| 1 |
5079
|
aaronmk
|
#!/usr/bin/env python
|
| 2 |
5323
|
aaronmk
|
# Scrubs the taxonconcepts in VegBIEN using TNRS.
|
| 3 |
5079
|
aaronmk
|
# Runs continuously until no new rows are added after max_pause.
|
| 4 |
|
|
|
| 5 |
|
|
import os.path
|
| 6 |
|
|
import StringIO
|
| 7 |
|
|
import sys
|
| 8 |
|
|
import time
|
| 9 |
|
|
|
| 10 |
|
|
sys.path.append(os.path.dirname(__file__)+"/../lib")
|
| 11 |
|
|
|
| 12 |
|
|
import csvs
|
| 13 |
|
|
import opts
|
| 14 |
5098
|
aaronmk
|
import profiling
|
| 15 |
5079
|
aaronmk
|
import sql
|
| 16 |
|
|
import sql_gen
|
| 17 |
|
|
import sql_io
|
| 18 |
|
|
import streams
|
| 19 |
|
|
import strings
|
| 20 |
|
|
import tnrs
|
| 21 |
|
|
|
| 22 |
|
|
# Config
|
| 23 |
5156
|
aaronmk
|
pause = 2*60*60 # sec; = 2 hr
|
| 24 |
|
|
max_pause = 9*60*60 # sec; = 9 hr; must be >= max partition import time (1.5 hr)
|
| 25 |
5079
|
aaronmk
|
assert pause <= max_pause
|
| 26 |
|
|
|
| 27 |
|
|
tnrs_data = sql_gen.Table('tnrs')
|
| 28 |
|
|
|
| 29 |
|
|
def main():
|
| 30 |
|
|
# Input
|
| 31 |
|
|
env_names = []
|
| 32 |
|
|
db_config = opts.get_env_vars(sql.db_config_names, None, env_names)
|
| 33 |
|
|
verbosity = float(opts.get_env_var('verbosity', 3, env_names))
|
| 34 |
5214
|
aaronmk
|
wait = opts.env_flag('wait', False, env_names)
|
| 35 |
5079
|
aaronmk
|
if not 'engine' in db_config: raise SystemExit('Usage: '
|
| 36 |
|
|
+opts.env_usage(env_names)+' '+sys.argv[0]+' 2>>log')
|
| 37 |
|
|
|
| 38 |
|
|
def log(msg, level=1):
|
| 39 |
|
|
'''Higher level -> more verbose'''
|
| 40 |
|
|
if level <= verbosity:
|
| 41 |
|
|
sys.stderr.write(strings.to_raw_str(msg.rstrip('\n')+'\n'))
|
| 42 |
|
|
|
| 43 |
|
|
# Connect to DB
|
| 44 |
|
|
db = sql.connect(db_config, log_debug=log)
|
| 45 |
|
|
|
| 46 |
5124
|
aaronmk
|
tnrs_profiler = profiling.ItersProfiler(iter_text='name')
|
| 47 |
5123
|
aaronmk
|
|
| 48 |
5323
|
aaronmk
|
# Iterate over unscrubbed verbatim taxonconcepts
|
| 49 |
5123
|
aaronmk
|
total_pause = 0
|
| 50 |
5322
|
aaronmk
|
tables = ['taxonconcept', sql_gen.Join('tnrs',
|
| 51 |
5159
|
aaronmk
|
{'Name_submitted': 'identifyingtaxonomicname'}, sql_gen.filter_out)]
|
| 52 |
|
|
# Has a concatenated name and not already linked to an accepted name
|
| 53 |
|
|
conds = [('identifyingtaxonomicname', sql_gen.CompareCond(None, '!=')),
|
| 54 |
5322
|
aaronmk
|
('canon_taxonconcept_id', None)]
|
| 55 |
5123
|
aaronmk
|
while True:
|
| 56 |
|
|
# Fetch next set
|
| 57 |
5159
|
aaronmk
|
cur = sql.select(db, tables, ['identifyingtaxonomicname'], conds,
|
| 58 |
|
|
limit=tnrs.max_names, cacheable=False)
|
| 59 |
5123
|
aaronmk
|
this_ct = cur.rowcount
|
| 60 |
5324
|
aaronmk
|
log('Processing '+str(this_ct)+' taxonconcepts')
|
| 61 |
5123
|
aaronmk
|
if this_ct == 0:
|
| 62 |
5213
|
aaronmk
|
if not wait: break
|
| 63 |
5212
|
aaronmk
|
log('Waited '+str(total_pause)+' sec total')
|
| 64 |
5123
|
aaronmk
|
total_pause += pause
|
| 65 |
|
|
if total_pause > max_pause: break
|
| 66 |
5212
|
aaronmk
|
log('Waiting '+str(pause)+' sec...')
|
| 67 |
5123
|
aaronmk
|
time.sleep(pause) # wait for more rows
|
| 68 |
|
|
continue # try again
|
| 69 |
|
|
# otherwise, rows found
|
| 70 |
5079
|
aaronmk
|
total_pause = 0
|
| 71 |
5123
|
aaronmk
|
names = list(sql.values(cur))
|
| 72 |
|
|
|
| 73 |
|
|
# Run TNRS
|
| 74 |
|
|
log('Making TNRS request')
|
| 75 |
|
|
tnrs_profiler.start()
|
| 76 |
|
|
try:
|
| 77 |
|
|
try: stream = tnrs.repeated_tnrs_request(names)
|
| 78 |
|
|
finally:
|
| 79 |
|
|
tnrs_profiler.stop(iter_ct=this_ct)
|
| 80 |
|
|
log('Cumulatively: '+tnrs_profiler.msg())
|
| 81 |
|
|
except tnrs.InvalidResponse: pass # skip set in case it caused error
|
| 82 |
|
|
else:
|
| 83 |
|
|
log('Storing TNRS response data')
|
| 84 |
|
|
stream_info = csvs.stream_info(stream, parse_header=True)
|
| 85 |
|
|
stream = streams.ProgressInputStream(stream, sys.stderr, n=1000)
|
| 86 |
|
|
sql_io.append_csv(db, tnrs_data, stream_info, stream)
|
| 87 |
5079
|
aaronmk
|
|
| 88 |
|
|
main()
|
