|
| 1 | +#!/usr/bin/env python |
| 2 | +# coding: utf-8 |
| 3 | + |
| 4 | +# In[26]: |
| 5 | + |
| 6 | + |
| 7 | + |
| 8 | + |
| 9 | + |
| 10 | +# In[27]: |
| 11 | + |
| 12 | + |
| 13 | +import logging |
| 14 | +logger = logging.getLogger() |
| 15 | +logger.setLevel(logging.INFO) #DEBUG, INFO, WARNING, ERROR, CRITICAL |
| 16 | + |
| 17 | + |
| 18 | + |
| 19 | + |
| 20 | +# In[28]: |
| 21 | + |
| 22 | + |
| 23 | +import json, pandas as pd |
| 24 | +from ProjectDomino.Neo4jDataAccess import Neo4jDataAccess |
| 25 | +from ProjectDomino.FirehoseJob import FirehoseJob |
| 26 | +from ProjectDomino.TwintPool import TwintPool |
| 27 | +from prefect.environments.storage import S3 |
| 28 | +from prefect import Flow,task |
| 29 | +from prefect.schedules import IntervalSchedule |
| 30 | +from datetime import timedelta, datetime |
| 31 | +from random import randrange |
| 32 | +from prefect.engine.executors import DaskExecutor |
| 33 | +import time |
| 34 | +import random |
| 35 | + |
| 36 | + |
| 37 | +# In[29]: |
| 38 | + |
| 39 | + |
| 40 | + |
| 41 | + |
| 42 | + |
| 43 | +# In[30]: |
| 44 | + |
| 45 | + |
| 46 | +S3_BUCKET = "wzy-project-domino" |
| 47 | + |
| 48 | + |
| 49 | +# In[31]: |
| 50 | + |
| 51 | + |
| 52 | +pd.set_option('display.max_colwidth', None) |
| 53 | +pd.set_option('display.max_rows', 500) |
| 54 | +pd.set_option('display.max_columns', 500) |
| 55 | +pd.set_option('display.width', 1000) |
| 56 | + |
| 57 | + |
| 58 | +# ## task |
| 59 | + |
| 60 | +# In[33]: |
| 61 | + |
| 62 | + |
| 63 | +def random_date(start, end): |
| 64 | + delta = end - start |
| 65 | + int_delta = (delta.days * 24 * 60 * 60) + delta.seconds |
| 66 | + random_second = randrange(int_delta) |
| 67 | + return start + timedelta(seconds=random_second) |
| 68 | + |
| 69 | +def get_creds(): |
| 70 | + neo4j_creds = None |
| 71 | + with open('/secrets/neo4jcreds.json') as json_file: |
| 72 | + neo4j_creds = json.load(json_file) |
| 73 | + return neo4j_creds |
| 74 | + |
| 75 | +@task(log_stdout=True, skip_on_upstream_skip=True) |
| 76 | +def run_stream(): |
| 77 | + creds = get_creds() |
| 78 | + start = datetime.strptime("2020-03-11 20:00:00", "%Y-%m-%d %H:%M:%S") |
| 79 | + current = datetime.strptime(datetime.now().strftime("%Y-%m-%d %H:%M:%S"), "%Y-%m-%d %H:%M:%S") |
| 80 | + #rand_dt=random_date(start, current) |
| 81 | + #2020-10-10 16:07:30 |
| 82 | + #2020-10-11 06:29:00 to 2020-10-11 06:29:30: |
| 83 | + tp = TwintPool(is_tor=True) |
| 84 | + fh = FirehoseJob(neo4j_creds=creds, PARQUET_SAMPLE_RATE_TIME_S=30, save_to_neo=True, writers={}) |
| 85 | + try: |
| 86 | + search = "covid OR corona OR virus OR pandemic" |
| 87 | + job_name = "covid multi test" |
| 88 | + limit = 10000000 |
| 89 | + for df in fh.search_time_range(tp=tp, Search=search, Since=str(start), Until=str(current), job_name=job_name, Limit=10000000, stride_sec=30): |
| 90 | + logger.info('got: %s', len(df) if not (df is None) else 'None') |
| 91 | + logger.info('proceed to next df') |
| 92 | + except Exception as e: |
| 93 | + logger.error("job exception", exc_info=True) |
| 94 | + raise e |
| 95 | + logger.info("job finished") |
| 96 | + |
| 97 | +# In[ ]: |
| 98 | + |
| 99 | + |
| 100 | +schedule = IntervalSchedule( |
| 101 | + start_date=datetime(2020, 9, 5), |
| 102 | + interval=timedelta(seconds=10), |
| 103 | +) |
| 104 | +storage = S3(bucket=S3_BUCKET) |
| 105 | + |
| 106 | +#with Flow("covid-19 stream", storage=storage, schedule=schedule) as flow: |
| 107 | +with Flow("covid-19 stream-single") as flow: |
| 108 | + run_stream() |
| 109 | +flow.run() |
| 110 | + |
| 111 | + |
| 112 | +# In[ ]: |
| 113 | + |
| 114 | + |
| 115 | + |
| 116 | + |
0 commit comments