PySpark/hbase.py at master · LinYuMingBejing/PySpark · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
from pyspark.sql.context import SQLContext
from pyspark import SparkContext
from pyspark.sql.functions import lit, when, col

from datetime import datetime, timedelta
from argparse import ArgumentParser

parser = ArgumentParser()
parser.add_argument('-start', help='--start_date', required=True)
parser.add_argument('-path', help='--output_directory', required=True)
args = parser.parse_args()

sc = SparkContext()
sqlContext = SQLContext(sc)

hbase_prefixes = ['a', 'b', 'c', 'd', 'e', 'f', 'g']
filtered_referrers = ['www.google.com', 'baidu.com']
datetime_format = '%Y-%m-%d'


def get_hbase_rowkey(start, end):
    rowkey = []
    for prefix in hbase_prefixes:
        row = {
            'start': f'{prefix}_{start.strftime(datetime_format)}',
            'end': f'{prefix}_{end.strftime(datetime_format)}'
        }
        rowkey.append(row)
    return rowkey


def pull_data(table, rowkey):
    df_all = None
    for key in rowkey:
        tmp_df = sqlContext.sql("select * from {} where key > '{}' and key < '{}' ".format(table, key['start'], key['end']))
        if df_all:
            df_all = df_all.union(tmp_df)
        else:
            df_all = tmp_df

    df_all = df_all.na.fill('')

    return df_all


def load_footprint_data(rowkey):
    footprint_catalog = ''.join("""{
                "table": {"namespace":"default", "name":"footprint"},
                "rowkey": "key1",
                "columns": {
                    "key": {"cf":"rowkey", "col":"key1", "type":"string", "length":"12"},
                    "ip": {"cf":"footprint","col":"ip","type":"string"},
                    "fp": {"cf":"footprint","col":"fp","type":"string"},
                    "url": {"cf":"footprint","col":"url","type":"string"},
                    "page_id": {"cf":"footprint","col":"page_id","type":"string"},
                    "referrer": {"cf":"footprint","col":"referrer","type":"string"}
                }
            }""".split())

    data = sqlContext.read.format('org.apache.spark.sql.execution.datasources.hbase') \
                    .options(catalog=footprint_catalog) \
                    .load()

    data.createOrReplaceTempView('footprint')
    footprint = pull_data('footprint', rowkey)
    footprint = footprint.filter(~footprint_pv['referrer'].isin(filtered_referrers)) \
                        .groupby('page_id', 'url', 'fp', 'ip')\
                        .count()\

    return footprint


def load_session_stay_data(rowkey):
    session_catalog = ''.join("""{
                "table": {"namespace":"default", "name":"session_stay"},
                "rowkey": "key1",
                "columns": {
                    "key": {"cf":"rowkey", "col":"key1", "type":"string", "length":"12"},
                    "ip": {"cf":"session_stay","col":"ip","type":"string"},
                    "fp": {"cf":"session_stay","col":"fp","type":"string"},
                    "page_id":{"cf":"session_stay","col":"page_id","type":"string"},
                    "session_stay(int)": {"cf":"session_stay","col":"session_stay(int)","type":"string"},
                }
            }""".split())

    data = sqlContext.read.format('org.apache.spark.sql.execution.datasources.hbase') \
            .options(catalog=session_catalog) \
            .load()

    data.createOrReplaceTempView('session_stay')
    session_stay = pull_data('session_stay', rowkey)

    session_stay = session_stay.filter((col('session_stay(int)') <= 1)) \
                            .groupby('page_id', 'ip', 'fp', 'session_stay(int)')\
                            .count()

    session_stay = session_stay.withColumn('stay_0_count',
                                when((session_stay['session_stay(int)'] == 1), lit(0))\
                                .otherwise(session_stay['count'])) \
                            .withColumn('stay_1_count', \
                                when((session_stay['session_stay(int)'] == 0), lit(0))\
                                .otherwise(session_stay['count']))\
                            .groupBy('page_id', 'ip', 'fp')\
                            .agg({'stay_0_count': 'sum', 'stay_1_count': 'sum'}) \
                            .withColumnRenamed('sum(stay_0_count)', 'stay_0_count') \
                            .withColumnRenamed('sum(stay_1_count)', 'stay_1_count')

    session_stay = session_stay.drop('count', 'session_stay(int)')

    return session_stay


def calculate(footprint, session_stay):
    df = footprint.join(session_stay, ['page_id', 'ip', 'fp'], how='left').fillna(0)
    df = df.withColumn('new_count',
                            when(((df['count'] - df['stay_0_count'] - df['stay_1_count']) > 0), (df['count'] - df['stay_0_count'] - df['stay_1_count']))\
                            .otherwise(lit(0)))

    df = df.withColumn('valid_pv', when((df['new_count'] > 3), lit(1)).otherwise(df['new_count'])) \
    df = df.withColumn('invalid_pv', when((df['new_count'] > 3), df['new_count'] - 1).otherwise(lit(0)))

    df = df.groupBy('page_id')\
                    .agg({
                        'url': 'last',
                        'count': 'sum',
                        'valid_pv': 'sum',
                        'invalid_pv': 'sum',
                        'new_count': 'sum'
                    })

    df = df.withColumn('datetime', lit(args.start_date)) \
            .withColumn('url', df['last(url)']) \
            .withColumn('pv', df['sum(new_count)'].cast('int')) \
            .withColumn('pv_valid', df['sum(valid_pv)'].cast('int')) \
            .withColumn('pv_invalid', df['sum(invalid_pv)'].cast('int'))

    df = df.select('datetime','page_id', 'url', 'pv', 'pv_valid', 'pv_invalid')

    return df


if __name__ == '__main__':
    # get_hbase_rowkey
    start = datetime.strptime(args.start_date, datetime_format)
    end = start + timedelta(days=1)
    rowkey = get_hbase_rowkey(start, end)

    # load data
    footprint_pv = load_footprint_data(rowkey)
    session_pv = load_session_stay_data(rowkey)

    # calculate pageviews
    result = calculate(footprint_pv, session_pv)

    # store
    result.repartition(1).write.format('com.databricks.spark.avro')\
            .save(args.output_directory)