Project

General

Profile

1
# Database access
2

    
3
import copy
4
import operator
5
import re
6
import warnings
7

    
8
import exc
9
import dicts
10
import iters
11
import lists
12
from Proxy import Proxy
13
import rand
14
import strings
15
import util
16

    
17
##### Exceptions
18

    
19
def get_cur_query(cur, input_query=None, input_params=None):
20
    raw_query = None
21
    if hasattr(cur, 'query'): raw_query = cur.query
22
    elif hasattr(cur, '_last_executed'): raw_query = cur._last_executed
23
    
24
    if raw_query != None: return raw_query
25
    else: return repr(input_query)+' % '+repr(input_params)
26

    
27
def _add_cursor_info(e, *args, **kw_args):
28
    '''For params, see get_cur_query()'''
29
    exc.add_msg(e, 'query: '+str(get_cur_query(*args, **kw_args)))
30

    
31
class DbException(exc.ExceptionWithCause):
32
    def __init__(self, msg, cause=None, cur=None):
33
        exc.ExceptionWithCause.__init__(self, msg, cause, cause_newline=True)
34
        if cur != None: _add_cursor_info(self, cur)
35

    
36
class ExceptionWithName(DbException):
37
    def __init__(self, name, cause=None):
38
        DbException.__init__(self, 'for name: '+str(name), cause)
39
        self.name = name
40

    
41
class ExceptionWithColumns(DbException):
42
    def __init__(self, cols, cause=None):
43
        DbException.__init__(self, 'for columns: '+(', '.join(cols)), cause)
44
        self.cols = cols
45

    
46
class NameException(DbException): pass
47

    
48
class DuplicateKeyException(ExceptionWithColumns): pass
49

    
50
class NullValueException(ExceptionWithColumns): pass
51

    
52
class DuplicateTableException(ExceptionWithName): pass
53

    
54
class DuplicateFunctionException(ExceptionWithName): pass
55

    
56
class EmptyRowException(DbException): pass
57

    
58
##### Warnings
59

    
60
class DbWarning(UserWarning): pass
61

    
62
##### Result retrieval
63

    
64
def col_names(cur): return (col[0] for col in cur.description)
65

    
66
def rows(cur): return iter(lambda: cur.fetchone(), None)
67

    
68
def consume_rows(cur):
69
    '''Used to fetch all rows so result will be cached'''
70
    iters.consume_iter(rows(cur))
71

    
72
def next_row(cur): return rows(cur).next()
73

    
74
def row(cur):
75
    row_ = next_row(cur)
76
    consume_rows(cur)
77
    return row_
78

    
79
def next_value(cur): return next_row(cur)[0]
80

    
81
def value(cur): return row(cur)[0]
82

    
83
def values(cur): return iters.func_iter(lambda: next_value(cur))
84

    
85
def value_or_none(cur):
86
    try: return value(cur)
87
    except StopIteration: return None
88

    
89
##### Input validation
90

    
91
def clean_name(name): return re.sub(r'\W', r'', name).lower()
92

    
93
def check_name(name):
94
    if re.search(r'\W', name) != None: raise NameException('Name "'+name
95
        +'" may contain only alphanumeric characters and _')
96

    
97
def esc_name_by_module(module, name, ignore_case=False):
98
    if module == 'psycopg2':
99
        if ignore_case:
100
            # Don't enclose in quotes because this disables case-insensitivity
101
            check_name(name)
102
            return name
103
        else: quote = '"'
104
    elif module == 'MySQLdb': quote = '`'
105
    else: raise NotImplementedError("Can't escape name for "+module+' database')
106
    return quote + name.replace(quote, '') + quote
107

    
108
def esc_name_by_engine(engine, name, **kw_args):
109
    return esc_name_by_module(db_engines[engine][0], name, **kw_args)
110

    
111
def esc_name(db, name, **kw_args):
112
    return esc_name_by_module(util.root_module(db.db), name, **kw_args)
113

    
114
def qual_name(db, schema, table):
115
    def esc_name_(name): return esc_name(db, name)
116
    table = esc_name_(table)
117
    if schema != None: return esc_name_(schema)+'.'+table
118
    else: return table
119

    
120
##### Database connections
121

    
122
db_config_names = ['engine', 'host', 'user', 'password', 'database', 'schemas']
123

    
124
db_engines = {
125
    'MySQL': ('MySQLdb', {'password': 'passwd', 'database': 'db'}),
126
    'PostgreSQL': ('psycopg2', {}),
127
}
128

    
129
DatabaseErrors_set = set([DbException])
130
DatabaseErrors = tuple(DatabaseErrors_set)
131

    
132
def _add_module(module):
133
    DatabaseErrors_set.add(module.DatabaseError)
134
    global DatabaseErrors
135
    DatabaseErrors = tuple(DatabaseErrors_set)
136

    
137
def db_config_str(db_config):
138
    return db_config['engine']+' database '+db_config['database']
139

    
140
def _query_lookup(query, params): return (query, dicts.make_hashable(params))
141

    
142
log_debug_none = lambda msg: None
143

    
144
class DbConn:
145
    def __init__(self, db_config, serializable=True, autocommit=False,
146
        caching=True, log_debug=log_debug_none):
147
        self.db_config = db_config
148
        self.serializable = serializable
149
        self.autocommit = autocommit
150
        self.caching = caching
151
        self.log_debug = log_debug
152
        self.debug = log_debug != log_debug_none
153
        
154
        self.__db = None
155
        self.query_results = {}
156
        self._savepoint = 0
157
    
158
    def __getattr__(self, name):
159
        if name == '__dict__': raise Exception('getting __dict__')
160
        if name == 'db': return self._db()
161
        else: raise AttributeError()
162
    
163
    def __getstate__(self):
164
        state = copy.copy(self.__dict__) # shallow copy
165
        state['log_debug'] = None # don't pickle the debug callback
166
        state['_DbConn__db'] = None # don't pickle the connection
167
        return state
168
    
169
    def connected(self): return self.__db != None
170
    
171
    def _db(self):
172
        if self.__db == None:
173
            # Process db_config
174
            db_config = self.db_config.copy() # don't modify input!
175
            schemas = db_config.pop('schemas', None)
176
            module_name, mappings = db_engines[db_config.pop('engine')]
177
            module = __import__(module_name)
178
            _add_module(module)
179
            for orig, new in mappings.iteritems():
180
                try: util.rename_key(db_config, orig, new)
181
                except KeyError: pass
182
            
183
            # Connect
184
            self.__db = module.connect(**db_config)
185
            
186
            # Configure connection
187
            if self.serializable and not self.autocommit:
188
                self.db.set_session(isolation_level='SERIALIZABLE')
189
            if schemas != None:
190
                schemas_ = ''.join((esc_name(self, s)+', '
191
                    for s in schemas.split(',')))
192
                run_raw_query(self, "SELECT set_config('search_path', \
193
%s || current_setting('search_path'), false)", [schemas_])
194
        
195
        return self.__db
196
    
197
    class DbCursor(Proxy):
198
        def __init__(self, outer):
199
            Proxy.__init__(self, outer.db.cursor())
200
            self.outer = outer
201
            self.query_results = outer.query_results
202
            self.query_lookup = None
203
            self.result = []
204
        
205
        def execute(self, query, params=None):
206
            self._is_insert = query.upper().find('INSERT') >= 0
207
            self.query_lookup = _query_lookup(query, params)
208
            try:
209
                try:
210
                    return_value = self.inner.execute(query, params)
211
                    self.outer.do_autocommit()
212
                finally: self.query = get_cur_query(self.inner)
213
            except Exception, e:
214
                _add_cursor_info(e, self, query, params)
215
                self.result = e # cache the exception as the result
216
                self._cache_result()
217
                raise
218
            # Fetch all rows so result will be cached
219
            if self.rowcount == 0 and not self._is_insert: consume_rows(self)
220
            return return_value
221
        
222
        def fetchone(self):
223
            row = self.inner.fetchone()
224
            if row != None: self.result.append(row)
225
            # otherwise, fetched all rows
226
            else: self._cache_result()
227
            return row
228
        
229
        def _cache_result(self):
230
            # For inserts, only cache exceptions since inserts are not
231
            # idempotent, but an invalid insert will always be invalid
232
            if self.query_results != None and (not self._is_insert
233
                or isinstance(self.result, Exception)):
234
                
235
                assert self.query_lookup != None
236
                self.query_results[self.query_lookup] = self.CacheCursor(
237
                    util.dict_subset(dicts.AttrsDictView(self),
238
                    ['query', 'result', 'rowcount', 'description']))
239
        
240
        class CacheCursor:
241
            def __init__(self, cached_result): self.__dict__ = cached_result
242
            
243
            def execute(self, *args, **kw_args):
244
                if isinstance(self.result, Exception): raise self.result
245
                # otherwise, result is a rows list
246
                self.iter = iter(self.result)
247
            
248
            def fetchone(self):
249
                try: return self.iter.next()
250
                except StopIteration: return None
251
    
252
    def run_query(self, query, params=None, cacheable=False):
253
        '''Translates known DB errors to typed exceptions:
254
        See self.DbCursor.execute().'''
255
        assert query != None
256
        
257
        if not self.caching: cacheable = False
258
        used_cache = False
259
        try:
260
            # Get cursor
261
            if cacheable:
262
                query_lookup = _query_lookup(query, params)
263
                try:
264
                    cur = self.query_results[query_lookup]
265
                    used_cache = True
266
                except KeyError: cur = self.DbCursor(self)
267
            else: cur = self.db.cursor()
268
            
269
            # Run query
270
            cur.execute(query, params)
271
        finally:
272
            if self.debug: # only compute msg if needed
273
                if used_cache: cache_status = 'Cache hit'
274
                elif cacheable: cache_status = 'Cache miss'
275
                else: cache_status = 'Non-cacheable'
276
                self.log_debug(cache_status+': '
277
                    +strings.one_line(str(get_cur_query(cur, query, params))))
278
        
279
        return cur
280
    
281
    def is_cached(self, query, params=None):
282
        return _query_lookup(query, params) in self.query_results
283
    
284
    def with_savepoint(self, func):
285
        savepoint = 'level_'+str(self._savepoint)
286
        self.run_query('SAVEPOINT '+savepoint)
287
        self._savepoint += 1
288
        try: 
289
            try: return_val = func()
290
            finally:
291
                self._savepoint -= 1
292
                assert self._savepoint >= 0
293
        except:
294
            self.run_query('ROLLBACK TO SAVEPOINT '+savepoint)
295
            raise
296
        else:
297
            self.run_query('RELEASE SAVEPOINT '+savepoint)
298
            self.do_autocommit()
299
            return return_val
300
    
301
    def do_autocommit(self):
302
        '''Autocommits if outside savepoint'''
303
        assert self._savepoint >= 0
304
        if self.autocommit and self._savepoint == 0:
305
            self.log_debug('Autocommiting')
306
            self.db.commit()
307

    
308
connect = DbConn
309

    
310
##### Querying
311

    
312
def run_raw_query(db, *args, **kw_args):
313
    '''For params, see DbConn.run_query()'''
314
    return db.run_query(*args, **kw_args)
315

    
316
def mogrify(db, query, params):
317
    module = util.root_module(db.db)
318
    if module == 'psycopg2': return db.db.cursor().mogrify(query, params)
319
    else: raise NotImplementedError("Can't mogrify query for "+module+
320
        ' database')
321

    
322
##### Recoverable querying
323

    
324
def with_savepoint(db, func): return db.with_savepoint(func)
325

    
326
def run_query(db, query, params=None, recover=None, cacheable=False):
327
    if recover == None: recover = False
328
    
329
    try:
330
        def run(): return run_raw_query(db, query, params, cacheable)
331
        if recover and not db.is_cached(query, params):
332
            return with_savepoint(db, run)
333
        else: return run() # don't need savepoint if cached
334
    except Exception, e:
335
        if not recover: raise # need savepoint to run index_cols()
336
        msg = str(e)
337
        match = re.search(r'duplicate key value violates unique constraint '
338
            r'"((_?[^\W_]+)_[^"]+)"', msg)
339
        if match:
340
            constraint, table = match.groups()
341
            try: cols = index_cols(db, table, constraint)
342
            except NotImplementedError: raise e
343
            else: raise DuplicateKeyException(cols, e)
344
        match = re.search(r'null value in column "(\w+)" violates not-null '
345
            'constraint', msg)
346
        if match: raise NullValueException([match.group(1)], e)
347
        match = re.search(r'relation "(\w+)" already exists', msg)
348
        if match: raise DuplicateTableException(match.group(1), e)
349
        match = re.search(r'function "(\w+)" already exists', msg)
350
        if match: raise DuplicateFunctionException(match.group(1), e)
351
        raise # no specific exception raised
352

    
353
##### Basic queries
354

    
355
def next_version(name):
356
    '''Prepends the version # so it won't be removed if the name is truncated'''
357
    version = 1 # first existing name was version 0
358
    match = re.match(r'^v(\d+)_(.*)$', name)
359
    if match:
360
        version = int(match.group(1))+1
361
        name = match.group(2)
362
    return 'v'+str(version)+'_'+name
363

    
364
def run_query_into(db, query, params, into_ref=None, *args, **kw_args):
365
    '''Outputs a query to a temp table.
366
    For params, see run_query().
367
    '''
368
    if into_ref == None: return run_query(db, query, params, *args, **kw_args)
369
    else: # place rows in temp table
370
        check_name(into_ref[0])
371
        kw_args['recover'] = True
372
        while True:
373
            try:
374
                create_query = 'CREATE'
375
                if not db.debug: create_query += ' TEMP'
376
                create_query += ' TABLE '+into_ref[0]+' AS '+query
377
                
378
                return run_query(db, create_query, params, *args, **kw_args)
379
                    # CREATE TABLE AS sets rowcount to # rows in query
380
            except DuplicateTableException, e:
381
                into_ref[0] = next_version(into_ref[0])
382
                # try again with next version of name
383

    
384
order_by_pkey = object() # tells mk_select() to order by the pkey
385

    
386
join_using = object() # tells mk_select() to join the column with USING
387

    
388
filter_out = object() # tells mk_select() to filter out rows that match the join
389

    
390
distinct_on_all = object() # tells mk_select() to SELECT DISTINCT ON all columns
391

    
392
def mk_select(db, tables, fields=None, conds=None, distinct_on=None, limit=None,
393
    start=None, order_by=order_by_pkey, table_is_esc=False):
394
    '''
395
    @param tables The single table to select from, or a list of tables to join
396
        together: [table0, (table1, joins), ...]
397
        
398
        joins has the format: dict(right_col=left_col, ...)
399
        * if left_col is join_using, left_col is set to right_col
400
        * if left_col is filter_out, the tables are LEFT JOINed together and the
401
          query is filtered by `right_col IS NULL` (indicating no match)
402
    @param fields Use None to select all fields in the table
403
    @param distinct_on The columns to SELECT DISTINCT ON, or distinct_on_all to
404
        use all columns
405
    @param table_is_esc Whether the table name has already been escaped
406
    @return tuple(query, params)
407
    '''
408
    def esc_name_(name): return esc_name(db, name)
409
    
410
    if not lists.is_seq(tables): tables = [tables]
411
    tables = list(tables) # don't modify input! (list() copies input)
412
    table0 = tables.pop(0) # first table is separate
413
    
414
    if conds == None: conds = {}
415
    assert limit == None or type(limit) == int
416
    assert start == None or type(start) == int
417
    if order_by == order_by_pkey:
418
        order_by = pkey(db, table0, recover=True, table_is_esc=table_is_esc)
419
    if not table_is_esc: table0 = esc_name_(table0)
420
    
421
    params = []
422
    
423
    def parse_col(field, default_table=None):
424
        '''Parses fields'''
425
        if field == None: field = (field,) # for None values, tuple is optional
426
        is_tuple = isinstance(field, tuple)
427
        if is_tuple and len(field) == 1: # field is literal value
428
            value, = field
429
            sql_ = '%s'
430
            params.append(value)
431
        elif is_tuple and len(field) == 2: # field is col with table
432
            table, col = field
433
            if not table_is_esc: table = esc_name_(table)
434
            sql_ = table+'.'+esc_name_(col)
435
        else:
436
            sql_ = esc_name_(field) # field is col name
437
            if default_table != None: sql_ = default_table+'.'+sql_
438
        return sql_
439
    def cond(entry):
440
        '''Parses conditions'''
441
        col, value = entry
442
        cond_ = parse_col(col)+' '
443
        if value == None: cond_ += 'IS'
444
        else: cond_ += '='
445
        cond_ += ' %s'
446
        return cond_
447
    
448
    query = 'SELECT'
449
    
450
    # DISTINCT ON columns
451
    if distinct_on != None:
452
        query += ' DISTINCT'
453
        if distinct_on != distinct_on_all:
454
            query += ' ON ('+(', '.join(map(parse_col, distinct_on)))+')'
455
    
456
    # Columns
457
    query += ' '
458
    if fields == None: query += '*'
459
    else: query += ', '.join(map(parse_col, fields))
460
    
461
    # Main table
462
    query += ' FROM '+table0
463
    
464
    # Add joins
465
    left_table = table0
466
    for table, joins in tables:
467
        if not table_is_esc: table = esc_name_(table)
468
        
469
        left_join_ref = [False]
470
        
471
        def join(entry):
472
            '''Parses non-USING joins'''
473
            right_col, left_col = entry
474
            
475
            # Parse special values
476
            if left_col == None: left_col = (left_col,)
477
                # for None values, tuple is optional
478
            elif left_col == join_using: left_col = right_col
479
            elif left_col == filter_out:
480
                left_col = right_col
481
                left_join_ref[0] = True
482
                conds[(table, right_col)] = None # filter query by no match
483
            
484
            # Create SQL
485
            right_col = table+'.'+esc_name_(right_col)
486
            sql_ = right_col+' '
487
            if isinstance(left_col, tuple) and len(left_col) == 1:
488
                # col is literal value
489
                value, = left_col
490
                if value == None: sql_ += 'IS'
491
                else: sql_ += '='
492
                sql_ += ' %s'
493
                params.append(value)
494
            else: # col is name
495
                left_col = parse_col(left_col, left_table)
496
                sql_ += ('= '+left_col+' OR ('+right_col+' IS NULL AND '
497
                    +left_col+' IS NULL)')
498
            
499
            return sql_
500
        
501
        # Create join condition and determine join type
502
        if reduce(operator.and_, (v == join_using for v in joins.itervalues())):
503
            # all cols w/ USING, so can use simpler USING syntax
504
            join_cond = 'USING ('+(', '.join(joins.iterkeys()))+')'
505
        else: join_cond = 'ON '+(' AND '.join(map(join, joins.iteritems())))
506
        
507
        # Create join
508
        if left_join_ref[0]: query += ' LEFT'
509
        query += ' JOIN '+table+' '+join_cond
510
        
511
        left_table = table
512
    
513
    missing = True
514
    if conds != {}:
515
        query += ' WHERE '+(' AND '.join(map(cond, conds.iteritems())))
516
        params += conds.values()
517
        missing = False
518
    if order_by != None: query += ' ORDER BY '+parse_col(order_by, table0)
519
    if limit != None: query += ' LIMIT '+str(limit); missing = False
520
    if start != None:
521
        if start != 0: query += ' OFFSET '+str(start)
522
        missing = False
523
    if missing: warnings.warn(DbWarning(
524
        'SELECT statement missing a WHERE, LIMIT, or OFFSET clause: '+query))
525
    
526
    return (query, params)
527

    
528
def select(db, *args, **kw_args):
529
    '''For params, see mk_select() and run_query()'''
530
    recover = kw_args.pop('recover', None)
531
    cacheable = kw_args.pop('cacheable', True)
532
    
533
    query, params = mk_select(db, *args, **kw_args)
534
    return run_query(db, query, params, recover, cacheable)
535

    
536
def mk_insert_select(db, table, cols=None, select_query=None, params=None,
537
    returning=None, embeddable=False, table_is_esc=False):
538
    '''
539
    @param returning str|None An inserted column (such as pkey) to return
540
    @param embeddable Whether the query should be embeddable as a nested SELECT.
541
        Warning: If you set this and cacheable=True when the query is run, the
542
        query will be fully cached, not just if it raises an exception.
543
    @param table_is_esc Whether the table name has already been escaped
544
    '''
545
    if select_query == None: select_query = 'DEFAULT VALUES'
546
    if cols == []: cols = None # no cols (all defaults) = unknown col names
547
    if not table_is_esc: check_name(table)
548
    
549
    # Build query
550
    query = 'INSERT INTO '+table
551
    if cols != None:
552
        map(check_name, cols)
553
        query += ' ('+', '.join(cols)+')'
554
    query += ' '+select_query
555
    
556
    if returning != None:
557
        check_name(returning)
558
        query += ' RETURNING '+returning
559
    
560
    if embeddable:
561
        # Create function
562
        function_name = '_'.join(map(clean_name, ['insert', table] + cols))
563
        return_type = 'SETOF '+table+'.'+returning+'%TYPE'
564
        while True:
565
            try:
566
                function = function_name
567
                if not db.debug: function = 'pg_temp.'+function
568
                
569
                function_query = '''\
570
CREATE FUNCTION '''+function+'''() RETURNS '''+return_type+'''
571
    LANGUAGE sql
572
    AS $$'''+mogrify(db, query, params)+''';$$;
573
'''
574
                run_query(db, function_query, recover=True, cacheable=True)
575
                break # this version was successful
576
            except DuplicateFunctionException, e:
577
                function_name = next_version(function_name)
578
                # try again with next version of name
579
        
580
        # Return query that uses function
581
        return mk_select(db, function+'() AS f ('+returning+')', start=0,
582
            order_by=None, table_is_esc=True)# AS clause requires function alias
583
    
584
    return (query, params)
585

    
586
def insert_select(db, *args, **kw_args):
587
    '''For params, see mk_insert_select() and run_query_into()
588
    @param into_ref List with name of temp table to place RETURNING values in
589
    '''
590
    into_ref = kw_args.pop('into_ref', None)
591
    if into_ref != None: kw_args['embeddable'] = True
592
    recover = kw_args.pop('recover', None)
593
    cacheable = kw_args.pop('cacheable', True)
594
    
595
    query, params = mk_insert_select(db, *args, **kw_args)
596
    return run_query_into(db, query, params, into_ref, recover=recover,
597
        cacheable=cacheable)
598

    
599
default = object() # tells insert() to use the default value for a column
600

    
601
def insert(db, table, row, *args, **kw_args):
602
    '''For params, see insert_select()'''
603
    if lists.is_seq(row): cols = None
604
    else:
605
        cols = row.keys()
606
        row = row.values()
607
    row = list(row) # ensure that "!= []" works
608
    
609
    # Check for special values
610
    labels = []
611
    values = []
612
    for value in row:
613
        if value == default: labels.append('DEFAULT')
614
        else:
615
            labels.append('%s')
616
            values.append(value)
617
    
618
    # Build query
619
    if values != []: query = ' VALUES ('+(', '.join(labels))+')'
620
    else: query = None
621
    
622
    return insert_select(db, table, cols, query, values, *args, **kw_args)
623

    
624
def last_insert_id(db):
625
    module = util.root_module(db.db)
626
    if module == 'psycopg2': return value(run_query(db, 'SELECT lastval()'))
627
    elif module == 'MySQLdb': return db.insert_id()
628
    else: return None
629

    
630
def truncate(db, table, schema='public'):
631
    return run_query(db, 'TRUNCATE '+qual_name(db, schema, table)+' CASCADE')
632

    
633
##### Database structure queries
634

    
635
def pkey(db, table, recover=None, table_is_esc=False):
636
    '''Assumed to be first column in table'''
637
    return col_names(select(db, table, limit=0, order_by=None, recover=recover,
638
        table_is_esc=table_is_esc)).next()
639

    
640
def index_cols(db, table, index):
641
    '''Can also use this for UNIQUE constraints, because a UNIQUE index is
642
    automatically created. When you don't know whether something is a UNIQUE
643
    constraint or a UNIQUE index, use this function.'''
644
    check_name(table)
645
    check_name(index)
646
    module = util.root_module(db.db)
647
    if module == 'psycopg2':
648
        return list(values(run_query(db, '''\
649
SELECT attname
650
FROM
651
(
652
        SELECT attnum, attname
653
        FROM pg_index
654
        JOIN pg_class index ON index.oid = indexrelid
655
        JOIN pg_class table_ ON table_.oid = indrelid
656
        JOIN pg_attribute ON attrelid = indrelid AND attnum = ANY (indkey)
657
        WHERE
658
            table_.relname = %(table)s
659
            AND index.relname = %(index)s
660
    UNION
661
        SELECT attnum, attname
662
        FROM
663
        (
664
            SELECT
665
                indrelid
666
                , (regexp_matches(indexprs, E':varattno (\\\\d+)', 'g'))[1]::int
667
                    AS indkey
668
            FROM pg_index
669
            JOIN pg_class index ON index.oid = indexrelid
670
            JOIN pg_class table_ ON table_.oid = indrelid
671
            WHERE
672
                table_.relname = %(table)s
673
                AND index.relname = %(index)s
674
        ) s
675
        JOIN pg_attribute ON attrelid = indrelid AND attnum = indkey
676
) s
677
ORDER BY attnum
678
''',
679
            {'table': table, 'index': index}, cacheable=True)))
680
    else: raise NotImplementedError("Can't list index columns for "+module+
681
        ' database')
682

    
683
def constraint_cols(db, table, constraint):
684
    check_name(table)
685
    check_name(constraint)
686
    module = util.root_module(db.db)
687
    if module == 'psycopg2':
688
        return list(values(run_query(db, '''\
689
SELECT attname
690
FROM pg_constraint
691
JOIN pg_class ON pg_class.oid = conrelid
692
JOIN pg_attribute ON attrelid = conrelid AND attnum = ANY (conkey)
693
WHERE
694
    relname = %(table)s
695
    AND conname = %(constraint)s
696
ORDER BY attnum
697
''',
698
            {'table': table, 'constraint': constraint})))
699
    else: raise NotImplementedError("Can't list constraint columns for "+module+
700
        ' database')
701

    
702
row_num_col = '_row_num'
703

    
704
def add_row_num(db, table):
705
    '''Adds a row number column to a table. Its name is in row_num_col. It will
706
    be the primary key.'''
707
    check_name(table)
708
    run_query(db, 'ALTER TABLE '+table+' ADD COLUMN '+row_num_col
709
        +' serial NOT NULL PRIMARY KEY')
710

    
711
def tables(db, schema='public', table_like='%'):
712
    module = util.root_module(db.db)
713
    params = {'schema': schema, 'table_like': table_like}
714
    if module == 'psycopg2':
715
        return values(run_query(db, '''\
716
SELECT tablename
717
FROM pg_tables
718
WHERE
719
    schemaname = %(schema)s
720
    AND tablename LIKE %(table_like)s
721
ORDER BY tablename
722
''',
723
            params, cacheable=True))
724
    elif module == 'MySQLdb':
725
        return values(run_query(db, 'SHOW TABLES LIKE %(table_like)s', params,
726
            cacheable=True))
727
    else: raise NotImplementedError("Can't list tables for "+module+' database')
728

    
729
##### Database management
730

    
731
def empty_db(db, schema='public', **kw_args):
732
    '''For kw_args, see tables()'''
733
    for table in tables(db, schema, **kw_args): truncate(db, table, schema)
734

    
735
##### Heuristic queries
736

    
737
def put(db, table, row, pkey_=None, row_ct_ref=None):
738
    '''Recovers from errors.
739
    Only works under PostgreSQL (uses INSERT RETURNING).
740
    '''
741
    if pkey_ == None: pkey_ = pkey(db, table, recover=True)
742
    
743
    try:
744
        cur = insert(db, table, row, pkey_, recover=True)
745
        if row_ct_ref != None and cur.rowcount >= 0:
746
            row_ct_ref[0] += cur.rowcount
747
        return value(cur)
748
    except DuplicateKeyException, e:
749
        return value(select(db, table, [pkey_],
750
            util.dict_subset_right_join(row, e.cols), recover=True))
751

    
752
def get(db, table, row, pkey, row_ct_ref=None, create=False):
753
    '''Recovers from errors'''
754
    try: return value(select(db, table, [pkey], row, limit=1, recover=True))
755
    except StopIteration:
756
        if not create: raise
757
        return put(db, table, row, pkey, row_ct_ref) # insert new row
758

    
759
def put_table(db, out_table, in_tables, mapping, limit=None, start=0,
760
    row_ct_ref=None, table_is_esc=False):
761
    '''Recovers from errors.
762
    Only works under PostgreSQL (uses INSERT RETURNING).
763
    @param in_tables The main input table to select from, followed by a list of
764
        tables to join with it using the main input table's pkey
765
    @return (table, col) Where the pkeys (from INSERT RETURNING) are made
766
        available
767
    '''
768
    temp_suffix = clean_name(out_table)
769
        # suffix, not prefix, so main name won't be removed if name is truncated
770
    pkeys_ref = ['pkeys_'+temp_suffix]
771
    
772
    # Join together input tables
773
    in_tables = in_tables[:] # don't modify input!
774
    in_tables0 = in_tables.pop(0) # first table is separate
775
    in_pkey = pkey(db, in_tables0, recover=True, table_is_esc=table_is_esc)
776
    insert_joins = [in_tables0]+[(t, {in_pkey: join_using}) for t in in_tables]
777
    
778
    out_pkey = pkey(db, out_table, recover=True, table_is_esc=table_is_esc)
779
    pkeys_cols = [in_pkey, out_pkey]
780
    
781
    pkeys_table_exists_ref = [False]
782
    def run_query_into_pkeys(query, params):
783
        if pkeys_table_exists_ref[0]:
784
            insert_select(db, pkeys_ref[0], pkeys_cols, query, params)
785
        else:
786
            run_query_into(db, query, params, into_ref=pkeys_ref)
787
            pkeys_table_exists_ref[0] = True
788
    
789
    conds = {}
790
    distinct_on = None
791
    def mk_main_select(cols):
792
        return mk_select(db, insert_joins, cols, conds, distinct_on,
793
            order_by=None, limit=limit, start=start, table_is_esc=table_is_esc)
794
    
795
    # Do inserts and selects
796
    out_pkeys_ref = ['out_pkeys_'+temp_suffix]
797
    while True:
798
        try:
799
            cur = insert_select(db, out_table, mapping.keys(),
800
                *mk_main_select(mapping.values()), returning=out_pkey,
801
                into_ref=out_pkeys_ref, recover=True, table_is_esc=table_is_esc)
802
            if row_ct_ref != None and cur.rowcount >= 0:
803
                row_ct_ref[0] += cur.rowcount
804
                add_row_num(db, out_pkeys_ref[0]) # for joining with input pkeys
805
            
806
            # Get input pkeys corresponding to rows in insert
807
            in_pkeys_ref = ['in_pkeys_'+temp_suffix]
808
            run_query_into(db, *mk_main_select([in_pkey]),
809
                into_ref=in_pkeys_ref)
810
            add_row_num(db, in_pkeys_ref[0]) # for joining with output pkeys
811
            
812
            # Join together output and input pkeys
813
            run_query_into_pkeys(*mk_select(db, [in_pkeys_ref[0],
814
                (out_pkeys_ref[0], {row_num_col: join_using})], pkeys_cols,
815
                start=0))
816
            
817
            break # insert successful
818
        except DuplicateKeyException, e:
819
            join_cols = util.dict_subset_right_join(mapping, e.cols)
820
            select_joins = insert_joins + [(out_table, join_cols)]
821
            
822
            # Get pkeys of already existing rows
823
            run_query_into_pkeys(*mk_select(db, select_joins, pkeys_cols,
824
                order_by=None, start=0, table_is_esc=table_is_esc))
825
            
826
            # Save existing pkeys in temp table for joining on
827
            existing_pkeys_ref = ['existing_pkeys_'+temp_suffix]
828
            run_query_into(db, *mk_select(db, pkeys_ref[0], [in_pkey],
829
                order_by=None, start=0, table_is_esc=True),
830
                into_ref=existing_pkeys_ref)
831
                # need table_is_esc=True to make table name case-insensitive
832
            
833
            # rerun loop with additional constraints
834
            break # but until NullValueExceptions are handled, end loop here
835
    
836
    return (pkeys_ref[0], out_pkey)
837

    
838
##### Data cleanup
839

    
840
def cleanup_table(db, table, cols, table_is_esc=False):
841
    def esc_name_(name): return esc_name(db, name)
842
    
843
    if not table_is_esc: check_name(table)
844
    cols = map(esc_name_, cols)
845
    
846
    run_query(db, 'UPDATE '+table+' SET\n'+(',\n'.join(('\n'+col
847
        +' = nullif(nullif(trim(both from '+col+"), %(null0)s), %(null1)s)"
848
            for col in cols))),
849
        dict(null0='', null1=r'\N'))
(22-22/34)