Project

General

Profile

1
#!/usr/bin/env python
2
# Maps one datasource to another, using a map spreadsheet if needed
3
# Exit status is the # of errors in the import, up to the maximum exit status
4
# Multi-safe (supports an input appearing multiple times).
5
# For outputting an XML file to a PostgreSQL database, use the general format of
6
# http://vegbank.org/vegdocs/xml/vegbank_example_ver1.0.2.xml
7

    
8
import copy
9
import csv
10
import itertools
11
import os.path
12
import warnings
13
import sys
14
import xml.dom.minidom as minidom
15

    
16
sys.path.append(os.path.dirname(__file__)+"/../lib")
17

    
18
import csvs
19
import db_xml
20
import exc
21
import iters
22
import maps
23
import opts
24
import parallelproc
25
import Parser
26
import profiling
27
import sql
28
import sql_gen
29
import sql_io
30
import streams
31
import strings
32
import term
33
import util
34
import xpath
35
import xml_dom
36
import xml_func
37
import xml_parse
38

    
39
def get_with_prefix(map_, prefixes, key):
40
    '''Gets all entries for the given key with any of the given prefixes
41
    @return tuple(found_key, found_value)
42
    '''
43
    values = []
44
    for key_ in strings.with_prefixes(['']+prefixes, key): # also with no prefix
45
        try: value = map_[key_]
46
        except KeyError, e: continue # keep going
47
        values.append((key_, value))
48
    
49
    if values != []: return values
50
    else: raise e # re-raise last KeyError
51

    
52
def metadata_value(name): return None # this feature has been removed
53

    
54
def cleanup(val):
55
    if val == None: return val
56
    return util.none_if(strings.cleanup(strings.ustr(val)), u'', u'\\N')
57

    
58
def main_():
59
    env_names = []
60
    def usage_err():
61
        raise SystemExit('Usage: '+opts.env_usage(env_names, True)+' '
62
            +sys.argv[0]+' [map_path...] [<input] [>output]\n'
63
            'Note: Row #s start with 1')
64
    
65
    ## Get config from env vars
66
    
67
    # Modes
68
    test = opts.env_flag('test', False, env_names)
69
    commit = opts.env_flag('commit', False, env_names) and not test
70
        # never commit in test mode
71
    redo = opts.env_flag('redo', test, env_names) and not commit
72
        # never redo in commit mode (run `make schemas/reinstall` instead)
73
    
74
    # Ranges
75
    start = util.cast(int, opts.get_env_var('start', 1, env_names)) # 1-based
76
    # Make start interally 0-based.
77
    # It's 1-based to the user to match up with the staging table row #s.
78
    start -= 1
79
    if test: n_default = 1
80
    else: n_default = None
81
    n = util.cast(int, util.none_if(opts.get_env_var('n', n_default, env_names),
82
        u''))
83
    end = n
84
    if end != None: end += start
85
    
86
    # Debugging
87
    verbosity = util.cast(float, opts.get_env_var('verbosity', None, env_names))
88
    opts.get_env_var('profile_to', None, env_names) # add to env_names
89
    
90
    # DB
91
    def get_db_config(prefix):
92
        return opts.get_env_vars(sql.db_config_names, prefix, env_names)
93
    in_db_config = get_db_config('in')
94
    out_db_config = get_db_config('out')
95
    in_is_db = 'engine' in in_db_config
96
    out_is_db = 'engine' in out_db_config
97
    in_schema = opts.get_env_var('in_schema', None, env_names)
98
    in_table = opts.get_env_var('in_table', None, env_names)
99
    if in_schema != None:
100
        for config in [in_db_config, out_db_config]:
101
            config['schemas'] += ','+in_schema
102
    
103
    # Optimization
104
    cache_sql = opts.env_flag('cache_sql', True, env_names)
105
    by_col = in_db_config == out_db_config and opts.env_flag('by_col', False,
106
        env_names) # by-column optimization only applies if mapping to same DB
107
    if test: cpus_default = 0
108
    else: cpus_default = 0 # or None to use parallel processing by default
109
    cpus = util.cast(int, util.none_if(opts.get_env_var('cpus', cpus_default,
110
        env_names), u''))
111
    
112
    # Set default verbosity. Must happen after by_col is set.
113
    if verbosity == None: 
114
        if test: verbosity = 0.5 # automated tests should not be verbose
115
        elif by_col: verbosity = 3 # show all queries to assist profiling
116
        else: verbosity = 1.1 # just show row progress
117
    
118
    ##
119
    
120
    # Logging
121
    verbose_errors = test and verbosity > 0
122
    debug = verbosity >= 1.5
123
    def log(msg, level=1):
124
        '''Higher level -> more verbose'''
125
        if level <= verbosity:
126
            if verbosity <= 2:
127
                if level == 1.5: msg = '# '+msg # msg is Redmine list item
128
                elif msg.startswith('DB query:'): # remove extra debug info
129
                    first_line, nl, msg = msg.partition('\n')
130
            elif level > 1: msg = '['+str(level)+'] '+msg # include level in msg
131
            
132
            sys.stderr.write(strings.to_raw_str(msg.rstrip('\n')+'\n'))
133
    if debug: log_debug = lambda msg, level=2: log(msg, level)
134
    else: log_debug = sql.log_debug_none
135
    
136
    # Parse args
137
    map_paths = sys.argv[1:]
138
    if map_paths == []:
139
        if in_is_db or not out_is_db: usage_err()
140
        else: map_paths = [None]
141
    
142
    def connect_db(db_config):
143
        log('Connecting to '+sql.db_config_str(db_config))
144
        return sql.connect(db_config, caching=cache_sql, autocommit=commit,
145
            debug_temp=verbosity > 3 and commit, log_debug=log_debug)
146
    
147
    if end != None: end_str = str(end-1) # end is one past the last #
148
    else: end_str = 'end'
149
    log('Processing input rows '+str(start)+'-'+end_str)
150
    
151
    ex_tracker = exc.ExPercentTracker(iter_text='row')
152
    profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
153
    
154
    # Parallel processing
155
    pool = parallelproc.MultiProducerPool(cpus)
156
    log('Using '+str(pool.process_ct)+' parallel CPUs')
157
    
158
    # Set up DB access
159
    row_ins_ct_ref = [0]
160
    if out_is_db:
161
        out_db = connect_db(out_db_config)
162
        def is_rel_func(name):
163
            return (name in db_xml.put_special_funcs
164
                or sql.function_exists(out_db, sql_gen.Function(name)))
165
    
166
    doc = xml_dom.create_doc()
167
    root = doc.documentElement
168
    out_is_xml_ref = [False]
169
    
170
    in_label_ref = [None]
171
    col_defaults = {}
172
    def update_in_label():
173
        if in_label_ref[0] != None and out_is_db:
174
            # TODO: Move this to the mappings as some kind of metadata
175
            col_defaults['datasource_id'] = db_xml.put(out_db, xpath.path2xml(
176
                'party/organizationname="'+in_label_ref[0]+'"'), row_ins_ct_ref)
177
    
178
    def prep_root():
179
        root.clear()
180
        update_in_label()
181
    prep_root()
182
    
183
    def process_input(root, row_ready, map_path):
184
        '''Inputs datasource to XML tree, mapping if needed'''
185
        # Load map header
186
        in_is_xpaths = True
187
        out_is_xpaths = True
188
        out_label = None
189
        if map_path != None:
190
            metadata = []
191
            mappings = []
192
            stream = open(map_path, 'rb')
193
            reader = csv.reader(stream)
194
            in_label, out_label = reader.next()[:2]
195
            
196
            def split_col_name(name):
197
                label, sep, root = name.partition(':')
198
                label, sep2, prefixes_str = label.partition('[')
199
                prefixes_str = strings.remove_suffix(']', prefixes_str)
200
                prefixes = strings.split(',', prefixes_str)
201
                return label, sep != '', root, prefixes
202
                    # extract datasrc from "datasrc[data_format]"
203
            
204
            in_label, in_root, prefixes = maps.col_info(in_label)
205
            in_is_xpaths = in_root != None
206
            in_label_ref[0] = in_label
207
            update_in_label()
208
            out_label, out_root = maps.col_info(out_label)[:2]
209
            out_is_xpaths = out_root != None
210
            if out_is_xpaths: has_types = out_root.find('/*s/') >= 0
211
                # outer elements are types
212
            
213
            for row in reader:
214
                in_, out = row[:2]
215
                if out != '': mappings.append([in_, out_root+out])
216
            
217
            stream.close()
218
            
219
            root.ownerDocument.documentElement.tagName = out_label
220
        in_is_xml = in_is_xpaths and not in_is_db
221
        out_is_xml_ref[0] = out_is_xpaths and not out_is_db
222
        
223
        def process_rows(process_row, rows, rows_start=0):
224
            '''Processes input rows      
225
            @param process_row(in_row, i)
226
            @rows_start The (0-based) row # of the first row in rows. Set this
227
                only if the pre-start rows have already been skipped.
228
            '''
229
            rows = iter(rows)
230
            
231
            if end != None: row_nums = xrange(rows_start, end)
232
            else: row_nums = itertools.count(rows_start)
233
            i = -1
234
            for i in row_nums:
235
                try: row = rows.next()
236
                except StopIteration:
237
                    i -= 1 # last row # didn't count
238
                    break # no more rows
239
                if i < start: continue # not at start row yet
240
                
241
                # Row # is interally 0-based, but 1-based to the user
242
                log('Processing input row #'+str(i+1), level=1.1)
243
                process_row(row, i)
244
                row_ready(i, row)
245
            row_ct = i-start+1
246
            return row_ct
247
        
248
        def map_rows(get_value, rows, **kw_args):
249
            '''Maps input rows
250
            @param get_value(in_, row):str
251
            '''
252
            # Prevent collisions if multiple inputs mapping to same output
253
            outputs_idxs = dict()
254
            for i, mapping in enumerate(mappings):
255
                in_, out = mapping
256
                default = util.NamedTuple(count=1, first=i)
257
                idxs = outputs_idxs.setdefault(out, default)
258
                if idxs is not default: # key existed, so there was a collision
259
                    if idxs.count == 1: # first key does not yet have /_alt/#
260
                        mappings[idxs.first][1] += '/_alt/0'
261
                    mappings[i][1] += '/_alt/'+str(idxs.count)
262
                    idxs.count += 1
263
            
264
            id_node = None
265
            if out_is_db:
266
                for i, mapping in enumerate(mappings):
267
                    in_, out = mapping
268
                    # All put_obj()s should return the same id_node
269
                    nodes, id_node = xpath.put_obj(root, out, '-1', has_types,
270
                        '$'+str(in_)) # value is placeholder that documents name
271
                    mappings[i] = [in_, nodes]
272
                if id_node == None:
273
                    warnings.warn(UserWarning('Map warning: No mappings'))                
274
                sys.stdout.write('Put template:\n'+str(root))
275
                sys.stdout.flush()
276
            
277
            def process_row(row, i):
278
                row_id = str(i)
279
                if id_node != None: xml_dom.set_value(id_node, row_id)
280
                for in_, out in mappings:
281
                    log_debug('Getting '+str(in_))
282
                    value = metadata_value(in_)
283
                    if value == None: value = cleanup(get_value(in_, row))
284
                    log_debug('Putting '+repr(value)+' to '+str(out))
285
                    if out_is_db: # out is list of XML nodes
286
                        for node in out: xml_dom.set_value(node, value)
287
                    elif value != None: # out is XPath
288
                        xpath.put_obj(root, out, row_id, has_types, value)
289
            return process_rows(process_row, rows, **kw_args)
290
        
291
        def map_table(col_names, rows, **kw_args):
292
            col_names_ct = len(col_names)
293
            col_idxs = util.list_flip(col_names)
294
            
295
            # Resolve prefixes
296
            mappings_orig = mappings[:] # save a copy
297
            mappings[:] = [] # empty existing elements
298
            for in_, out in mappings_orig:
299
                if metadata_value(in_) == None:
300
                    try: cols = get_with_prefix(col_idxs, prefixes, in_)
301
                    except KeyError: pass
302
                    else: mappings[len(mappings):] = [[db_xml.ColRef(*col), out]
303
                        for col in cols] # can't use += because that uses =
304
            
305
            def get_value(in_, row): return row.list[in_.idx]
306
            def wrap_row(row):
307
                return util.ListDict(util.list_as_length(row, col_names_ct),
308
                    col_names, col_idxs) # handle CSV rows of different lengths
309
            
310
            return map_rows(get_value, util.WrapIter(wrap_row, rows), **kw_args)
311
        
312
        if in_is_db:
313
            def on_error(e): ex_tracker.track(e)
314
            
315
            if by_col: in_db = out_db
316
            else: in_db = connect_db(in_db_config)
317
            
318
            # Get table and schema name
319
            schema = in_schema # modified, so can't have same name as outer var
320
            table = in_table # modified, so can't have same name as outer var
321
            if table == None:
322
                assert in_is_xpaths
323
                schema, sep, table = in_root.partition('.')
324
                if sep == '': # only the table name was specified
325
                    table = schema
326
                    schema = None
327
            table = sql_gen.Table(table, schema)
328
            
329
            # Fetch rows
330
            if by_col: limit = 0 # only fetch column names
331
            else: limit = n
332
            cur = sql.select(in_db, table, limit=limit, start=start,
333
                cacheable=False)
334
            col_names = list(sql.col_names(cur))
335
            
336
            if by_col:
337
                map_table(col_names, []) # just create the template
338
                
339
                if start == 0 and n == None: # doing full re-import
340
                    log('Clearing errors table')
341
                    sql.truncate(in_db, sql_io.errors_table(in_db, table))
342
                
343
                # Strip XML functions not in the DB
344
                xml_func.process(root, is_rel_func=is_rel_func)
345
                if debug: log_debug('Putting stripped:\n'+str(root))
346
                    # only calc if debug
347
                
348
                # Import rows
349
                in_row_ct_ref = [0]
350
                db_xml.put_table(in_db, root.firstChild, table, in_row_ct_ref,
351
                    row_ins_ct_ref, n, start, on_error, col_defaults)
352
                row_ct = in_row_ct_ref[0]
353
            else:
354
                # Use normal by-row method
355
                row_ct = map_table(col_names, sql.rows(cur), rows_start=start)
356
                    # rows_start: pre-start rows have been skipped
357
                
358
                in_db.db.close()
359
        elif in_is_xml:
360
            stdin = streams.LineCountStream(sys.stdin)
361
            def on_error(e):
362
                exc.add_msg(e, term.emph('input line #:')+' '
363
                    +str(stdin.line_num))
364
                ex_tracker.track(e)
365
            
366
            def get_rows(doc2rows):
367
                return iters.flatten(itertools.imap(doc2rows,
368
                    xml_parse.docs_iter(stdin, on_error)))
369
            
370
            if map_path == None:
371
                def doc2rows(in_xml_root):
372
                    iter_ = xml_dom.NodeElemIter(in_xml_root)
373
                    util.skip(iter_, xml_dom.is_text) # skip metadata
374
                    return iter_
375
                
376
                row_ct = process_rows(lambda row, i: root.appendChild(row),
377
                    get_rows(doc2rows))
378
            else:
379
                def doc2rows(in_xml_root):
380
                    rows = xpath.get(in_xml_root, in_root, limit=end)
381
                    if rows == []: warnings.warn(UserWarning('Map warning: '
382
                        'Root "'+in_root+'" not found in input'))
383
                    return rows
384
                
385
                def get_value(in_, row):
386
                    in_ = './{'+(','.join(strings.with_prefixes(
387
                        ['']+prefixes, in_)))+'}' # also with no prefix
388
                    nodes = xpath.get(row, in_, allow_rooted=False)
389
                    if nodes != []: return xml_dom.value(nodes[0])
390
                    else: return None
391
                
392
                row_ct = map_rows(get_value, get_rows(doc2rows))
393
        else: # input is CSV
394
            reader, col_names = csvs.reader_and_header(sys.stdin)
395
            row_ct = map_table(col_names, reader)
396
        
397
        return row_ct
398
    
399
    def process_inputs(root, row_ready):
400
        row_ct = 0
401
        for map_path in map_paths:
402
            row_ct += process_input(root, row_ready, map_path)
403
        return row_ct
404
    
405
    pool.share_vars(locals())
406
    if out_is_db:
407
        try:
408
            if redo: sql.empty_db(out_db)
409
            pool.share_vars(locals())
410
            
411
            def row_ready(row_num, input_row):
412
                row_str_ = [None]
413
                def row_str():
414
                    if row_str_[0] == None:
415
                        # Row # is interally 0-based, but 1-based to the user
416
                        row_str_[0] = (term.emph('row #:')+' '+str(row_num+1)
417
                            +'\n'+term.emph('input row:')+'\n'+str(input_row))
418
                        if verbose_errors: row_str_[0] += ('\n'
419
                            +term.emph('output row:')+'\n'+str(root))
420
                    return row_str_[0]
421
                
422
                if debug: log_debug(row_str()) # only calc if debug
423
                
424
                def on_error(e):
425
                    exc.add_msg(e, row_str())
426
                    ex_tracker.track(e, row_num, detail=verbose_errors)
427
                pool.share_vars(locals())
428
                
429
                row_root = root.cloneNode(True) # deep copy so don't modify root
430
                xml_func.process(row_root, on_error, is_rel_func, out_db)
431
                if debug: log_debug('Putting processed:\n'+str(row_root))
432
                    # only calc if debug
433
                if not xml_dom.is_empty(row_root):
434
                    assert xml_dom.has_one_child(row_root)
435
                    try:
436
                        sql.with_savepoint(out_db,
437
                            lambda: db_xml.put(out_db, row_root.firstChild,
438
                                row_ins_ct_ref, on_error, col_defaults))
439
                    except sql.DatabaseErrors, e: on_error(e)
440
            
441
            row_ct = process_inputs(root, row_ready)
442
            sys.stdout.write('Inserted '+str(row_ins_ct_ref[0])+
443
                ' new rows into database\n')
444
            sys.stdout.flush()
445
            
446
            # Consume asynchronous tasks
447
            pool.main_loop()
448
        finally: out_db.close()
449
    else:
450
        def on_error(e): ex_tracker.track(e)
451
        def row_ready(row_num, input_row): pass
452
        row_ct = process_inputs(root, row_ready)
453
        xml_func.process(root, on_error)
454
        if out_is_xml_ref[0]:
455
            doc.writexml(sys.stdout, **xml_dom.prettyxml_config)
456
        else: # output is CSV
457
            raise NotImplementedError('CSV output not supported yet')
458
    
459
    # Consume any asynchronous tasks not already consumed above
460
    pool.main_loop()
461
    
462
    profiler.stop(row_ct)
463
    if not by_col: ex_tracker.add_iters(row_ct) # only if errors are done by row
464
    log('Processed '+str(row_ct)+' input rows')
465
    log(profiler.msg())
466
    log(ex_tracker.msg())
467
    ex_tracker.exit()
468

    
469
def main():
470
    try: main_()
471
    except Parser.SyntaxError, e: raise SystemExit(str(e))
472

    
473
if __name__ == '__main__':
474
    profile_to = opts.get_env_var('profile_to', None)
475
    if profile_to != None:
476
        import cProfile
477
        sys.stderr.write('Profiling to '+profile_to+'\n')
478
        cProfile.run(main.func_code, profile_to)
479
    else: main()
(28-28/54)