc
        proc.terminate()
def setup_redis_for_bitmapist():
    setup_redis('default', '192.168.15.100', 6379)
    setup_redis('default_copy', 'l92.168.8.101', 6380)



def start_redis_server(port):
    """
    Helper function starting Redis server
    """
    devzero = open(os.devnull, 'r')
    devnull = open(os.devnull, 'w')
    #查看本身的redis-serverm命令
    proc = subprocess.Popen(['/usr/local/redis/bin/redis-server', '--port', str(port)],
                            stdin=devzero, stdout=devnull, stderr=devnull,
                            close_fds=True)#close_fds關閉子進程
    atexit.register(lambda: proc.terminate())
    return proc


def is_socket_open(host, port):
    """
    Helper function which tests is the socket open
    """
    sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
    sock.settimeout(0.1)
    return sock.connect_ex((host, port)) == 0


def wait_for_socket(host, port, seconds=3):
    """
    Check if socket is up for :param:`seconds` sec, raise an error otherwise
    """
    polling_interval = 0.1
    iterations = int(seconds / polling_interval)

    sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
    sock.settimeout(0.1)
    for _ in range(iterations):
        result = sock.connect_ex((host, port))
        if result == 0:
            sock.close()
            break
        time.sleep(polling_interval)
    else:
        raise RuntimeError('Service at %s:%d is unreachable' % (host, port))

########################################################################################
def base_test():
    try:
        setup_redis_for_bitmapist()
        redis_server()
        #mark_event的兩個重要參數now=None, track_hourly=None,
        # track_hourly =True標記用戶小時活躍  默認只標記到月周天
        #now=None時指如今時間   不然 指定到特定的用戶活躍時間
        mark_event('active', 123, track_hourly=True)
        now = datetime.utcnow()
        unmark_event('active', 124, track_hourly=True)#取消標記
        print MonthEvents('active', now.year, now.month).has_events_marked()
        print set(get_event_names(batch=2))#獲取全部的事件名稱set(["active"])
        print set(get_event_names(prefix='b', batch=2))#獲取全部以b開頭的事件名稱set（）set([])
        # Month
        assert 123 in MonthEvents('active', now.year, now.month)
        assert 124 not in MonthEvents('active', now.year, now.month)
        # Week
        assert 123 in WeekEvents('active', now.year, now.isocalendar()[1])
        assert 124 not in WeekEvents('active', now.year, now.isocalendar()[1])
        # Day
        assert 123 in DayEvents('active', now.year, now.month, now.day)
        assert 124 not in DayEvents('active', now.year, now.month, now.day)
        # Hour
        assert 123 in HourEvents('active', now.year, now.month, now.day, now.hour)
        assert 124 not in HourEvents('active', now.year, now.month, now.day, now.hour)
        assert 124 not in HourEvents('active', now.year, now.month, now.day, now.hour-1)
        mark_event("active",124)
        #標記用戶125本月活躍
        mark_event('active', 125)
        assert 125 in MonthEvents('active', now.year, now.month)
        #取消用戶125本月活躍
        unmark_event('active', 125)
        assert 125 not in MonthEvents('active', now.year, now.month)
        yesterday = now - timedelta(days=1)
        mark_event('active', 126, now=now)
        mark_event('active', 127, now=yesterday)#把用戶活躍標記到昨天的月周天上
        #獲取活躍的用戶數 提示錯誤沒有bitcount命令 要求redis>= 2.6.0  info查看redis版本
        print MonthEvents('active', now.year, now.month).get_count()#4
        print list(WeekEvents('active', now.year, now.isocalendar()[1]))# [123, 124, 126, 127]
        print list(MonthEvents('active', now.year, now.month))#[123, 124, 126, 127]
        ev = DayEvents('active', now.year, now.month,now.day) # [123, 124, 126]
        print list(ev)#3
        print len(ev)#3
        print list(DayEvents('active',yesterday.year,yesterday.month,yesterday.day))#[127]


        last_month = datetime.utcnow() - timedelta(days=30)
        ago3_month = datetime.utcnow() - timedelta(days=60)
        # 123，127 128上月活躍，127 上上月活躍
        mark_event('active', 127, now=ago3_month)
        mark_event('active', 123, now=last_month)
        mark_event('active', 127, now=last_month)
        mark_event('active', 128, now=last_month)
        # 近三個月都活躍的用戶(交集)
        active_3_months = BitOpAnd(
            MonthEvents('active', ago3_month.year, ago3_month.month),
            MonthEvents('active', last_month.year, last_month.month),
            MonthEvents('active', now.year, now.month)
        )
        print list(active_3_months)#[127]
        active_3_months.delete()

        #嵌套查詢
        active_3_months = BitOpAnd(
            MonthEvents('active', ago3_month.year, ago3_month.month),
            MonthEvents('active', last_month.year, last_month.month),
            BitOpAnd(
            MonthEvents('active', now.year, now.month))
        )
        print list(active_3_months)#[127]

        # 近三個月活躍過的用戶（並集)
        print list(BitOpOr(
            MonthEvents('active', ago3_month.year, ago3_month.month),
            MonthEvents('active', last_month.year, last_month.month),
            MonthEvents('active', now.year, now.month)))#[123, 124, 126, 127, 128]

        #用備份redis查詢
        active_2_months = BitOpAnd(
            'default_copy',
            MonthEvents('active', last_month.year, last_month.month),
            MonthEvents('active', now.year, now.month)
        )
        print list(active_2_months)
        active_2_months.delete()
        delete_all_events()#清理全部的bitmap

    except:
        traceback.print_exc()




def test_bit_operations_complex():
    now = datetime.utcnow()
    tom = now + timedelta(days=1)

    mark_event('task1', 111, now=now)
    mark_event('task1', 111, now=tom)
    mark_event('task2', 111, now=now)
    mark_event('task2', 111, now=tom)
    mark_event('task1', 222, now=now)
    mark_event('task1', 222, now=tom)
    mark_event('task2', 222, now=now)
    mark_event('task2', 222, now=tom)

    now_events = BitOpAnd(
        DayEvents('task1', now.year, now.month, now.day),
        DayEvents('task2', now.year, now.month, now.day)
    )

    tom_events = BitOpAnd(
        DayEvents('task1', tom.year, tom.month, tom.day),
        DayEvents('task2', tom.year, tom.month, tom.day)
    )

    both_events = BitOpAnd(now_events, tom_events)
    print list(both_events)


def test_bitop_key_sharing():
    #測試key共享
    today = datetime.utcnow()
    #假設task1爲唱歌 task2爲跳舞
    mark_event('task1', 111, now=today)
    mark_event('task2', 111, now=today)
    mark_event('task1', 222, now=today)
    mark_event('task2', 222, now=today)

    ev1_task1 = DayEvents('task1', today.year, today.month, today.day)
    ev1_task2 = DayEvents('task2', today.year, today.month, today.day)
    ev1_both = BitOpAnd(ev1_task1, ev1_task2)#今天即唱歌又跳舞的

    ev2_task1 = DayEvents('task1', today.year, today.month, today.day)
    ev2_task2 = DayEvents('task2', today.year, today.month, today.day)
    ev2_both = BitOpAnd(ev2_task1, ev2_task2)
    print ev1_both,ev2_both
    print ev1_both.redis_key,ev2_both.redis_key
    print  len(ev1_both), len(ev2_both)#2,2
    ev1_both.delete()
    print  len(ev1_both), len(ev2_both)#0,0 同一個查詢的實例相同





def test_bit_operations_magic():
    delete_all_events()
    mark_event('foo', 1)
    mark_event('foo', 2)
    mark_event('bar', 2)
    mark_event('bar', 3)
    foo = DayEvents('foo')
    bar = DayEvents('bar')
    print list(foo & bar)#交集
    print  list(foo | bar)#並集
    print  list(foo ^ bar)#異或 （foo和bar不相同的）
    print  list(~foo & bar)
    #[2]
    #[1, 2, 3]
    #[1, 3]
    #[3]



def test_cohort():
    '''測試場景：當天註冊用戶 三天內天天有多少活躍過'''
    today = datetime.utcnow()#20160608
    yes = today - timedelta(days=1)
    ago2 = today - timedelta(days=2)#20160606
    tomorrow = today +timedelta(days=1)
    after2 = today +timedelta(days=2)
    ####################################前天註冊用戶追蹤
    #前天註冊
    mark_event("regist",111,now=ago2)
    mark_event("regist",112,now=ago2)
    mark_event("regist",113,now=ago2)
    mark_event("regist",114,now=ago2)
    mark_event("regist",115,now=ago2)
    #前天活躍
    mark_event("active",111,now=ago2)
    mark_event("active",112,now=ago2)
    mark_event("active",113,now=ago2)
    mark_event("active",114,now=ago2)
    #昨天活躍的
    mark_event("active",111,now=yes)
    mark_event("active",112,now=yes)
    mark_event("active",113,now=yes)
    #今天活躍的
    mark_event("active",115,now=today)

    ############################昨天註冊用戶追蹤
    #昨天註冊
    mark_event("regist",116,now=yes)
    mark_event("regist",117,now=yes)
    mark_event("regist",118,now=yes)
    mark_event("regist",119,now=yes)
    mark_event("regist",120,now=yes)
    #昨天活躍
    mark_event("active",116,now=yes)
    mark_event("active",117,now=yes)
    mark_event("active",118,now=yes)
    mark_event("active",119,now=yes)
    #今天活躍
    mark_event("active",119,now=today)
    mark_event("active",120,now=today)
    #明天活躍
    mark_event("active",116,now=tomorrow)

########################################################今天註冊用戶追蹤
    #今天註冊
    mark_event("regist",121,now=today)
    mark_event("regist",122,now=today)
    mark_event("regist",123,now=today)
    #今天活躍
    mark_event("active",121,now=today)
    mark_event("active",122,now=today)
    #明天活躍
    mark_event("active",121,now=tomorrow)
    #後天活躍
    mark_event("active",123,now=after2)


    for select1,select1b,select2,select2b in [('regist', None, 'active', None)]:
        '''select1:初始條件 本例指註冊
        select1b:select1的附屬條件 例如中國註冊用戶（mark_event("registchina",123,now=today)）
        select2:初始條件下的過濾
        select2b:同select1b
        time_group:時間跨度 `days`, `weeks`, `months`, `years`
        as_precent：0輸出通過select過濾後的數量  1百分比
        num_results：time_group+num_results 獲得關注的時間範圍  如近三天、三週、三月 三年
        num_of_rows：對select1（或select1+select1b）條件下的select2（或select2+select2b）
                     輸出將來幾天（周/月/年）的
        本例：近三天（time_group+num_results=3days）當天註冊用戶(select1='regist')
              三天內（num_of_rows=3）天天的活躍（select2='active'）人數
        '''
        r = get_dates_data(select1=select1, select1b=select1b,
                       select2=select2, select2b=select2b,
                       time_group='days', as_precent=0,
                       num_results=3, num_of_rows=3)
        print list(r)
                                                     #當天註冊   #當天活躍  #將來三天活躍
# [[datetime.datetime(2016, 6, 6, 6, 21, 43, 845437), 5,               4,     3, 1, 0.0],
#  [datetime.datetime(2016, 6, 7, 6, 21, 43, 845437), 5,               4,     2, 1, 0.0],
#  [datetime.datetime(2016, 6, 8, 6, 21, 43, 845437), 3,               2,     1, 1, '']]
#以6號爲例  20160606 當天註冊5人 4人活躍  將來三天20160607 在20160606註冊的人中其中3人活躍 20160608 1人活躍 20160609 0人活躍(0.0表明有人活躍但不在20160606的註冊人中 '' 表明壓根沒人活躍）
#redis中查找key：
#:6379> BITCOUNT trackist_regist_2016-6-7
#6379> BITCOUNT trackist_regist_W2016-27
#6379> BITCOUNT trackist_buyid_regist_2016-6
if __name__ == '__main__':
    test_bitop_key_sharing()
    delete_all_events()
    test_bit_operations_magic()
    delete_all_events()
    test_cohort()