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

    
13
sys.path.append(os.path.dirname(__file__)+"/../lib")
14

    
15
import csvs
16
import db_xml
17
import exc
18
import iters
19
import maps
20
import opts
21
import parallelproc
22
import Parser
23
import profiling
24
import sql
25
import sql_gen
26
import streams
27
import strings
28
import term
29
import util
30
import xpath
31
import xml_dom
32
import xml_func
33
import xml_parse
34

    
35
def get_with_prefix(map_, prefixes, key):
36
    '''Gets all entries for the given key with any of the given prefixes
37
    @return tuple(found_key, found_value)
38
    '''
39
    values = []
40
    for key_ in strings.with_prefixes(['']+prefixes, key): # also with no prefix
41
        try: value = map_[key_]
42
        except KeyError, e: continue # keep going
43
        values.append((key_, value))
44
    
45
    if values != []: return values
46
    else: raise e # re-raise last KeyError
47

    
48
def metadata_value(name): return None # this feature has been removed
49

    
50
def cleanup(val):
51
    if val == None: return val
52
    return util.none_if(strings.cleanup(strings.ustr(val)), u'', u'\\N')
53

    
54
def main_():
55
    env_names = []
56
    def usage_err():
57
        raise SystemExit('Usage: '+opts.env_usage(env_names, True)+' '
58
            +sys.argv[0]+' [map_path...] [<input] [>output]\n'
59
            'Note: Row #s start with 1')
60
    
61
    ## Get config from env vars
62
    
63
    # Modes
64
    test = opts.env_flag('test', False, env_names)
65
    commit = opts.env_flag('commit', False, env_names) and not test
66
        # never commit in test mode
67
    redo = opts.env_flag('redo', test, env_names) and not commit
68
        # never redo in commit mode (manually run `make empty_db` instead)
69
    
70
    # Ranges
71
    start = util.cast(int, opts.get_env_var('start', 1, env_names)) # 1-based
72
    # Make start interally 0-based.
73
    # It's 1-based to the user to match up with the staging table row #s.
74
    start -= 1
75
    if test: n_default = 1
76
    else: n_default = None
77
    n = util.cast(int, util.none_if(opts.get_env_var('n', n_default, env_names),
78
        u''))
79
    end = n
80
    if end != None: end += start
81
    
82
    # Debugging
83
    if test: default_verbosity = 0.5
84
    else: default_verbosity = 1
85
    verbosity = util.cast(float, opts.get_env_var('verbosity',
86
        default_verbosity, env_names))
87
    opts.get_env_var('profile_to', None, env_names) # add to env_names
88
    
89
    # DB
90
    def get_db_config(prefix):
91
        return opts.get_env_vars(sql.db_config_names, prefix, env_names)
92
    in_db_config = get_db_config('in')
93
    out_db_config = get_db_config('out')
94
    in_is_db = 'engine' in in_db_config
95
    out_is_db = 'engine' in out_db_config
96
    in_schema = opts.get_env_var('in_schema', None, env_names)
97
    in_table = opts.get_env_var('in_table', None, env_names)
98
    
99
    # Optimization
100
    cache_sql = opts.env_flag('cache_sql', True, env_names)
101
    by_col = in_db_config == out_db_config and opts.env_flag('by_col', False,
102
        env_names) # by-column optimization only applies if mapping to same DB
103
    if test: cpus_default = 0
104
    else: cpus_default = 0 # or None to use parallel processing by default
105
    cpus = util.cast(int, util.none_if(opts.get_env_var('cpus', cpus_default,
106
        env_names), u''))
107
    
108
    ##
109
    
110
    # Logging
111
    verbose_errors = test and verbosity > 0
112
    debug = verbosity > 1
113
    def log(msg, level=1):
114
        '''Higher level -> more verbose'''
115
        if level <= verbosity:
116
            if level > 1: msg = '['+str(level)+'] '+msg # include level in msg
117
            sys.stderr.write(msg+'\n')
118
    if debug: log_debug = lambda msg, level=2: log(msg, level)
119
    else: log_debug = sql.log_debug_none
120
    
121
    # Parse args
122
    map_paths = sys.argv[1:]
123
    if map_paths == []:
124
        if in_is_db or not out_is_db: usage_err()
125
        else: map_paths = [None]
126
    
127
    def connect_db(db_config):
128
        log('Connecting to '+sql.db_config_str(db_config))
129
        return sql.connect(db_config, caching=cache_sql,
130
            autocommit=debug and commit, log_debug=log_debug)
131
    
132
    if end != None: end_str = str(end-1) # end is one past the last #
133
    else: end_str = 'end'
134
    log('Processing input rows '+str(start)+'-'+end_str)
135
    
136
    ex_tracker = exc.ExPercentTracker(iter_text='row')
137
    profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
138
    
139
    # Parallel processing
140
    pool = parallelproc.MultiProducerPool(cpus)
141
    log('Using '+str(pool.process_ct)+' parallel CPUs')
142
    
143
    doc = xml_dom.create_doc()
144
    root = doc.documentElement
145
    out_is_xml_ref = [False]
146
    in_label_ref = [None]
147
    def update_in_label():
148
        if in_label_ref[0] != None:
149
            xpath.get(root, '/_ignore/inLabel="'+in_label_ref[0]+'"', True)
150
    def prep_root():
151
        root.clear()
152
        update_in_label()
153
    prep_root()
154
    
155
    # Define before the out_is_db section because it's used by by_col
156
    row_ins_ct_ref = [0]
157
    
158
    def process_input(root, row_ready, map_path):
159
        '''Inputs datasource to XML tree, mapping if needed'''
160
        # Load map header
161
        in_is_xpaths = True
162
        out_is_xpaths = True
163
        out_label = None
164
        if map_path != None:
165
            metadata = []
166
            mappings = []
167
            stream = open(map_path, 'rb')
168
            reader = csv.reader(stream)
169
            in_label, out_label = reader.next()[:2]
170
            
171
            def split_col_name(name):
172
                label, sep, root = name.partition(':')
173
                label, sep2, prefixes_str = label.partition('[')
174
                prefixes_str = strings.remove_suffix(']', prefixes_str)
175
                prefixes = strings.split(',', prefixes_str)
176
                return label, sep != '', root, prefixes
177
                    # extract datasrc from "datasrc[data_format]"
178
            
179
            in_label, in_root, prefixes = maps.col_info(in_label)
180
            in_is_xpaths = in_root != None
181
            in_label_ref[0] = in_label
182
            update_in_label()
183
            out_label, out_root = maps.col_info(out_label)[:2]
184
            out_is_xpaths = out_root != None
185
            if out_is_xpaths: has_types = out_root.find('/*s/') >= 0
186
                # outer elements are types
187
            
188
            for row in reader:
189
                in_, out = row[:2]
190
                if out != '': mappings.append([in_, out_root+out])
191
            
192
            stream.close()
193
            
194
            root.ownerDocument.documentElement.tagName = out_label
195
        in_is_xml = in_is_xpaths and not in_is_db
196
        out_is_xml_ref[0] = out_is_xpaths and not out_is_db
197
        
198
        def process_rows(process_row, rows, rows_start=0):
199
            '''Processes input rows      
200
            @param process_row(in_row, i)
201
            @rows_start The (0-based) row # of the first row in rows. Set this
202
                only if the pre-start rows have already been skipped.
203
            '''
204
            rows = iter(rows)
205
            
206
            if end != None: row_nums = xrange(rows_start, end)
207
            else: row_nums = itertools.count(rows_start)
208
            i = -1
209
            for i in row_nums:
210
                try: row = rows.next()
211
                except StopIteration:
212
                    i -= 1 # last row # didn't count
213
                    break # no more rows
214
                if i < start: continue # not at start row yet
215
                
216
                process_row(row, i)
217
                row_ready(i, row)
218
            row_ct = i-start+1
219
            return row_ct
220
        
221
        def map_rows(get_value, rows, **kw_args):
222
            '''Maps input rows
223
            @param get_value(in_, row):str
224
            '''
225
            # Prevent collisions if multiple inputs mapping to same output
226
            outputs_idxs = dict()
227
            for i, mapping in enumerate(mappings):
228
                in_, out = mapping
229
                default = util.NamedTuple(count=1, first=i)
230
                idxs = outputs_idxs.setdefault(out, default)
231
                if idxs is not default: # key existed, so there was a collision
232
                    if idxs.count == 1: # first key does not yet have /_alt/#
233
                        mappings[idxs.first][1] += '/_alt/0'
234
                    mappings[i][1] += '/_alt/'+str(idxs.count)
235
                    idxs.count += 1
236
            
237
            id_node = None
238
            if out_is_db:
239
                for i, mapping in enumerate(mappings):
240
                    in_, out = mapping
241
                    # All put_obj()s should return the same id_node
242
                    nodes, id_node = xpath.put_obj(root, out, '-1', has_types,
243
                        '$'+str(in_)) # value is placeholder that documents name
244
                    mappings[i] = [in_, nodes]
245
                assert id_node != None
246
                
247
                if debug: # only calc if debug
248
                    log_debug('Put template:\n'+str(root))
249
            
250
            def process_row(row, i):
251
                row_id = str(i)
252
                if id_node != None: xml_dom.set_value(id_node, row_id)
253
                for in_, out in mappings:
254
                    log_debug('Getting '+str(in_))
255
                    value = metadata_value(in_)
256
                    if value == None: value = cleanup(get_value(in_, row))
257
                    log_debug('Putting '+repr(value)+' to '+str(out))
258
                    if out_is_db: # out is list of XML nodes
259
                        for node in out: xml_dom.set_value(node, value)
260
                    elif value != None: # out is XPath
261
                        xpath.put_obj(root, out, row_id, has_types, value)
262
            return process_rows(process_row, rows, **kw_args)
263
        
264
        def map_table(col_names, rows, **kw_args):
265
            col_names_ct = len(col_names)
266
            col_idxs = util.list_flip(col_names)
267
            
268
            # Resolve prefixes
269
            mappings_orig = mappings[:] # save a copy
270
            mappings[:] = [] # empty existing elements
271
            for in_, out in mappings_orig:
272
                if metadata_value(in_) == None:
273
                    try: cols = get_with_prefix(col_idxs, prefixes, in_)
274
                    except KeyError: pass
275
                    else: mappings[len(mappings):] = [[db_xml.ColRef(*col), out]
276
                        for col in cols] # can't use += because that uses =
277
            
278
            def get_value(in_, row): return row.list[in_.idx]
279
            def wrap_row(row):
280
                return util.ListDict(util.list_as_length(row, col_names_ct),
281
                    col_names, col_idxs) # handle CSV rows of different lengths
282
            
283
            return map_rows(get_value, util.WrapIter(wrap_row, rows), **kw_args)
284
        
285
        stdin = streams.LineCountStream(sys.stdin)
286
        def on_error(e):
287
            exc.add_msg(e, term.emph('input line #:')+' '+str(stdin.line_num))
288
            ex_tracker.track(e)
289
        
290
        if in_is_db:
291
            in_db = connect_db(in_db_config)
292
            
293
            # Get table and schema name
294
            schema = in_schema # modified, so can't have same name as outer var
295
            table = in_table # modified, so can't have same name as outer var
296
            if table == None:
297
                assert in_is_xpaths
298
                schema, sep, table = in_root.partition('.')
299
                if sep == '': # only the table name was specified
300
                    table = schema
301
                    schema = None
302
            table = sql_gen.Table(table, schema)
303
            
304
            # Fetch rows
305
            if by_col: limit = 0 # only fetch column names
306
            else: limit = n
307
            cur = sql.select(in_db, table, limit=limit, start=start,
308
                cacheable=False)
309
            col_names = list(sql.col_names(cur))
310
            
311
            if by_col:
312
                map_table(col_names, []) # just create the template
313
                xml_func.strip(root, ['_simplifyPath'])
314
                if debug: log_debug('Putting stripped:\n'+str(root))
315
                    # only calc if debug
316
                in_row_ct_ref = [0]
317
                db_xml.put_table(in_db, root.firstChild, table, commit,
318
                    in_row_ct_ref, row_ins_ct_ref, n, start)
319
                row_ct = in_row_ct_ref[0]
320
            else:
321
                # Use normal by-row method
322
                row_ct = map_table(col_names, sql.rows(cur), rows_start=start)
323
                    # rows_start: pre-start rows have been skipped
324
            
325
            in_db.db.close()
326
        elif in_is_xml:
327
            def get_rows(doc2rows):
328
                return iters.flatten(itertools.imap(doc2rows,
329
                    xml_parse.docs_iter(stdin, on_error)))
330
            
331
            if map_path == None:
332
                def doc2rows(in_xml_root):
333
                    iter_ = xml_dom.NodeElemIter(in_xml_root)
334
                    util.skip(iter_, xml_dom.is_text) # skip metadata
335
                    return iter_
336
                
337
                row_ct = process_rows(lambda row, i: root.appendChild(row),
338
                    get_rows(doc2rows))
339
            else:
340
                def doc2rows(in_xml_root):
341
                    rows = xpath.get(in_xml_root, in_root, limit=end)
342
                    if rows == []: raise SystemExit('Map error: Root "'
343
                        +in_root+'" not found in input')
344
                    return rows
345
                
346
                def get_value(in_, row):
347
                    in_ = './{'+(','.join(strings.with_prefixes(
348
                        ['']+prefixes, in_)))+'}' # also with no prefix
349
                    nodes = xpath.get(row, in_, allow_rooted=False)
350
                    if nodes != []: return xml_dom.value(nodes[0])
351
                    else: return None
352
                
353
                row_ct = map_rows(get_value, get_rows(doc2rows))
354
        else: # input is CSV
355
            map_ = dict(mappings)
356
            reader, col_names = csvs.reader_and_header(sys.stdin)
357
            row_ct = map_table(col_names, reader)
358
        
359
        return row_ct
360
    
361
    def process_inputs(root, row_ready):
362
        row_ct = 0
363
        for map_path in map_paths:
364
            row_ct += process_input(root, row_ready, map_path)
365
        return row_ct
366
    
367
    pool.share_vars(locals())
368
    if out_is_db:
369
        out_db = connect_db(out_db_config)
370
        try:
371
            if redo: sql.empty_db(out_db)
372
            pool.share_vars(locals())
373
            
374
            def row_ready(row_num, input_row):
375
                row_str_ = [None]
376
                def row_str():
377
                    if row_str_[0] == None:
378
                        # Row # is interally 0-based, but 1-based to the user
379
                        row_str_[0] = (term.emph('row #:')+' '+str(row_num+1)
380
                            +'\n'+term.emph('input row:')+'\n'+str(input_row))
381
                        if verbose_errors: row_str_[0] += ('\n'
382
                            +term.emph('output row:')+'\n'+str(root))
383
                    return row_str_[0]
384
                
385
                if debug: log_debug(row_str()) # only calc if debug
386
                
387
                def on_error(e):
388
                    exc.add_msg(e, row_str())
389
                    ex_tracker.track(e, row_num, detail=verbose_errors)
390
                pool.share_vars(locals())
391
                
392
                row_root = root.cloneNode(True) # deep copy so don't modify root
393
                xml_func.process(row_root, on_error, out_db)
394
                if not xml_dom.is_empty(row_root):
395
                    assert xml_dom.has_one_child(row_root)
396
                    try:
397
                        sql.with_savepoint(out_db,
398
                            lambda: db_xml.put(out_db, row_root.firstChild,
399
                                row_ins_ct_ref, on_error))
400
                        if commit: out_db.db.commit()
401
                    except sql.DatabaseErrors, e: on_error(e)
402
            
403
            row_ct = process_inputs(root, row_ready)
404
            sys.stdout.write('Inserted '+str(row_ins_ct_ref[0])+
405
                ' new rows into database\n')
406
            
407
            # Consume asynchronous tasks
408
            pool.main_loop()
409
        finally:
410
            if out_db.connected():
411
                out_db.db.rollback()
412
                out_db.db.close()
413
    else:
414
        def on_error(e): ex_tracker.track(e)
415
        def row_ready(row_num, input_row): pass
416
        row_ct = process_inputs(root, row_ready)
417
        xml_func.process(root, on_error)
418
        if out_is_xml_ref[0]:
419
            doc.writexml(sys.stdout, **xml_dom.prettyxml_config)
420
        else: # output is CSV
421
            raise NotImplementedError('CSV output not supported yet')
422
    
423
    # Consume any asynchronous tasks not already consumed above
424
    pool.main_loop()
425
    
426
    profiler.stop(row_ct)
427
    ex_tracker.add_iters(row_ct)
428
    log('Processed '+str(row_ct)+' input rows')
429
    log(profiler.msg())
430
    log(ex_tracker.msg())
431
    ex_tracker.exit()
432

    
433
def main():
434
    try: main_()
435
    except Parser.SyntaxError, e: raise SystemExit(str(e))
436

    
437
if __name__ == '__main__':
438
    profile_to = opts.get_env_var('profile_to', None)
439
    if profile_to != None:
440
        import cProfile
441
        sys.stderr.write('Profiling to '+profile_to+'\n')
442
        cProfile.run(main.func_code, profile_to)
443
    else: main()
(25-25/48)