1 |
53
|
aaronmk
|
#!/usr/bin/env python
|
2 |
|
|
# Maps one datasource to another, using a map spreadsheet if needed
|
3 |
986
|
aaronmk
|
# Exit status is the # of errors in the import, up to the maximum exit status
|
4 |
53
|
aaronmk
|
# For outputting an XML file to a PostgreSQL database, use the general format of
|
5 |
|
|
# http://vegbank.org/vegdocs/xml/vegbank_example_ver1.0.2.xml
|
6 |
|
|
|
7 |
1014
|
aaronmk
|
import csv
|
8 |
53
|
aaronmk
|
import os.path
|
9 |
|
|
import sys
|
10 |
299
|
aaronmk
|
import xml.dom.minidom as minidom
|
11 |
53
|
aaronmk
|
|
12 |
266
|
aaronmk
|
sys.path.append(os.path.dirname(__file__)+"/../lib")
|
13 |
53
|
aaronmk
|
|
14 |
344
|
aaronmk
|
import exc
|
15 |
64
|
aaronmk
|
import opts
|
16 |
281
|
aaronmk
|
import Parser
|
17 |
982
|
aaronmk
|
import profiling
|
18 |
131
|
aaronmk
|
import sql
|
19 |
715
|
aaronmk
|
import strings
|
20 |
828
|
aaronmk
|
import term
|
21 |
310
|
aaronmk
|
import util
|
22 |
1014
|
aaronmk
|
import xpath
|
23 |
133
|
aaronmk
|
import xml_dom
|
24 |
86
|
aaronmk
|
import xml_func
|
25 |
53
|
aaronmk
|
|
26 |
1018
|
aaronmk
|
def metadata_value(name): return None # this feature has been removed
|
27 |
84
|
aaronmk
|
|
28 |
847
|
aaronmk
|
def main_():
|
29 |
131
|
aaronmk
|
env_names = []
|
30 |
|
|
def usage_err():
|
31 |
944
|
aaronmk
|
raise SystemExit('Usage: '+opts.env_usage(env_names, True)+' '
|
32 |
847
|
aaronmk
|
+sys.argv[0]+' [map_path...] [<input] [>output]')
|
33 |
838
|
aaronmk
|
|
34 |
|
|
# Get db config from env vars
|
35 |
|
|
db_config_names = ['engine', 'host', 'user', 'password', 'database']
|
36 |
|
|
def get_db_config(prefix):
|
37 |
|
|
return opts.get_env_vars(db_config_names, prefix, env_names)
|
38 |
|
|
in_db_config = get_db_config('in')
|
39 |
|
|
out_db_config = get_db_config('out')
|
40 |
|
|
in_is_db = 'engine' in in_db_config
|
41 |
|
|
out_is_db = 'engine' in out_db_config
|
42 |
|
|
|
43 |
|
|
# Get other config from env vars
|
44 |
|
|
end = util.cast(int, opts.get_env_var('n', None, env_names))
|
45 |
|
|
start = util.cast(int, opts.get_env_var('start', '0', env_names))
|
46 |
|
|
if end != None: end += start
|
47 |
946
|
aaronmk
|
test = opts.env_flag('test', False, env_names)
|
48 |
947
|
aaronmk
|
commit = opts.env_flag('commit', False, env_names) and not test
|
49 |
944
|
aaronmk
|
# never commit in test mode
|
50 |
947
|
aaronmk
|
redo = opts.env_flag('redo', test, env_names) and not commit
|
51 |
|
|
# never redo in commit mode (manually run `make empty_db` instead)
|
52 |
946
|
aaronmk
|
debug = opts.env_flag('debug', False, env_names)
|
53 |
859
|
aaronmk
|
sql.run_raw_query.debug = debug
|
54 |
946
|
aaronmk
|
verbose = debug or opts.env_flag('verbose', False, env_names)
|
55 |
944
|
aaronmk
|
opts.get_env_var('profile_to', None, env_names) # add to env_names
|
56 |
131
|
aaronmk
|
|
57 |
838
|
aaronmk
|
# Logging
|
58 |
662
|
aaronmk
|
def log(msg, on=verbose):
|
59 |
|
|
if on: sys.stderr.write(msg)
|
60 |
859
|
aaronmk
|
def log_start(action, on=verbose): log(action+'...\n', on)
|
61 |
662
|
aaronmk
|
|
62 |
53
|
aaronmk
|
# Parse args
|
63 |
510
|
aaronmk
|
map_paths = sys.argv[1:]
|
64 |
512
|
aaronmk
|
if map_paths == []:
|
65 |
|
|
if in_is_db or not out_is_db: usage_err()
|
66 |
|
|
else: map_paths = [None]
|
67 |
53
|
aaronmk
|
|
68 |
646
|
aaronmk
|
def connect_db(db_config):
|
69 |
662
|
aaronmk
|
log_start('Connecting to '+sql.db_config_str(db_config))
|
70 |
1014
|
aaronmk
|
return sql.connect(db_config)
|
71 |
646
|
aaronmk
|
|
72 |
1014
|
aaronmk
|
ex_tracker = exc.ExPercentTracker(iter_text='row')
|
73 |
|
|
profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
|
74 |
|
|
|
75 |
|
|
doc = xml_dom.create_doc()
|
76 |
|
|
root = doc.documentElement
|
77 |
751
|
aaronmk
|
out_is_xml_ref = [False]
|
78 |
1014
|
aaronmk
|
in_label_ref = [None]
|
79 |
|
|
def update_in_label():
|
80 |
|
|
if in_label_ref[0] != None:
|
81 |
|
|
xpath.get(root, '/_ignore/inLabel="'+in_label_ref[0]+'"', True)
|
82 |
|
|
def prep_root():
|
83 |
|
|
root.clear()
|
84 |
|
|
update_in_label()
|
85 |
|
|
prep_root()
|
86 |
751
|
aaronmk
|
|
87 |
838
|
aaronmk
|
def process_input(root, row_ready, map_path):
|
88 |
512
|
aaronmk
|
'''Inputs datasource to XML tree, mapping if needed'''
|
89 |
|
|
# Load map header
|
90 |
|
|
in_is_xpaths = True
|
91 |
751
|
aaronmk
|
out_is_xpaths = True
|
92 |
512
|
aaronmk
|
out_label = None
|
93 |
|
|
if map_path != None:
|
94 |
|
|
metadata = []
|
95 |
|
|
mappings = []
|
96 |
|
|
stream = open(map_path, 'rb')
|
97 |
|
|
reader = csv.reader(stream)
|
98 |
|
|
in_label, out_label = reader.next()[:2]
|
99 |
|
|
def split_col_name(name):
|
100 |
|
|
name, sep, root = name.partition(':')
|
101 |
1122
|
aaronmk
|
return name.partition('-')[0], sep != '', root
|
102 |
|
|
# extract datasrc from "datasrc-data_format"
|
103 |
512
|
aaronmk
|
in_label, in_is_xpaths, in_root = split_col_name(in_label)
|
104 |
1014
|
aaronmk
|
in_label_ref[0] = in_label
|
105 |
|
|
update_in_label()
|
106 |
512
|
aaronmk
|
out_label, out_is_xpaths, out_root = split_col_name(out_label)
|
107 |
|
|
has_types = out_root.startswith('/*s/') # outer elements are types
|
108 |
|
|
for row in reader:
|
109 |
|
|
in_, out = row[:2]
|
110 |
|
|
if out != '':
|
111 |
|
|
if out_is_xpaths: out = xpath.parse(out_root+out)
|
112 |
|
|
mappings.append((in_, out))
|
113 |
|
|
stream.close()
|
114 |
|
|
|
115 |
|
|
root.ownerDocument.documentElement.tagName = out_label
|
116 |
|
|
in_is_xml = in_is_xpaths and not in_is_db
|
117 |
751
|
aaronmk
|
out_is_xml_ref[0] = out_is_xpaths and not out_is_db
|
118 |
56
|
aaronmk
|
|
119 |
512
|
aaronmk
|
if in_is_xml:
|
120 |
|
|
doc0 = minidom.parse(sys.stdin)
|
121 |
1006
|
aaronmk
|
doc0_root = doc0.documentElement
|
122 |
|
|
if out_label == None: out_label = doc0_root.tagName
|
123 |
53
|
aaronmk
|
|
124 |
838
|
aaronmk
|
def process_rows(process_row, rows):
|
125 |
|
|
'''Processes input rows
|
126 |
|
|
@param process_row(in_row, i)
|
127 |
297
|
aaronmk
|
'''
|
128 |
838
|
aaronmk
|
i = -1 # in case for loop does not execute
|
129 |
314
|
aaronmk
|
for i, row in enumerate(rows):
|
130 |
838
|
aaronmk
|
if i < start: continue
|
131 |
|
|
if end != None and i >= end: break
|
132 |
|
|
process_row(row, i)
|
133 |
|
|
row_ready(i, row)
|
134 |
978
|
aaronmk
|
row_ct = i-start+1
|
135 |
982
|
aaronmk
|
return row_ct
|
136 |
838
|
aaronmk
|
|
137 |
|
|
def map_rows(get_value, rows):
|
138 |
|
|
'''Maps input rows
|
139 |
|
|
@param get_value(in_, row):str
|
140 |
|
|
'''
|
141 |
|
|
def process_row(row, i):
|
142 |
316
|
aaronmk
|
row_id = str(i)
|
143 |
|
|
for in_, out in mappings:
|
144 |
|
|
value = metadata_value(in_)
|
145 |
662
|
aaronmk
|
if value == None:
|
146 |
857
|
aaronmk
|
log_start('Getting '+str(in_), debug)
|
147 |
662
|
aaronmk
|
value = get_value(in_, row)
|
148 |
715
|
aaronmk
|
if value != None: xpath.put_obj(root, out, row_id,
|
149 |
|
|
has_types, strings.cleanup(value))
|
150 |
982
|
aaronmk
|
return process_rows(process_row, rows)
|
151 |
297
|
aaronmk
|
|
152 |
310
|
aaronmk
|
if map_path == None:
|
153 |
1006
|
aaronmk
|
iter_ = xml_dom.NodeElemIter(doc0_root)
|
154 |
310
|
aaronmk
|
util.skip(iter_, xml_dom.is_text) # skip metadata
|
155 |
982
|
aaronmk
|
row_ct = process_rows(lambda row, i: root.appendChild(row), iter_)
|
156 |
309
|
aaronmk
|
elif in_is_db:
|
157 |
130
|
aaronmk
|
assert in_is_xpaths
|
158 |
126
|
aaronmk
|
|
159 |
117
|
aaronmk
|
import db_xml
|
160 |
|
|
|
161 |
161
|
aaronmk
|
in_root_xml = xpath.path2xml(in_root)
|
162 |
164
|
aaronmk
|
for i, mapping in enumerate(mappings):
|
163 |
|
|
in_, out = mapping
|
164 |
|
|
if metadata_value(in_) == None:
|
165 |
168
|
aaronmk
|
mappings[i] = (xpath.path2xml(in_root+'/'+in_), out)
|
166 |
126
|
aaronmk
|
|
167 |
646
|
aaronmk
|
in_db = connect_db(in_db_config)
|
168 |
133
|
aaronmk
|
in_pkeys = {}
|
169 |
297
|
aaronmk
|
def get_value(in_, row):
|
170 |
167
|
aaronmk
|
pkey, = row
|
171 |
297
|
aaronmk
|
in_ = in_.cloneNode(True) # don't modify orig value!
|
172 |
886
|
aaronmk
|
xml_dom.set_id(xpath.get(in_, in_root)[0], pkey)
|
173 |
297
|
aaronmk
|
value = sql.value_or_none(db_xml.get(in_db, in_, in_pkeys))
|
174 |
|
|
if value != None: return str(value)
|
175 |
|
|
else: return None
|
176 |
982
|
aaronmk
|
row_ct = map_rows(get_value, sql.rows(db_xml.get(in_db, in_root_xml,
|
177 |
865
|
aaronmk
|
in_pkeys, end, 0)))
|
178 |
117
|
aaronmk
|
in_db.close()
|
179 |
161
|
aaronmk
|
elif in_is_xml:
|
180 |
297
|
aaronmk
|
def get_value(in_, row):
|
181 |
1005
|
aaronmk
|
nodes = xpath.get(row, in_, allow_rooted=False)
|
182 |
886
|
aaronmk
|
if nodes != []: return xml_dom.value(nodes[0])
|
183 |
297
|
aaronmk
|
else: return None
|
184 |
1006
|
aaronmk
|
rows = xpath.get(doc0_root, in_root, limit=end)
|
185 |
886
|
aaronmk
|
if rows == []: raise SystemExit('Map error: Root "'+in_root
|
186 |
883
|
aaronmk
|
+'" not found in input')
|
187 |
982
|
aaronmk
|
row_ct = map_rows(get_value, rows)
|
188 |
56
|
aaronmk
|
else: # input is CSV
|
189 |
133
|
aaronmk
|
map_ = dict(mappings)
|
190 |
59
|
aaronmk
|
reader = csv.reader(sys.stdin)
|
191 |
84
|
aaronmk
|
cols = reader.next()
|
192 |
162
|
aaronmk
|
col_idxs = dict([(value, idx) for idx, value in enumerate(cols)])
|
193 |
164
|
aaronmk
|
for i, mapping in enumerate(mappings):
|
194 |
|
|
in_, out = mapping
|
195 |
|
|
if metadata_value(in_) == None:
|
196 |
|
|
try: mappings[i] = (col_idxs[in_], out)
|
197 |
|
|
except KeyError: pass
|
198 |
162
|
aaronmk
|
|
199 |
297
|
aaronmk
|
def get_value(in_, row):
|
200 |
1133
|
aaronmk
|
try: value = row[in_]
|
201 |
|
|
except KeyError: pass
|
202 |
297
|
aaronmk
|
if value != '': return value
|
203 |
|
|
else: return None
|
204 |
982
|
aaronmk
|
row_ct = map_rows(get_value, reader)
|
205 |
|
|
|
206 |
|
|
return row_ct
|
207 |
53
|
aaronmk
|
|
208 |
838
|
aaronmk
|
def process_inputs(root, row_ready):
|
209 |
982
|
aaronmk
|
row_ct = 0
|
210 |
|
|
for map_path in map_paths:
|
211 |
|
|
row_ct += process_input(root, row_ready, map_path)
|
212 |
|
|
return row_ct
|
213 |
512
|
aaronmk
|
|
214 |
130
|
aaronmk
|
if out_is_db:
|
215 |
53
|
aaronmk
|
import db_xml
|
216 |
|
|
|
217 |
646
|
aaronmk
|
out_db = connect_db(out_db_config)
|
218 |
310
|
aaronmk
|
out_pkeys = {}
|
219 |
53
|
aaronmk
|
try:
|
220 |
947
|
aaronmk
|
if redo: sql.empty_db(out_db)
|
221 |
982
|
aaronmk
|
row_ins_ct_ref = [0]
|
222 |
449
|
aaronmk
|
|
223 |
838
|
aaronmk
|
def row_ready(row_num, input_row):
|
224 |
452
|
aaronmk
|
def on_error(e):
|
225 |
835
|
aaronmk
|
exc.add_msg(e, term.emph('row #:')+' '+str(row_num))
|
226 |
828
|
aaronmk
|
exc.add_msg(e, term.emph('input row:')+'\n'+str(input_row))
|
227 |
|
|
exc.add_msg(e, term.emph('output row:')+'\n'+str(root))
|
228 |
452
|
aaronmk
|
ex_tracker.track(e)
|
229 |
|
|
|
230 |
449
|
aaronmk
|
xml_func.process(root, on_error)
|
231 |
442
|
aaronmk
|
if not xml_dom.is_empty(root):
|
232 |
|
|
assert xml_dom.has_one_child(root)
|
233 |
|
|
try:
|
234 |
982
|
aaronmk
|
sql.with_savepoint(out_db,
|
235 |
|
|
lambda: db_xml.put(out_db, root.firstChild,
|
236 |
|
|
out_pkeys, row_ins_ct_ref, on_error))
|
237 |
442
|
aaronmk
|
if commit: out_db.commit()
|
238 |
449
|
aaronmk
|
except sql.DatabaseErrors, e: on_error(e)
|
239 |
1010
|
aaronmk
|
prep_root()
|
240 |
449
|
aaronmk
|
|
241 |
982
|
aaronmk
|
row_ct = process_inputs(root, row_ready)
|
242 |
|
|
sys.stdout.write('Inserted '+str(row_ins_ct_ref[0])+
|
243 |
460
|
aaronmk
|
' new rows into database\n')
|
244 |
53
|
aaronmk
|
finally:
|
245 |
133
|
aaronmk
|
out_db.rollback()
|
246 |
|
|
out_db.close()
|
247 |
751
|
aaronmk
|
else:
|
248 |
759
|
aaronmk
|
def on_error(e): ex_tracker.track(e)
|
249 |
838
|
aaronmk
|
def row_ready(row_num, input_row): pass
|
250 |
982
|
aaronmk
|
row_ct = process_inputs(root, row_ready)
|
251 |
759
|
aaronmk
|
xml_func.process(root, on_error)
|
252 |
751
|
aaronmk
|
if out_is_xml_ref[0]:
|
253 |
|
|
doc.writexml(sys.stdout, **xml_dom.prettyxml_config)
|
254 |
|
|
else: # output is CSV
|
255 |
|
|
raise NotImplementedError('CSV output not supported yet')
|
256 |
985
|
aaronmk
|
|
257 |
982
|
aaronmk
|
profiler.stop(row_ct)
|
258 |
|
|
ex_tracker.add_iters(row_ct)
|
259 |
990
|
aaronmk
|
if verbose:
|
260 |
|
|
sys.stderr.write('Processed '+str(row_ct)+' input rows\n')
|
261 |
|
|
sys.stderr.write(profiler.msg()+'\n')
|
262 |
|
|
sys.stderr.write(ex_tracker.msg()+'\n')
|
263 |
985
|
aaronmk
|
ex_tracker.exit()
|
264 |
53
|
aaronmk
|
|
265 |
847
|
aaronmk
|
def main():
|
266 |
|
|
try: main_()
|
267 |
|
|
except Parser.SyntaxException, e: raise SystemExit(str(e))
|
268 |
|
|
|
269 |
846
|
aaronmk
|
if __name__ == '__main__':
|
270 |
847
|
aaronmk
|
profile_to = opts.get_env_var('profile_to', None)
|
271 |
|
|
if profile_to != None:
|
272 |
|
|
import cProfile
|
273 |
|
|
cProfile.run(main.func_code, profile_to)
|
274 |
|
|
else: main()
|