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