/
extract_columns.py
executable file
·236 lines (197 loc) · 11.5 KB
/
extract_columns.py
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
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
#!/usr/bin/env python
"""
Pull all the temperature and salinity data out of a STOQS database no
matter what platform and write it out in Parquet file format.
This is a companion to select_data_in_columns_for_data_science.ipynb
where we operationalize the explorations demonstrated in this Notebook:
https://nbviewer.jupyter.org/github/stoqs/stoqs/blob/master/stoqs/contrib/notebooks/select_data_in_columns_for_data_science.ipynb
Sample command line executions:
(base) ➜ docker git:(master) ✗ docker-compose exec stoqs stoqs/contrib/parquet/extract_columns.py --db stoqs_canon_october2020 --platforms dorado -o dorado.parquet -v
INFO 2021-02-24 21:35:53,588 extract_columns.py _estimate_memory():161 Estimated required_memory = 146584085.76
INFO 2021-02-24 21:35:53,588 extract_columns.py _sql_to_df():65 Reading from SQL query into DataFrame...
INFO 2021-02-24 21:36:11,430 extract_columns.py _sql_to_df():77 df.shape: (2245467, 8) - read_sql_query() in 17.8 sec
INFO 2021-02-24 21:36:11,433 extract_columns.py _sql_to_df():78 df.memory_usage().sum(): 143710016
INFO 2021-02-24 21:36:13,876 extract_columns.py pivot_table_to_parquet():172 Writing data to file dorado.parquet...
INFO 2021-02-24 21:36:14,378 extract_columns.py pivot_table_to_parquet():176 dfp.shape: (169159, 16) - to_parquet() in 0.5 sec
INFO 2021-02-24 21:36:14,378 extract_columns.py pivot_table_to_parquet():177 Done
stoqs container peak memory usage: 2.1 GB
(base) ➜ docker git:(master) ✗ docker-compose exec stoqs stoqs/contrib/parquet/extract_columns.py --db stoqs_canon_october2020 --platforms pontus makai -o lrauv.parquet -v
INFO 2021-02-24 21:38:42,623 extract_columns.py _estimate_memory():161 Estimated required_memory = 645795521.28
INFO 2021-02-24 21:38:42,624 extract_columns.py _sql_to_df():65 Reading from SQL query into DataFrame...
INFO 2021-02-24 21:40:13,936 extract_columns.py _sql_to_df():77 df.shape: (9892701, 8) - read_sql_query() in 91.3 sec
INFO 2021-02-24 21:40:13,938 extract_columns.py _sql_to_df():78 df.memory_usage().sum(): 633132992
INFO 2021-02-24 21:40:30,517 extract_columns.py pivot_table_to_parquet():172 Writing data to file lrauv.parquet...
INFO 2021-02-24 21:40:31,798 extract_columns.py pivot_table_to_parquet():176 dfp.shape: (744662, 25) - to_parquet() in 1.3 sec
INFO 2021-02-24 21:40:31,798 extract_columns.py pivot_table_to_parquet():177 Done
stoqs container peak memory usage: 7.9 GB
(base) ➜ docker git:(master) ✗ docker-compose exec stoqs stoqs/contrib/parquet/extract_columns.py --db stoqs_canon_october2020 -o all_plats.parquet -v
INFO 2021-02-24 21:41:53,151 extract_columns.py _estimate_memory():161 Estimated required_memory = 896823624.96
INFO 2021-02-24 21:41:53,153 extract_columns.py _sql_to_df():65 Reading from SQL query into DataFrame...
INFO 2021-02-24 21:43:56,723 extract_columns.py _sql_to_df():77 df.shape: (13738107, 8) - read_sql_query() in 123.6 sec
INFO 2021-02-24 21:43:56,725 extract_columns.py _sql_to_df():78 df.memory_usage().sum(): 879238976
INFO 2021-02-24 21:44:20,578 extract_columns.py pivot_table_to_parquet():172 Writing data to file all_plats.parquet...
INFO 2021-02-24 21:44:23,349 extract_columns.py pivot_table_to_parquet():176 dfp.shape: (1123909, 61) - to_parquet() in 2.8 sec
INFO 2021-02-24 21:44:23,349 extract_columns.py pivot_table_to_parquet():177 Done
stoqs container peak memory usage: 11 GB
A regression of estimated df size to container memory usage gives a factor of 12.3
Mike McCann
MBARI 29 January 2021
"""
import os
import sys
# Insert Django App directory (parent of config) into python path
sys.path.insert(0, os.path.abspath(os.path.join(
os.path.dirname(__file__), "../../")))
if 'DJANGO_SETTINGS_MODULE' not in os.environ:
os.environ['DJANGO_SETTINGS_MODULE'] = 'config.settings.local'
# django >=1.7
try:
import django
django.setup()
except AttributeError:
pass
import argparse
import logging
import pandas as pd
from django.db import connections
from stoqs.models import Platform
from time import time
class Columnar():
logger = logging.getLogger(__name__)
_handler = logging.StreamHandler()
_formatter = logging.Formatter('%(levelname)s %(asctime)s %(filename)s '
'%(funcName)s():%(lineno)d %(message)s')
_handler.setFormatter(_formatter)
_log_levels = (logging.WARN, logging.INFO, logging.DEBUG)
logger.addHandler(_handler)
# Set to GB of RAM that have been resourced to the Docker engine
MAX_CONTAINER_MEMORY = 16
DF_TO_RAM_FACTOR = 12.3
def _set_platforms(self):
'''Set plats and plat_list member variables
'''
platforms = (self.args.platforms or
Platform.objects.using(self.args.db).all()
.values_list('name', flat=True).order_by('name'))
self.logger.debug(platforms)
self.plats = ''
self.plat_list = []
for platform in platforms:
if platform in self.args.platforms_omit:
# Omit some platforms for shorter execution times
continue
self.plats += f"'{platform}',"
self.plat_list.append(platform)
self.plats = self.plats[:-2] + "'"
def _sql_to_df(self, sql, extract=False):
if extract:
self.logger.info('Reading from SQL query into DataFrame...')
# More than 10 GB of RAM is needed in Docker Desktop for reading data
# from stoqs_canon_october2020. The chunksize option in read_sql_query()
# does not help reduce the server side memory usage.
# See: https://stackoverflow.com/a/31843091/1281657
# https://github.com/pandas-dev/pandas/issues/12265#issuecomment-181809005
# https://github.com/pandas-dev/pandas/issues/35689
stime = time()
df = pd.read_sql_query(sql, connections[self.args.db])
etime = time() - stime
if extract:
self.logger.info(f"df.shape: {df.shape} <- read_sql_query() in {etime:.1f} sec")
self.logger.info(f"Actual df.memory_usage().sum():"
f" {(df.memory_usage().sum()/1.e9):.3f} GB")
self.logger.debug(f"Head of original df:\n{df.head()}")
return df
def _build_sql(self, limit=None, order=True, count=False):
self._set_platforms()
# Base query that's similar to the one behind the api/measuredparameter.csv request
sql = f'''\nFROM public.stoqs_measuredparameter
INNER JOIN stoqs_measurement ON (stoqs_measuredparameter.measurement_id = stoqs_measurement.id)
INNER JOIN stoqs_instantpoint ON (stoqs_measurement.instantpoint_id = stoqs_instantpoint.id)
INNER JOIN stoqs_activity ON (stoqs_instantpoint.activity_id = stoqs_activity.id)
INNER JOIN stoqs_platform ON (stoqs_activity.platform_id = stoqs_platform.id)
INNER JOIN stoqs_parameter ON (stoqs_measuredparameter.parameter_id = stoqs_parameter.id)
WHERE stoqs_platform.name IN ({self.plats})
AND stoqs_parameter.{self.args.collect} is not null'''
if count:
sql = 'SELECT count(*) ' + sql
else:
sql = f'''SELECT stoqs_platform.name as platform,
stoqs_instantpoint.timevalue, stoqs_measurement.depth,
ST_X(stoqs_measurement.geom) as longitude,
ST_Y(stoqs_measurement.geom) as latitude,
stoqs_parameter.{self.args.collect},
stoqs_measuredparameter.datavalue {sql}'''
if order:
sql += ('\nORDER BY stoqs_platform.name, stoqs_instantpoint.timevalue,'
' stoqs_measurement.depth, stoqs_parameter.name')
if limit:
sql += f"\nLIMIT {limit}"
self.logger.debug(f'sql = {sql}')
return sql
def _estimate_memory(self):
'''Perform a small query on the selection and extrapolate
to estimate the server-side memory required for the full extraction.
'''
SAMPLE_SIZE = 100
sql = self._build_sql(limit=SAMPLE_SIZE, order=False)
df = self._sql_to_df(sql)
sample_memory = df.memory_usage().sum()
self.logger.debug(f"{sample_memory} B for {SAMPLE_SIZE} records")
total_recs = self._sql_to_df(self._build_sql(count=True))['count'][0]
self.logger.debug(f"total_recs = {total_recs}")
required_memory = total_recs * sample_memory / SAMPLE_SIZE / 1.e9
container_memory = self.DF_TO_RAM_FACTOR * required_memory
self.logger.info(f"Estimated required_memory:"
f" {required_memory:.3f} GB for DataFrame,"
f" {container_memory:.3f} GB for container RAM,")
if container_memory > self.MAX_CONTAINER_MEMORY:
self.logger.exception(f"Request of {container_memory:.3f} GB would"
f" exceed {self.MAX_CONTAINER_MEMORY} GB"
f" of RAM available")
sys.exit(-1)
def pivot_table_to_parquet(self):
'''Approach 4. Use Pandas do a pivot on data read into a DataFrame
'''
self._estimate_memory()
sql = self._build_sql()
df = self._sql_to_df(sql, extract=True)
context = ['platform', 'timevalue', 'depth', 'latitude', 'longitude']
dfp = df.pivot_table(index=context, columns=self.args.collect, values='datavalue')
self.logger.debug(dfp.shape)
self.logger.info(f'Writing data to file {self.args.output}...')
stime = time()
dfp.to_parquet(self.args.output)
etime = time() - stime
self.logger.info(f"dfp.shape: {dfp.shape} -> to_parquet() in {etime:.1f} sec")
self.logger.debug(f"Head of pivoted df:\n{dfp.head()}")
self.logger.info('Done')
def process_command_line(self):
parser = argparse.ArgumentParser(description='Transform STOQS data into columnar Parquet file format')
parser.add_argument('--platforms', action='store', nargs='*',
help='Restrict to just these platforms')
parser.add_argument('--platforms_omit', action='store', nargs='*', default=[],
help='Restrict to all but these platforms')
parser.add_argument('--collect', action='store', default='name',
choices=['name', 'standard_name'],
help='The column to collect: name or standard_name')
parser.add_argument('--db', action='store', required=True,
help='Database alias, e.g. stoqs_canon_october2020')
parser.add_argument('-o', '--output', action='store', required=True,
help='Output file name')
parser.add_argument('--start', action='store', help='Start time in YYYYMMDDTHHMMSS format',
default='19000101T000000')
parser.add_argument('--end', action='store', help='End time in YYYYMMDDTHHMMSS format',
default='22000101T000000')
parser.add_argument('-v', '--verbose', type=int, choices=range(3),
action='store', default=0, const=1, nargs='?',
help="verbosity level: " + ', '.join(
[f"{i}: {v}" for i, v, in enumerate(('WARN', 'INFO', 'DEBUG'))]))
self.args = parser.parse_args()
self.commandline = ' '.join(sys.argv)
self.logger.setLevel(self._log_levels[self.args.verbose])
self.logger.debug(f"Using databases at DATABASE_URL ="
f" {os.environ['DATABASE_URL']}")
if __name__ == '__main__':
c = Columnar()
c.process_command_line()
c.pivot_table_to_parquet()