| 1 |
1631
|
aaronmk
|
#!/usr/bin/env python
|
| 2 |
|
|
# Downloads REMIB data for all nodes
|
| 3 |
4415
|
aaronmk
|
# Usage: env [start=...] [n=...] self
|
| 4 |
1631
|
aaronmk
|
|
| 5 |
1633
|
aaronmk
|
import csv
|
| 6 |
|
|
import itertools
|
| 7 |
1631
|
aaronmk
|
import os.path
|
| 8 |
|
|
import StringIO
|
| 9 |
|
|
import sys
|
| 10 |
|
|
import urllib2
|
| 11 |
|
|
|
| 12 |
4416
|
aaronmk
|
sys.path.append(os.path.dirname(__file__)+"/../../lib")
|
| 13 |
1631
|
aaronmk
|
|
| 14 |
1637
|
aaronmk
|
import exc
|
| 15 |
1646
|
aaronmk
|
import opts
|
| 16 |
1631
|
aaronmk
|
import profiling
|
| 17 |
|
|
import streams
|
| 18 |
|
|
import strings
|
| 19 |
1637
|
aaronmk
|
import timeout
|
| 20 |
1631
|
aaronmk
|
import util
|
| 21 |
|
|
|
| 22 |
1669
|
aaronmk
|
# Config
|
| 23 |
1673
|
aaronmk
|
timeout = 20 # sec
|
| 24 |
1669
|
aaronmk
|
max_consec_empty_responses = 10
|
| 25 |
1637
|
aaronmk
|
|
| 26 |
1633
|
aaronmk
|
alphabet = map(chr, xrange(ord('A'), ord('Z')+1))
|
| 27 |
|
|
|
| 28 |
1637
|
aaronmk
|
class InputException(Exception): pass
|
| 29 |
|
|
|
| 30 |
1651
|
aaronmk
|
class EmptyResponseException(InputException): pass
|
| 31 |
|
|
|
| 32 |
1631
|
aaronmk
|
def is_ignore(line):
|
| 33 |
|
|
line = strings.remove_line_ending(line)
|
| 34 |
|
|
return line == '' or line.startswith('\t') or line.find(',') < 0
|
| 35 |
|
|
|
| 36 |
|
|
def main():
|
| 37 |
1646
|
aaronmk
|
# Get config from env vars
|
| 38 |
|
|
start = util.cast(int, opts.get_env_var('start', 1))
|
| 39 |
|
|
end = util.cast(int, util.none_if(opts.get_env_var('n', None), u''))
|
| 40 |
1669
|
aaronmk
|
if end != None: end += start
|
| 41 |
1646
|
aaronmk
|
|
| 42 |
4415
|
aaronmk
|
log_ = open(sys.argv[0]+'.log', 'a')
|
| 43 |
|
|
def clear_line(): log_.write('\n')
|
| 44 |
1631
|
aaronmk
|
log_indent = 0
|
| 45 |
|
|
def log(msg, line_ending='\n'):
|
| 46 |
4415
|
aaronmk
|
log_.write((' '*log_indent)+msg+line_ending)
|
| 47 |
1631
|
aaronmk
|
|
| 48 |
1643
|
aaronmk
|
os.chdir(os.path.dirname(__file__)) # dir of output files
|
| 49 |
1631
|
aaronmk
|
|
| 50 |
|
|
# Get by family ('familia') because that is the most general level at which
|
| 51 |
|
|
# an identification can be made. This assumes all records have a family.
|
| 52 |
|
|
url_template = ('http://www.conabio.gob.mx/remib/cgi-bin/'
|
| 53 |
|
|
'remib_distribucion.cgi?lengua=EN&niveltax=familia&taxon=[prefix]%25&'
|
| 54 |
|
|
'pais=Todos&pais_otro=&estado=100&formato=csv&mapa=no&mapabase=estados'
|
| 55 |
|
|
'&coleccion=id%3D[node_id]')
|
| 56 |
1646
|
aaronmk
|
|
| 57 |
1669
|
aaronmk
|
if end != None: node_ids = xrange(start, end)
|
| 58 |
|
|
else: node_ids = itertools.count(start)
|
| 59 |
|
|
consec_empty_responses = 0
|
| 60 |
|
|
for node_id in node_ids:
|
| 61 |
|
|
if consec_empty_responses > max_consec_empty_responses: break
|
| 62 |
|
|
|
| 63 |
1631
|
aaronmk
|
log('Processing node #'+str(node_id)+'...')
|
| 64 |
|
|
log_indent += 1
|
| 65 |
|
|
profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
|
| 66 |
|
|
|
| 67 |
4417
|
aaronmk
|
filename = 'Specimen/node.'+str(node_id)+'.csv'
|
| 68 |
1684
|
aaronmk
|
out = streams.LineCountStream(open(filename, 'w'))
|
| 69 |
1649
|
aaronmk
|
def log_ex(e):
|
| 70 |
|
|
clear_line()
|
| 71 |
|
|
log('! Output line '+str(out.line_num)+': '+exc.str_(e))
|
| 72 |
1643
|
aaronmk
|
start_line_num = out.line_num
|
| 73 |
1631
|
aaronmk
|
node_url_template = url_template.replace('[node_id]', str(node_id))
|
| 74 |
1643
|
aaronmk
|
|
| 75 |
1633
|
aaronmk
|
for prefix_chars in itertools.product(alphabet, repeat=2):
|
| 76 |
|
|
prefix = ''.join(prefix_chars)
|
| 77 |
1631
|
aaronmk
|
log('Processing prefix '+prefix+'...')
|
| 78 |
1633
|
aaronmk
|
row_ct = 0
|
| 79 |
|
|
def print_status(line_ending='\n'):
|
| 80 |
|
|
log('Processed '+str(row_ct)+' row(s)', line_ending)
|
| 81 |
1631
|
aaronmk
|
log_indent += 1
|
| 82 |
|
|
|
| 83 |
|
|
url = node_url_template.replace('[prefix]', prefix)
|
| 84 |
1637
|
aaronmk
|
stream = streams.StreamIter(streams.TimeoutInputStream(
|
| 85 |
1673
|
aaronmk
|
urllib2.urlopen(url), timeout))
|
| 86 |
1631
|
aaronmk
|
|
| 87 |
1669
|
aaronmk
|
is_empty_response = False
|
| 88 |
1637
|
aaronmk
|
try:
|
| 89 |
1641
|
aaronmk
|
util.skip(stream, is_ignore) # skip header
|
| 90 |
1650
|
aaronmk
|
try: metadata_row = csv.reader(stream).next()
|
| 91 |
1651
|
aaronmk
|
except StopIteration: raise EmptyResponseException()
|
| 92 |
1650
|
aaronmk
|
if metadata_row[0] != 'COLLECTION': raise InputException(
|
| 93 |
|
|
'Invalid metadata row: '+str(metadata_row))
|
| 94 |
1641
|
aaronmk
|
|
| 95 |
1650
|
aaronmk
|
# Copy lines
|
| 96 |
1637
|
aaronmk
|
for line in stream:
|
| 97 |
|
|
if is_ignore(line):
|
| 98 |
|
|
error = strings.remove_prefix('\t\t', line)
|
| 99 |
|
|
if len(error) != len(line): raise InputException(error)
|
| 100 |
|
|
break
|
| 101 |
1647
|
aaronmk
|
out.write(line)
|
| 102 |
|
|
|
| 103 |
|
|
row_ct += 1
|
| 104 |
1637
|
aaronmk
|
if row_ct % 100 == 0: print_status('\r')
|
| 105 |
|
|
# CR at end so next print overwrites msg
|
| 106 |
1651
|
aaronmk
|
except EmptyResponseException, e: # must come before InputException
|
| 107 |
1669
|
aaronmk
|
is_empty_response = True
|
| 108 |
1651
|
aaronmk
|
log_ex(e)
|
| 109 |
|
|
break # assume node doesn't exist, so abort node
|
| 110 |
1649
|
aaronmk
|
except InputException, e: log_ex(e)
|
| 111 |
|
|
except timeout.TimeoutException, e:
|
| 112 |
|
|
log_ex(e)
|
| 113 |
|
|
break # assume node is down, so abort node
|
| 114 |
|
|
finally: # still run if break is called
|
| 115 |
|
|
stream.close()
|
| 116 |
|
|
|
| 117 |
|
|
profiler.add_iters(row_ct)
|
| 118 |
|
|
print_status()
|
| 119 |
|
|
log_indent -= 1
|
| 120 |
1669
|
aaronmk
|
|
| 121 |
|
|
if is_empty_response: consec_empty_responses += 1
|
| 122 |
|
|
else: consec_empty_responses = 0 # reset count
|
| 123 |
1631
|
aaronmk
|
|
| 124 |
1633
|
aaronmk
|
profiler.stop()
|
| 125 |
1631
|
aaronmk
|
log(profiler.msg())
|
| 126 |
|
|
log_indent -= 1
|
| 127 |
|
|
|
| 128 |
|
|
main()
|
