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

    
14
sys.path.append(os.path.dirname(__file__)+"/../lib")
15

    
16
import csvs
17
import db_xml
18
import exc
19
import iters
20
import maps
21
import opts
22
import parallelproc
23
import Parser
24
import profiling
25
import sql
26
import sql_gen
27
import sql_io
28
import streams
29
import strings
30
import term
31
import util
32
import xpath
33
import xml_dom
34
import xml_func
35
import xml_parse
36

    
37
def get_with_prefix(map_, prefixes, key):
38
    '''Gets all entries for the given key with any of the given prefixes
39
    @return tuple(found_key, found_value)
40
    '''
41
    values = []
42
    for key_ in strings.with_prefixes(['']+prefixes, key): # also with no prefix
43
        try: value = map_[key_]
44
        except KeyError, e: continue # keep going
45
        values.append((key_, value))
46
    
47
    if values != []: return values
48
    else: raise e # re-raise last KeyError
49

    
50
def metadata_value(name): return None # this feature has been removed
51

    
52
def cleanup(val):
53
    if val == None: return val
54
    return util.none_if(strings.cleanup(strings.ustr(val)), u'', u'\\N')
55

    
56
def main_():
57
    env_names = []
58
    def usage_err():
59
        raise SystemExit('Usage: '+opts.env_usage(env_names, True)+' '
60
            +sys.argv[0]+' [map_path...] [<input] [>output]\n'
61
            'Note: Row #s start with 1')
62
    
63
    ## Get config from env vars
64
    
65
    # Modes
66
    test = opts.env_flag('test', False, env_names)
67
    commit = opts.env_flag('commit', False, env_names) and not test
68
        # never commit in test mode
69
    redo = opts.env_flag('redo', test, env_names) and not commit
70
        # never redo in commit mode (run `make schemas/reinstall` instead)
71
    
72
    # Ranges
73
    start = util.cast(int, opts.get_env_var('start', 1, env_names)) # 1-based
74
    # Make start interally 0-based.
75
    # It's 1-based to the user to match up with the staging table row #s.
76
    start -= 1
77
    if test: n_default = 1
78
    else: n_default = None
79
    n = util.cast(int, util.none_if(opts.get_env_var('n', n_default, env_names),
80
        u''))
81
    end = n
82
    if end != None: end += start
83
    
84
    # Debugging
85
    verbosity = util.cast(float, opts.get_env_var('verbosity', None, env_names))
86
    opts.get_env_var('profile_to', None, env_names) # add to env_names
87
    
88
    # DB
89
    def get_db_config(prefix):
90
        return opts.get_env_vars(sql.db_config_names, prefix, env_names)
91
    in_db_config = get_db_config('in')
92
    out_db_config = get_db_config('out')
93
    in_is_db = 'engine' in in_db_config
94
    out_is_db = 'engine' in out_db_config
95
    in_schema = opts.get_env_var('in_schema', None, env_names)
96
    in_table = opts.get_env_var('in_table', None, env_names)
97
    
98
    # Optimization
99
    cache_sql = opts.env_flag('cache_sql', True, env_names)
100
    by_col = in_db_config == out_db_config and opts.env_flag('by_col', False,
101
        env_names) # by-column optimization only applies if mapping to same DB
102
    if test: cpus_default = 0
103
    else: cpus_default = 0 # or None to use parallel processing by default
104
    cpus = util.cast(int, util.none_if(opts.get_env_var('cpus', cpus_default,
105
        env_names), u''))
106
    
107
    # Set default verbosity. Must happen after by_col is set.
108
    if verbosity == None: 
109
        if test: verbosity = 0.5 # automated tests should not be verbose
110
        elif by_col: verbosity = 3 # show all queries to assist profiling
111
        else: verbosity = 1.1 # just show row progress
112
    
113
    ##
114
    
115
    # Logging
116
    verbose_errors = test and verbosity > 0
117
    debug = verbosity >= 1.5
118
    def log(msg, level=1):
119
        '''Higher level -> more verbose'''
120
        if level <= verbosity:
121
            if verbosity <= 2:
122
                if level == 1.5: msg = '# '+msg # msg is Redmine list item
123
                elif msg.startswith('DB query:'): # remove extra debug info
124
                    first_line, nl, msg = msg.partition('\n')
125
            elif level > 1: msg = '['+str(level)+'] '+msg # include level in msg
126
            
127
            sys.stderr.write(strings.to_raw_str(msg.rstrip('\n')+'\n'))
128
    if debug: log_debug = lambda msg, level=2: log(msg, level)
129
    else: log_debug = sql.log_debug_none
130
    
131
    # Parse args
132
    map_paths = sys.argv[1:]
133
    if map_paths == []:
134
        if in_is_db or not out_is_db: usage_err()
135
        else: map_paths = [None]
136
    
137
    def connect_db(db_config):
138
        log('Connecting to '+sql.db_config_str(db_config))
139
        return sql.connect(db_config, caching=cache_sql, autocommit=commit,
140
            debug_temp=verbosity > 3 and commit, log_debug=log_debug)
141
    
142
    if end != None: end_str = str(end-1) # end is one past the last #
143
    else: end_str = 'end'
144
    log('Processing input rows '+str(start)+'-'+end_str)
145
    
146
    ex_tracker = exc.ExPercentTracker(iter_text='row')
147
    profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
148
    
149
    # Parallel processing
150
    pool = parallelproc.MultiProducerPool(cpus)
151
    log('Using '+str(pool.process_ct)+' parallel CPUs')
152
    
153
    doc = xml_dom.create_doc()
154
    root = doc.documentElement
155
    out_is_xml_ref = [False]
156
    in_label_ref = [None]
157
    def update_in_label():
158
        if in_label_ref[0] != None:
159
            xpath.get(root, '/_ignore/inLabel="'+in_label_ref[0]+'"', True)
160
    def prep_root():
161
        root.clear()
162
        update_in_label()
163
    prep_root()
164
    
165
    # Define before the out_is_db section because it's used by by_col
166
    row_ins_ct_ref = [0]
167
    
168
    if out_is_db:
169
        out_db = connect_db(out_db_config)
170
        rel_funcs = set(sql.tables(out_db, schema_like='%',
171
            table_like=r'\__%'))
172
    
173
    def process_input(root, row_ready, map_path):
174
        '''Inputs datasource to XML tree, mapping if needed'''
175
        # Load map header
176
        in_is_xpaths = True
177
        out_is_xpaths = True
178
        out_label = None
179
        if map_path != None:
180
            metadata = []
181
            mappings = []
182
            stream = open(map_path, 'rb')
183
            reader = csv.reader(stream)
184
            in_label, out_label = reader.next()[:2]
185
            
186
            def split_col_name(name):
187
                label, sep, root = name.partition(':')
188
                label, sep2, prefixes_str = label.partition('[')
189
                prefixes_str = strings.remove_suffix(']', prefixes_str)
190
                prefixes = strings.split(',', prefixes_str)
191
                return label, sep != '', root, prefixes
192
                    # extract datasrc from "datasrc[data_format]"
193
            
194
            in_label, in_root, prefixes = maps.col_info(in_label)
195
            in_is_xpaths = in_root != None
196
            in_label_ref[0] = in_label
197
            update_in_label()
198
            out_label, out_root = maps.col_info(out_label)[:2]
199
            out_is_xpaths = out_root != None
200
            if out_is_xpaths: has_types = out_root.find('/*s/') >= 0
201
                # outer elements are types
202
            
203
            for row in reader:
204
                in_, out = row[:2]
205
                if out != '': mappings.append([in_, out_root+out])
206
            
207
            stream.close()
208
            
209
            root.ownerDocument.documentElement.tagName = out_label
210
        in_is_xml = in_is_xpaths and not in_is_db
211
        out_is_xml_ref[0] = out_is_xpaths and not out_is_db
212
        
213
        def process_rows(process_row, rows, rows_start=0):
214
            '''Processes input rows      
215
            @param process_row(in_row, i)
216
            @rows_start The (0-based) row # of the first row in rows. Set this
217
                only if the pre-start rows have already been skipped.
218
            '''
219
            rows = iter(rows)
220
            
221
            if end != None: row_nums = xrange(rows_start, end)
222
            else: row_nums = itertools.count(rows_start)
223
            i = -1
224
            for i in row_nums:
225
                try: row = rows.next()
226
                except StopIteration:
227
                    i -= 1 # last row # didn't count
228
                    break # no more rows
229
                if i < start: continue # not at start row yet
230
                
231
                # Row # is interally 0-based, but 1-based to the user
232
                log('Processing input row #'+str(i+1), level=1.1)
233
                process_row(row, i)
234
                row_ready(i, row)
235
            row_ct = i-start+1
236
            return row_ct
237
        
238
        def map_rows(get_value, rows, **kw_args):
239
            '''Maps input rows
240
            @param get_value(in_, row):str
241
            '''
242
            # Prevent collisions if multiple inputs mapping to same output
243
            outputs_idxs = dict()
244
            for i, mapping in enumerate(mappings):
245
                in_, out = mapping
246
                default = util.NamedTuple(count=1, first=i)
247
                idxs = outputs_idxs.setdefault(out, default)
248
                if idxs is not default: # key existed, so there was a collision
249
                    if idxs.count == 1: # first key does not yet have /_alt/#
250
                        mappings[idxs.first][1] += '/_alt/0'
251
                    mappings[i][1] += '/_alt/'+str(idxs.count)
252
                    idxs.count += 1
253
            
254
            id_node = None
255
            if out_is_db:
256
                for i, mapping in enumerate(mappings):
257
                    in_, out = mapping
258
                    # All put_obj()s should return the same id_node
259
                    nodes, id_node = xpath.put_obj(root, out, '-1', has_types,
260
                        '$'+str(in_)) # value is placeholder that documents name
261
                    mappings[i] = [in_, nodes]
262
                assert id_node != None
263
                
264
                if debug: # only calc if debug
265
                    log_debug('Put template:\n'+str(root))
266
            
267
            def process_row(row, i):
268
                row_id = str(i)
269
                if id_node != None: xml_dom.set_value(id_node, row_id)
270
                for in_, out in mappings:
271
                    log_debug('Getting '+str(in_))
272
                    value = metadata_value(in_)
273
                    if value == None: value = cleanup(get_value(in_, row))
274
                    log_debug('Putting '+repr(value)+' to '+str(out))
275
                    if out_is_db: # out is list of XML nodes
276
                        for node in out: xml_dom.set_value(node, value)
277
                    elif value != None: # out is XPath
278
                        xpath.put_obj(root, out, row_id, has_types, value)
279
            return process_rows(process_row, rows, **kw_args)
280
        
281
        def map_table(col_names, rows, **kw_args):
282
            col_names_ct = len(col_names)
283
            col_idxs = util.list_flip(col_names)
284
            
285
            # Resolve prefixes
286
            mappings_orig = mappings[:] # save a copy
287
            mappings[:] = [] # empty existing elements
288
            for in_, out in mappings_orig:
289
                if metadata_value(in_) == None:
290
                    try: cols = get_with_prefix(col_idxs, prefixes, in_)
291
                    except KeyError: pass
292
                    else: mappings[len(mappings):] = [[db_xml.ColRef(*col), out]
293
                        for col in cols] # can't use += because that uses =
294
            
295
            def get_value(in_, row): return row.list[in_.idx]
296
            def wrap_row(row):
297
                return util.ListDict(util.list_as_length(row, col_names_ct),
298
                    col_names, col_idxs) # handle CSV rows of different lengths
299
            
300
            return map_rows(get_value, util.WrapIter(wrap_row, rows), **kw_args)
301
        
302
        if in_is_db:
303
            def on_error(e): ex_tracker.track(e)
304
            
305
            if by_col: in_db = out_db
306
            else: in_db = connect_db(in_db_config)
307
            
308
            # Get table and schema name
309
            schema = in_schema # modified, so can't have same name as outer var
310
            table = in_table # modified, so can't have same name as outer var
311
            if table == None:
312
                assert in_is_xpaths
313
                schema, sep, table = in_root.partition('.')
314
                if sep == '': # only the table name was specified
315
                    table = schema
316
                    schema = None
317
            table = sql_gen.Table(table, schema)
318
            
319
            # Fetch rows
320
            if by_col: limit = 0 # only fetch column names
321
            else: limit = n
322
            cur = sql.select(in_db, table, limit=limit, start=start,
323
                cacheable=False)
324
            col_names = list(sql.col_names(cur))
325
            
326
            if by_col:
327
                map_table(col_names, []) # just create the template
328
                
329
                if n == None: # doing full import
330
                    log('Clearing errors table')
331
                    sql.truncate(in_db, sql_io.errors_table(in_db, table))
332
                
333
                # Strip XML functions not in the DB
334
                special_funcs = db_xml.put_table_special_funcs | rel_funcs
335
                xml_func.process(root, rel_funcs=special_funcs)
336
                if debug: log_debug('Putting stripped:\n'+str(root))
337
                    # only calc if debug
338
                
339
                # Import rows
340
                in_row_ct_ref = [0]
341
                db_xml.put_table(in_db, root.firstChild, table, in_row_ct_ref,
342
                    row_ins_ct_ref, n, start, on_error)
343
                row_ct = in_row_ct_ref[0]
344
            else:
345
                # Use normal by-row method
346
                row_ct = map_table(col_names, sql.rows(cur), rows_start=start)
347
                    # rows_start: pre-start rows have been skipped
348
                
349
                in_db.db.close()
350
        elif in_is_xml:
351
            stdin = streams.LineCountStream(sys.stdin)
352
            def on_error(e):
353
                exc.add_msg(e, term.emph('input line #:')+' '
354
                    +str(stdin.line_num))
355
                ex_tracker.track(e)
356
            
357
            def get_rows(doc2rows):
358
                return iters.flatten(itertools.imap(doc2rows,
359
                    xml_parse.docs_iter(stdin, on_error)))
360
            
361
            if map_path == None:
362
                def doc2rows(in_xml_root):
363
                    iter_ = xml_dom.NodeElemIter(in_xml_root)
364
                    util.skip(iter_, xml_dom.is_text) # skip metadata
365
                    return iter_
366
                
367
                row_ct = process_rows(lambda row, i: root.appendChild(row),
368
                    get_rows(doc2rows))
369
            else:
370
                def doc2rows(in_xml_root):
371
                    rows = xpath.get(in_xml_root, in_root, limit=end)
372
                    if rows == []: raise SystemExit('Map error: Root "'
373
                        +in_root+'" not found in input')
374
                    return rows
375
                
376
                def get_value(in_, row):
377
                    in_ = './{'+(','.join(strings.with_prefixes(
378
                        ['']+prefixes, in_)))+'}' # also with no prefix
379
                    nodes = xpath.get(row, in_, allow_rooted=False)
380
                    if nodes != []: return xml_dom.value(nodes[0])
381
                    else: return None
382
                
383
                row_ct = map_rows(get_value, get_rows(doc2rows))
384
        else: # input is CSV
385
            map_ = dict(mappings)
386
            reader, col_names = csvs.reader_and_header(sys.stdin)
387
            row_ct = map_table(col_names, reader)
388
        
389
        return row_ct
390
    
391
    def process_inputs(root, row_ready):
392
        row_ct = 0
393
        for map_path in map_paths:
394
            row_ct += process_input(root, row_ready, map_path)
395
        return row_ct
396
    
397
    pool.share_vars(locals())
398
    if out_is_db:
399
        try:
400
            if redo: sql.empty_db(out_db)
401
            pool.share_vars(locals())
402
            
403
            def row_ready(row_num, input_row):
404
                row_str_ = [None]
405
                def row_str():
406
                    if row_str_[0] == None:
407
                        # Row # is interally 0-based, but 1-based to the user
408
                        row_str_[0] = (term.emph('row #:')+' '+str(row_num+1)
409
                            +'\n'+term.emph('input row:')+'\n'+str(input_row))
410
                        if verbose_errors: row_str_[0] += ('\n'
411
                            +term.emph('output row:')+'\n'+str(root))
412
                    return row_str_[0]
413
                
414
                if debug: log_debug(row_str()) # only calc if debug
415
                
416
                def on_error(e):
417
                    exc.add_msg(e, row_str())
418
                    ex_tracker.track(e, row_num, detail=verbose_errors)
419
                pool.share_vars(locals())
420
                
421
                row_root = root.cloneNode(True) # deep copy so don't modify root
422
                xml_func.process(row_root, on_error, rel_funcs, out_db)
423
                if not xml_dom.is_empty(row_root):
424
                    assert xml_dom.has_one_child(row_root)
425
                    try:
426
                        sql.with_savepoint(out_db,
427
                            lambda: db_xml.put(out_db, row_root.firstChild,
428
                                row_ins_ct_ref, on_error))
429
                    except sql.DatabaseErrors, e: on_error(e)
430
            
431
            row_ct = process_inputs(root, row_ready)
432
            sys.stdout.write('Inserted '+str(row_ins_ct_ref[0])+
433
                ' new rows into database\n')
434
            
435
            # Consume asynchronous tasks
436
            pool.main_loop()
437
        finally: out_db.close()
438
    else:
439
        def on_error(e): ex_tracker.track(e)
440
        def row_ready(row_num, input_row): pass
441
        row_ct = process_inputs(root, row_ready)
442
        xml_func.process(root, on_error)
443
        if out_is_xml_ref[0]:
444
            doc.writexml(sys.stdout, **xml_dom.prettyxml_config)
445
        else: # output is CSV
446
            raise NotImplementedError('CSV output not supported yet')
447
    
448
    # Consume any asynchronous tasks not already consumed above
449
    pool.main_loop()
450
    
451
    profiler.stop(row_ct)
452
    if not by_col: ex_tracker.add_iters(row_ct) # only if errors are done by row
453
    log('Processed '+str(row_ct)+' input rows')
454
    log(profiler.msg())
455
    log(ex_tracker.msg())
456
    ex_tracker.exit()
457

    
458
def main():
459
    try: main_()
460
    except Parser.SyntaxError, e: raise SystemExit(str(e))
461

    
462
if __name__ == '__main__':
463
    profile_to = opts.get_env_var('profile_to', None)
464
    if profile_to != None:
465
        import cProfile
466
        sys.stderr.write('Profiling to '+profile_to+'\n')
467
        cProfile.run(main.func_code, profile_to)
468
    else: main()
(26-26/51)