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 streams
28
import strings
29
import term
30
import util
31
import xpath
32
import xml_dom
33
import xml_func
34
import xml_parse
35

    
36
def get_with_prefix(map_, prefixes, key):
37
    '''Gets all entries for the given key with any of the given prefixes
38
    @return tuple(found_key, found_value)
39
    '''
40
    values = []
41
    for key_ in strings.with_prefixes(['']+prefixes, key): # also with no prefix
42
        try: value = map_[key_]
43
        except KeyError, e: continue # keep going
44
        values.append((key_, value))
45
    
46
    if values != []: return values
47
    else: raise e # re-raise last KeyError
48

    
49
def metadata_value(name): return None # this feature has been removed
50

    
51
def cleanup(val):
52
    if val == None: return val
53
    return util.none_if(strings.cleanup(strings.ustr(val)), u'', u'\\N')
54

    
55
def main_():
56
    env_names = []
57
    def usage_err():
58
        raise SystemExit('Usage: '+opts.env_usage(env_names, True)+' '
59
            +sys.argv[0]+' [map_path...] [<input] [>output]\n'
60
            'Note: Row #s start with 1')
61
    
62
    ## Get config from env vars
63
    
64
    # Modes
65
    test = opts.env_flag('test', False, env_names)
66
    commit = opts.env_flag('commit', False, env_names) and not test
67
        # never commit in test mode
68
    redo = opts.env_flag('redo', test, env_names) and not commit
69
        # never redo in commit mode (manually run `make empty_db` instead)
70
    
71
    # Ranges
72
    start = util.cast(int, opts.get_env_var('start', 1, env_names)) # 1-based
73
    # Make start interally 0-based.
74
    # It's 1-based to the user to match up with the staging table row #s.
75
    start -= 1
76
    if test: n_default = 1
77
    else: n_default = None
78
    n = util.cast(int, util.none_if(opts.get_env_var('n', n_default, env_names),
79
        u''))
80
    end = n
81
    if end != None: end += start
82
    
83
    # Debugging
84
    if test: default_verbosity = 0.5
85
    else: default_verbosity = 1
86
    verbosity = util.cast(float, opts.get_env_var('verbosity',
87
        default_verbosity, 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
    
100
    # Optimization
101
    cache_sql = opts.env_flag('cache_sql', True, env_names)
102
    by_col = in_db_config == out_db_config and opts.env_flag('by_col', False,
103
        env_names) # by-column optimization only applies if mapping to same DB
104
    if test: cpus_default = 0
105
    else: cpus_default = 0 # or None to use parallel processing by default
106
    cpus = util.cast(int, util.none_if(opts.get_env_var('cpus', cpus_default,
107
        env_names), u''))
108
    
109
    ##
110
    
111
    # Logging
112
    verbose_errors = test and verbosity > 0
113
    debug = verbosity > 1
114
    def log(msg, level=1):
115
        '''Higher level -> more verbose'''
116
        if level <= verbosity:
117
            if verbosity <= 2:
118
                if level == 1.5: msg = '# '+msg # msg is Redmine list item
119
                elif msg.startswith('DB query:'): # remove extra debug info
120
                    first_line, nl, msg = msg.partition('\n')
121
            elif level > 1: msg = '['+str(level)+'] '+msg # include level in msg
122
            
123
            sys.stderr.write(msg.rstrip('\n')+'\n')
124
    if debug: log_debug = lambda msg, level=2: log(msg, level)
125
    else: log_debug = sql.log_debug_none
126
    
127
    # Parse args
128
    map_paths = sys.argv[1:]
129
    if map_paths == []:
130
        if in_is_db or not out_is_db: usage_err()
131
        else: map_paths = [None]
132
    
133
    def connect_db(db_config):
134
        log('Connecting to '+sql.db_config_str(db_config))
135
        return sql.connect(db_config, caching=cache_sql,
136
            autocommit=debug and commit, log_debug=log_debug)
137
    
138
    if end != None: end_str = str(end-1) # end is one past the last #
139
    else: end_str = 'end'
140
    log('Processing input rows '+str(start)+'-'+end_str)
141
    
142
    ex_tracker = exc.ExPercentTracker(iter_text='row')
143
    profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
144
    
145
    # Parallel processing
146
    pool = parallelproc.MultiProducerPool(cpus)
147
    log('Using '+str(pool.process_ct)+' parallel CPUs')
148
    
149
    doc = xml_dom.create_doc()
150
    root = doc.documentElement
151
    out_is_xml_ref = [False]
152
    in_label_ref = [None]
153
    def update_in_label():
154
        if in_label_ref[0] != None:
155
            xpath.get(root, '/_ignore/inLabel="'+in_label_ref[0]+'"', True)
156
    def prep_root():
157
        root.clear()
158
        update_in_label()
159
    prep_root()
160
    
161
    # Define before the out_is_db section because it's used by by_col
162
    row_ins_ct_ref = [0]
163
    
164
    def process_input(root, row_ready, map_path):
165
        '''Inputs datasource to XML tree, mapping if needed'''
166
        # Load map header
167
        in_is_xpaths = True
168
        out_is_xpaths = True
169
        out_label = None
170
        if map_path != None:
171
            metadata = []
172
            mappings = []
173
            stream = open(map_path, 'rb')
174
            reader = csv.reader(stream)
175
            in_label, out_label = reader.next()[:2]
176
            
177
            def split_col_name(name):
178
                label, sep, root = name.partition(':')
179
                label, sep2, prefixes_str = label.partition('[')
180
                prefixes_str = strings.remove_suffix(']', prefixes_str)
181
                prefixes = strings.split(',', prefixes_str)
182
                return label, sep != '', root, prefixes
183
                    # extract datasrc from "datasrc[data_format]"
184
            
185
            in_label, in_root, prefixes = maps.col_info(in_label)
186
            in_is_xpaths = in_root != None
187
            in_label_ref[0] = in_label
188
            update_in_label()
189
            out_label, out_root = maps.col_info(out_label)[:2]
190
            out_is_xpaths = out_root != None
191
            if out_is_xpaths: has_types = out_root.find('/*s/') >= 0
192
                # outer elements are types
193
            
194
            for row in reader:
195
                in_, out = row[:2]
196
                if out != '': mappings.append([in_, out_root+out])
197
            
198
            stream.close()
199
            
200
            root.ownerDocument.documentElement.tagName = out_label
201
        in_is_xml = in_is_xpaths and not in_is_db
202
        out_is_xml_ref[0] = out_is_xpaths and not out_is_db
203
        
204
        def process_rows(process_row, rows, rows_start=0):
205
            '''Processes input rows      
206
            @param process_row(in_row, i)
207
            @rows_start The (0-based) row # of the first row in rows. Set this
208
                only if the pre-start rows have already been skipped.
209
            '''
210
            rows = iter(rows)
211
            
212
            if end != None: row_nums = xrange(rows_start, end)
213
            else: row_nums = itertools.count(rows_start)
214
            i = -1
215
            for i in row_nums:
216
                try: row = rows.next()
217
                except StopIteration:
218
                    i -= 1 # last row # didn't count
219
                    break # no more rows
220
                if i < start: continue # not at start row yet
221
                
222
                process_row(row, i)
223
                row_ready(i, row)
224
            row_ct = i-start+1
225
            return row_ct
226
        
227
        def map_rows(get_value, rows, **kw_args):
228
            '''Maps input rows
229
            @param get_value(in_, row):str
230
            '''
231
            # Prevent collisions if multiple inputs mapping to same output
232
            outputs_idxs = dict()
233
            for i, mapping in enumerate(mappings):
234
                in_, out = mapping
235
                default = util.NamedTuple(count=1, first=i)
236
                idxs = outputs_idxs.setdefault(out, default)
237
                if idxs is not default: # key existed, so there was a collision
238
                    if idxs.count == 1: # first key does not yet have /_alt/#
239
                        mappings[idxs.first][1] += '/_alt/0'
240
                    mappings[i][1] += '/_alt/'+str(idxs.count)
241
                    idxs.count += 1
242
            
243
            id_node = None
244
            if out_is_db:
245
                for i, mapping in enumerate(mappings):
246
                    in_, out = mapping
247
                    # All put_obj()s should return the same id_node
248
                    nodes, id_node = xpath.put_obj(root, out, '-1', has_types,
249
                        '$'+str(in_)) # value is placeholder that documents name
250
                    mappings[i] = [in_, nodes]
251
                assert id_node != None
252
                
253
                if debug: # only calc if debug
254
                    log_debug('Put template:\n'+str(root))
255
            
256
            def process_row(row, i):
257
                row_id = str(i)
258
                if id_node != None: xml_dom.set_value(id_node, row_id)
259
                for in_, out in mappings:
260
                    log_debug('Getting '+str(in_))
261
                    value = metadata_value(in_)
262
                    if value == None: value = cleanup(get_value(in_, row))
263
                    log_debug('Putting '+repr(value)+' to '+str(out))
264
                    if out_is_db: # out is list of XML nodes
265
                        for node in out: xml_dom.set_value(node, value)
266
                    elif value != None: # out is XPath
267
                        xpath.put_obj(root, out, row_id, has_types, value)
268
            return process_rows(process_row, rows, **kw_args)
269
        
270
        def map_table(col_names, rows, **kw_args):
271
            col_names_ct = len(col_names)
272
            col_idxs = util.list_flip(col_names)
273
            
274
            # Resolve prefixes
275
            mappings_orig = mappings[:] # save a copy
276
            mappings[:] = [] # empty existing elements
277
            for in_, out in mappings_orig:
278
                if metadata_value(in_) == None:
279
                    try: cols = get_with_prefix(col_idxs, prefixes, in_)
280
                    except KeyError: pass
281
                    else: mappings[len(mappings):] = [[db_xml.ColRef(*col), out]
282
                        for col in cols] # can't use += because that uses =
283
            
284
            def get_value(in_, row): return row.list[in_.idx]
285
            def wrap_row(row):
286
                return util.ListDict(util.list_as_length(row, col_names_ct),
287
                    col_names, col_idxs) # handle CSV rows of different lengths
288
            
289
            return map_rows(get_value, util.WrapIter(wrap_row, rows), **kw_args)
290
        
291
        stdin = streams.LineCountStream(sys.stdin)
292
        def on_error(e):
293
            exc.add_msg(e, term.emph('input line #:')+' '+str(stdin.line_num))
294
            ex_tracker.track(e)
295
        
296
        if in_is_db:
297
            in_db = connect_db(in_db_config)
298
            
299
            # Get table and schema name
300
            schema = in_schema # modified, so can't have same name as outer var
301
            table = in_table # modified, so can't have same name as outer var
302
            if table == None:
303
                assert in_is_xpaths
304
                schema, sep, table = in_root.partition('.')
305
                if sep == '': # only the table name was specified
306
                    table = schema
307
                    schema = None
308
            table = sql_gen.Table(table, schema)
309
            
310
            # Fetch rows
311
            if by_col: limit = 0 # only fetch column names
312
            else: limit = n
313
            cur = sql.select(in_db, table, limit=limit, start=start,
314
                cacheable=False)
315
            col_names = list(sql.col_names(cur))
316
            
317
            if by_col:
318
                map_table(col_names, []) # just create the template
319
                
320
                # Strip XML functions not in the DB
321
                preserve_funcs = copy.copy(db_xml.put_table_special_funcs)
322
                preserve_funcs.add(sql.tables(in_db, schema_like='%',
323
                    table_like=r'\__%'))
324
                xml_func.strip(root, preserve_funcs)
325
                if debug: log_debug('Putting stripped:\n'+str(root))
326
                    # only calc if debug
327
                
328
                # Import rows
329
                in_row_ct_ref = [0]
330
                db_xml.put_table(in_db, root.firstChild, table, commit,
331
                    in_row_ct_ref, row_ins_ct_ref, n, start)
332
                row_ct = in_row_ct_ref[0]
333
            else:
334
                # Use normal by-row method
335
                row_ct = map_table(col_names, sql.rows(cur), rows_start=start)
336
                    # rows_start: pre-start rows have been skipped
337
            
338
            in_db.db.close()
339
        elif in_is_xml:
340
            def get_rows(doc2rows):
341
                return iters.flatten(itertools.imap(doc2rows,
342
                    xml_parse.docs_iter(stdin, on_error)))
343
            
344
            if map_path == None:
345
                def doc2rows(in_xml_root):
346
                    iter_ = xml_dom.NodeElemIter(in_xml_root)
347
                    util.skip(iter_, xml_dom.is_text) # skip metadata
348
                    return iter_
349
                
350
                row_ct = process_rows(lambda row, i: root.appendChild(row),
351
                    get_rows(doc2rows))
352
            else:
353
                def doc2rows(in_xml_root):
354
                    rows = xpath.get(in_xml_root, in_root, limit=end)
355
                    if rows == []: raise SystemExit('Map error: Root "'
356
                        +in_root+'" not found in input')
357
                    return rows
358
                
359
                def get_value(in_, row):
360
                    in_ = './{'+(','.join(strings.with_prefixes(
361
                        ['']+prefixes, in_)))+'}' # also with no prefix
362
                    nodes = xpath.get(row, in_, allow_rooted=False)
363
                    if nodes != []: return xml_dom.value(nodes[0])
364
                    else: return None
365
                
366
                row_ct = map_rows(get_value, get_rows(doc2rows))
367
        else: # input is CSV
368
            map_ = dict(mappings)
369
            reader, col_names = csvs.reader_and_header(sys.stdin)
370
            row_ct = map_table(col_names, reader)
371
        
372
        return row_ct
373
    
374
    def process_inputs(root, row_ready):
375
        row_ct = 0
376
        for map_path in map_paths:
377
            row_ct += process_input(root, row_ready, map_path)
378
        return row_ct
379
    
380
    pool.share_vars(locals())
381
    if out_is_db:
382
        out_db = connect_db(out_db_config)
383
        try:
384
            if redo: sql.empty_db(out_db)
385
            pool.share_vars(locals())
386
            
387
            def row_ready(row_num, input_row):
388
                row_str_ = [None]
389
                def row_str():
390
                    if row_str_[0] == None:
391
                        # Row # is interally 0-based, but 1-based to the user
392
                        row_str_[0] = (term.emph('row #:')+' '+str(row_num+1)
393
                            +'\n'+term.emph('input row:')+'\n'+str(input_row))
394
                        if verbose_errors: row_str_[0] += ('\n'
395
                            +term.emph('output row:')+'\n'+str(root))
396
                    return row_str_[0]
397
                
398
                if debug: log_debug(row_str()) # only calc if debug
399
                
400
                def on_error(e):
401
                    exc.add_msg(e, row_str())
402
                    ex_tracker.track(e, row_num, detail=verbose_errors)
403
                pool.share_vars(locals())
404
                
405
                row_root = root.cloneNode(True) # deep copy so don't modify root
406
                xml_func.process(row_root, on_error, out_db)
407
                if not xml_dom.is_empty(row_root):
408
                    assert xml_dom.has_one_child(row_root)
409
                    try:
410
                        sql.with_savepoint(out_db,
411
                            lambda: db_xml.put(out_db, row_root.firstChild,
412
                                row_ins_ct_ref, on_error))
413
                        if commit: out_db.db.commit()
414
                    except sql.DatabaseErrors, e: on_error(e)
415
            
416
            row_ct = process_inputs(root, row_ready)
417
            sys.stdout.write('Inserted '+str(row_ins_ct_ref[0])+
418
                ' new rows into database\n')
419
            
420
            # Consume asynchronous tasks
421
            pool.main_loop()
422
        finally:
423
            if out_db.connected():
424
                out_db.db.rollback()
425
                out_db.db.close()
426
    else:
427
        def on_error(e): ex_tracker.track(e)
428
        def row_ready(row_num, input_row): pass
429
        row_ct = process_inputs(root, row_ready)
430
        xml_func.process(root, on_error)
431
        if out_is_xml_ref[0]:
432
            doc.writexml(sys.stdout, **xml_dom.prettyxml_config)
433
        else: # output is CSV
434
            raise NotImplementedError('CSV output not supported yet')
435
    
436
    # Consume any asynchronous tasks not already consumed above
437
    pool.main_loop()
438
    
439
    profiler.stop(row_ct)
440
    ex_tracker.add_iters(row_ct)
441
    log('Processed '+str(row_ct)+' input rows')
442
    log(profiler.msg())
443
    log(ex_tracker.msg())
444
    ex_tracker.exit()
445

    
446
def main():
447
    try: main_()
448
    except Parser.SyntaxError, e: raise SystemExit(str(e))
449

    
450
if __name__ == '__main__':
451
    profile_to = opts.get_env_var('profile_to', None)
452
    if profile_to != None:
453
        import cProfile
454
        sys.stderr.write('Profiling to '+profile_to+'\n')
455
        cProfile.run(main.func_code, profile_to)
456
    else: main()
(26-26/49)