| 1 |
53
|
aaronmk
|
#!/usr/bin/env python
|
| 2 |
|
|
# Maps one datasource to another, using a map spreadsheet if needed
|
| 3 |
|
|
# For outputting an XML file to a PostgreSQL database, use the general format of
|
| 4 |
|
|
# http://vegbank.org/vegdocs/xml/vegbank_example_ver1.0.2.xml
|
| 5 |
|
|
|
| 6 |
|
|
import os.path
|
| 7 |
|
|
import sys
|
| 8 |
299
|
aaronmk
|
import xml.dom.minidom as minidom
|
| 9 |
53
|
aaronmk
|
|
| 10 |
266
|
aaronmk
|
sys.path.append(os.path.dirname(__file__)+"/../lib")
|
| 11 |
53
|
aaronmk
|
|
| 12 |
344
|
aaronmk
|
import exc
|
| 13 |
64
|
aaronmk
|
import opts
|
| 14 |
281
|
aaronmk
|
import Parser
|
| 15 |
131
|
aaronmk
|
import sql
|
| 16 |
715
|
aaronmk
|
import strings
|
| 17 |
310
|
aaronmk
|
import util
|
| 18 |
133
|
aaronmk
|
import xml_dom
|
| 19 |
86
|
aaronmk
|
import xml_func
|
| 20 |
53
|
aaronmk
|
|
| 21 |
84
|
aaronmk
|
def metadata_value(name):
|
| 22 |
164
|
aaronmk
|
if type(name) == str and name.startswith(':'): return name[1:]
|
| 23 |
84
|
aaronmk
|
else: return None
|
| 24 |
|
|
|
| 25 |
53
|
aaronmk
|
def main():
|
| 26 |
434
|
aaronmk
|
ex_tracker = exc.ExTracker()
|
| 27 |
|
|
|
| 28 |
131
|
aaronmk
|
env_names = []
|
| 29 |
|
|
def usage_err():
|
| 30 |
662
|
aaronmk
|
raise SystemExit('Usage: ' + opts.env_usage(env_names, True)
|
| 31 |
|
|
+' [commit=1] [test=1] [verbose=1] [debug=1] '+sys.argv[0]
|
| 32 |
|
|
+' [map_path...] [<input] [>output]')
|
| 33 |
146
|
aaronmk
|
limit = opts.get_env_var('n', None, env_names)
|
| 34 |
|
|
if limit != None: limit = int(limit)
|
| 35 |
400
|
aaronmk
|
test = opts.env_flag('test')
|
| 36 |
|
|
commit = not test and opts.env_flag('commit') # never commit in test mode
|
| 37 |
662
|
aaronmk
|
debug = opts.env_flag('debug')
|
| 38 |
|
|
verbose = debug or opts.env_flag('verbose')
|
| 39 |
131
|
aaronmk
|
|
| 40 |
662
|
aaronmk
|
def log(msg, on=verbose):
|
| 41 |
|
|
if on: sys.stderr.write(msg)
|
| 42 |
663
|
aaronmk
|
def log_start(action, on=verbose): log(action+'...', on)
|
| 43 |
|
|
def log_done(on=verbose): log('Done\n', on)
|
| 44 |
662
|
aaronmk
|
|
| 45 |
53
|
aaronmk
|
# Get db config from env vars
|
| 46 |
131
|
aaronmk
|
db_config_names = ['engine', 'host', 'user', 'password', 'database']
|
| 47 |
53
|
aaronmk
|
def get_db_config(prefix):
|
| 48 |
64
|
aaronmk
|
return opts.get_env_vars(db_config_names, prefix, env_names)
|
| 49 |
67
|
aaronmk
|
in_db_config = get_db_config('in')
|
| 50 |
|
|
out_db_config = get_db_config('out')
|
| 51 |
131
|
aaronmk
|
in_is_db = 'engine' in in_db_config
|
| 52 |
|
|
out_is_db = 'engine' in out_db_config
|
| 53 |
53
|
aaronmk
|
|
| 54 |
|
|
# Parse args
|
| 55 |
510
|
aaronmk
|
map_paths = sys.argv[1:]
|
| 56 |
512
|
aaronmk
|
if map_paths == []:
|
| 57 |
|
|
if in_is_db or not out_is_db: usage_err()
|
| 58 |
|
|
else: map_paths = [None]
|
| 59 |
53
|
aaronmk
|
|
| 60 |
646
|
aaronmk
|
def connect_db(db_config):
|
| 61 |
662
|
aaronmk
|
log_start('Connecting to '+sql.db_config_str(db_config))
|
| 62 |
646
|
aaronmk
|
db = sql.connect(db_config)
|
| 63 |
662
|
aaronmk
|
log_done()
|
| 64 |
646
|
aaronmk
|
return db
|
| 65 |
|
|
|
| 66 |
751
|
aaronmk
|
out_is_xml_ref = [False]
|
| 67 |
|
|
|
| 68 |
512
|
aaronmk
|
def process_input(root, process_row, map_path):
|
| 69 |
|
|
'''Inputs datasource to XML tree, mapping if needed'''
|
| 70 |
|
|
# Load map header
|
| 71 |
|
|
in_is_xpaths = True
|
| 72 |
751
|
aaronmk
|
out_is_xpaths = True
|
| 73 |
512
|
aaronmk
|
out_label = None
|
| 74 |
|
|
if map_path != None:
|
| 75 |
|
|
import copy
|
| 76 |
|
|
import csv
|
| 77 |
|
|
|
| 78 |
|
|
import xpath
|
| 79 |
|
|
|
| 80 |
|
|
metadata = []
|
| 81 |
|
|
mappings = []
|
| 82 |
|
|
stream = open(map_path, 'rb')
|
| 83 |
|
|
reader = csv.reader(stream)
|
| 84 |
|
|
in_label, out_label = reader.next()[:2]
|
| 85 |
|
|
def split_col_name(name):
|
| 86 |
|
|
name, sep, root = name.partition(':')
|
| 87 |
|
|
return name, sep != '', root
|
| 88 |
|
|
in_label, in_is_xpaths, in_root = split_col_name(in_label)
|
| 89 |
|
|
out_label, out_is_xpaths, out_root = split_col_name(out_label)
|
| 90 |
|
|
has_types = out_root.startswith('/*s/') # outer elements are types
|
| 91 |
|
|
for row in reader:
|
| 92 |
|
|
in_, out = row[:2]
|
| 93 |
|
|
if out != '':
|
| 94 |
|
|
if out_is_xpaths: out = xpath.parse(out_root+out)
|
| 95 |
|
|
mappings.append((in_, out))
|
| 96 |
|
|
stream.close()
|
| 97 |
|
|
|
| 98 |
|
|
root.ownerDocument.documentElement.tagName = out_label
|
| 99 |
|
|
in_is_xml = in_is_xpaths and not in_is_db
|
| 100 |
751
|
aaronmk
|
out_is_xml_ref[0] = out_is_xpaths and not out_is_db
|
| 101 |
56
|
aaronmk
|
|
| 102 |
512
|
aaronmk
|
if in_is_xml:
|
| 103 |
|
|
doc0 = minidom.parse(sys.stdin)
|
| 104 |
|
|
if out_label == None: out_label = doc0.documentElement.tagName
|
| 105 |
53
|
aaronmk
|
|
| 106 |
309
|
aaronmk
|
def process_rows(get_value, rows):
|
| 107 |
297
|
aaronmk
|
'''Processes input values
|
| 108 |
|
|
@param get_value f(in_, row):str
|
| 109 |
|
|
'''
|
| 110 |
314
|
aaronmk
|
for i, row in enumerate(rows):
|
| 111 |
316
|
aaronmk
|
if not (limit == None or i < limit): break
|
| 112 |
|
|
row_id = str(i)
|
| 113 |
|
|
for in_, out in mappings:
|
| 114 |
|
|
value = metadata_value(in_)
|
| 115 |
662
|
aaronmk
|
if value == None:
|
| 116 |
|
|
log_start('Getting '+str(in_), debug)
|
| 117 |
|
|
value = get_value(in_, row)
|
| 118 |
|
|
log_done(debug)
|
| 119 |
715
|
aaronmk
|
if value != None: xpath.put_obj(root, out, row_id,
|
| 120 |
|
|
has_types, strings.cleanup(value))
|
| 121 |
452
|
aaronmk
|
process_row(row)
|
| 122 |
460
|
aaronmk
|
sys.stderr.write('Processed '+str(i+1)+' input rows\n')
|
| 123 |
297
|
aaronmk
|
|
| 124 |
310
|
aaronmk
|
if map_path == None:
|
| 125 |
|
|
iter_ = xml_dom.NodeElemIter(doc0.documentElement)
|
| 126 |
|
|
util.skip(iter_, xml_dom.is_text) # skip metadata
|
| 127 |
317
|
aaronmk
|
for child in iter_:
|
| 128 |
|
|
root.appendChild(child)
|
| 129 |
452
|
aaronmk
|
process_row(child)
|
| 130 |
309
|
aaronmk
|
elif in_is_db:
|
| 131 |
130
|
aaronmk
|
assert in_is_xpaths
|
| 132 |
126
|
aaronmk
|
|
| 133 |
117
|
aaronmk
|
import db_xml
|
| 134 |
|
|
|
| 135 |
161
|
aaronmk
|
in_root_xml = xpath.path2xml(in_root)
|
| 136 |
164
|
aaronmk
|
for i, mapping in enumerate(mappings):
|
| 137 |
|
|
in_, out = mapping
|
| 138 |
|
|
if metadata_value(in_) == None:
|
| 139 |
168
|
aaronmk
|
mappings[i] = (xpath.path2xml(in_root+'/'+in_), out)
|
| 140 |
126
|
aaronmk
|
|
| 141 |
646
|
aaronmk
|
in_db = connect_db(in_db_config)
|
| 142 |
133
|
aaronmk
|
in_pkeys = {}
|
| 143 |
297
|
aaronmk
|
def get_value(in_, row):
|
| 144 |
167
|
aaronmk
|
pkey, = row
|
| 145 |
297
|
aaronmk
|
in_ = in_.cloneNode(True) # don't modify orig value!
|
| 146 |
|
|
xml_dom.set_id(xpath.get(in_, in_root), pkey)
|
| 147 |
|
|
value = sql.value_or_none(db_xml.get(in_db, in_, in_pkeys))
|
| 148 |
|
|
if value != None: return str(value)
|
| 149 |
|
|
else: return None
|
| 150 |
309
|
aaronmk
|
process_rows(get_value, sql.rows(db_xml.get(in_db, in_root_xml,
|
| 151 |
|
|
in_pkeys, limit)))
|
| 152 |
117
|
aaronmk
|
in_db.close()
|
| 153 |
161
|
aaronmk
|
elif in_is_xml:
|
| 154 |
297
|
aaronmk
|
def get_value(in_, row):
|
| 155 |
|
|
node = xpath.get(row, in_)
|
| 156 |
|
|
if node != None: return xml_dom.value(node)
|
| 157 |
|
|
else: return None
|
| 158 |
|
|
row0 = xpath.get(doc0.documentElement, in_root)
|
| 159 |
309
|
aaronmk
|
process_rows(get_value, xml_dom.NodeElemIter(row0.parentNode))
|
| 160 |
56
|
aaronmk
|
else: # input is CSV
|
| 161 |
133
|
aaronmk
|
map_ = dict(mappings)
|
| 162 |
59
|
aaronmk
|
reader = csv.reader(sys.stdin)
|
| 163 |
84
|
aaronmk
|
cols = reader.next()
|
| 164 |
162
|
aaronmk
|
col_idxs = dict([(value, idx) for idx, value in enumerate(cols)])
|
| 165 |
164
|
aaronmk
|
for i, mapping in enumerate(mappings):
|
| 166 |
|
|
in_, out = mapping
|
| 167 |
|
|
if metadata_value(in_) == None:
|
| 168 |
|
|
try: mappings[i] = (col_idxs[in_], out)
|
| 169 |
|
|
except KeyError: pass
|
| 170 |
162
|
aaronmk
|
|
| 171 |
297
|
aaronmk
|
def get_value(in_, row):
|
| 172 |
|
|
value = row[in_]
|
| 173 |
|
|
if value != '': return value
|
| 174 |
|
|
else: return None
|
| 175 |
309
|
aaronmk
|
process_rows(get_value, reader)
|
| 176 |
53
|
aaronmk
|
|
| 177 |
512
|
aaronmk
|
def process_inputs(root, process_row):
|
| 178 |
|
|
for map_path in map_paths: process_input(root, process_row, map_path)
|
| 179 |
|
|
|
| 180 |
53
|
aaronmk
|
# Output XML tree
|
| 181 |
512
|
aaronmk
|
doc = xml_dom.create_doc()
|
| 182 |
316
|
aaronmk
|
root = doc.documentElement
|
| 183 |
130
|
aaronmk
|
if out_is_db:
|
| 184 |
53
|
aaronmk
|
import db_xml
|
| 185 |
|
|
|
| 186 |
646
|
aaronmk
|
out_db = connect_db(out_db_config)
|
| 187 |
310
|
aaronmk
|
out_pkeys = {}
|
| 188 |
53
|
aaronmk
|
try:
|
| 189 |
400
|
aaronmk
|
if test: sql.empty_db(out_db)
|
| 190 |
53
|
aaronmk
|
row_ct_ref = [0]
|
| 191 |
449
|
aaronmk
|
|
| 192 |
452
|
aaronmk
|
def process_row(input_row):
|
| 193 |
|
|
def on_error(e):
|
| 194 |
|
|
exc.add_msg(e, 'output row:\n'+str(root))
|
| 195 |
|
|
exc.add_msg(e, 'input row:\n'+str(input_row))
|
| 196 |
|
|
ex_tracker.track(e)
|
| 197 |
|
|
|
| 198 |
449
|
aaronmk
|
xml_func.process(root, on_error)
|
| 199 |
442
|
aaronmk
|
if not xml_dom.is_empty(root):
|
| 200 |
|
|
assert xml_dom.has_one_child(root)
|
| 201 |
|
|
try:
|
| 202 |
|
|
sql.with_savepoint(out_db, lambda: db_xml.put(out_db,
|
| 203 |
462
|
aaronmk
|
root.firstChild, out_pkeys, row_ct_ref, on_error))
|
| 204 |
442
|
aaronmk
|
if commit: out_db.commit()
|
| 205 |
449
|
aaronmk
|
except sql.DatabaseErrors, e: on_error(e)
|
| 206 |
316
|
aaronmk
|
root.clear()
|
| 207 |
449
|
aaronmk
|
|
| 208 |
512
|
aaronmk
|
process_inputs(root, process_row)
|
| 209 |
717
|
aaronmk
|
sys.stdout.write('Inserted '+str(row_ct_ref[0])+
|
| 210 |
460
|
aaronmk
|
' new rows into database\n')
|
| 211 |
53
|
aaronmk
|
finally:
|
| 212 |
133
|
aaronmk
|
out_db.rollback()
|
| 213 |
|
|
out_db.close()
|
| 214 |
751
|
aaronmk
|
else:
|
| 215 |
759
|
aaronmk
|
def on_error(e): ex_tracker.track(e)
|
| 216 |
452
|
aaronmk
|
def process_row(input_row): pass
|
| 217 |
512
|
aaronmk
|
process_inputs(root, process_row)
|
| 218 |
759
|
aaronmk
|
xml_func.process(root, on_error)
|
| 219 |
751
|
aaronmk
|
if out_is_xml_ref[0]:
|
| 220 |
|
|
doc.writexml(sys.stdout, **xml_dom.prettyxml_config)
|
| 221 |
|
|
else: # output is CSV
|
| 222 |
|
|
raise NotImplementedError('CSV output not supported yet')
|
| 223 |
53
|
aaronmk
|
|
| 224 |
133
|
aaronmk
|
try: main()
|
| 225 |
294
|
aaronmk
|
except Parser.SyntaxException, e: raise SystemExit(str(e))
|
