Project

General

Profile

1
# Database access
2

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

    
8
import exc
9
import dicts
10
import iters
11
import lists
12
import profiling
13
from Proxy import Proxy
14
import rand
15
import regexp
16
import sql_gen
17
import strings
18
import util
19

    
20
##### Exceptions
21

    
22
def get_cur_query(cur, input_query=None):
23
    raw_query = None
24
    if hasattr(cur, 'query'): raw_query = cur.query
25
    elif hasattr(cur, '_last_executed'): raw_query = cur._last_executed
26
    
27
    if raw_query != None: return raw_query
28
    else: return '[input] '+strings.ustr(input_query)
29

    
30
def _add_cursor_info(e, *args, **kw_args):
31
    '''For params, see get_cur_query()'''
32
    exc.add_msg(e, 'query: '+strings.ustr(get_cur_query(*args, **kw_args)))
33

    
34
class DbException(exc.ExceptionWithCause):
35
    def __init__(self, msg, cause=None, cur=None):
36
        exc.ExceptionWithCause.__init__(self, msg, cause, cause_newline=True)
37
        if cur != None: _add_cursor_info(self, cur)
38

    
39
class ExceptionWithName(DbException):
40
    def __init__(self, name, cause=None):
41
        DbException.__init__(self, 'for name: '
42
            +strings.as_tt(strings.ustr(name)), cause)
43
        self.name = name
44

    
45
class ExceptionWithValue(DbException):
46
    def __init__(self, value, cause=None):
47
        DbException.__init__(self, 'for value: '
48
            +strings.as_tt(strings.urepr(value)), cause)
49
        self.value = value
50

    
51
class ExceptionWithNameType(DbException):
52
    def __init__(self, type_, name, cause=None):
53
        DbException.__init__(self, 'for type: '+strings.as_tt(strings.ustr(
54
            type_))+'; name: '+strings.as_tt(name), cause)
55
        self.type = type_
56
        self.name = name
57

    
58
class ConstraintException(DbException):
59
    def __init__(self, name, cond, cols, cause=None):
60
        msg = 'Violated '+strings.as_tt(name)+' constraint'
61
        if cond != None: msg += ' with condition '+cond
62
        if cols != []: msg += ' on columns: '+strings.as_tt(', '.join(cols))
63
        DbException.__init__(self, msg, cause)
64
        self.name = name
65
        self.cond = cond
66
        self.cols = cols
67

    
68
class MissingCastException(DbException):
69
    def __init__(self, type_, col=None, cause=None):
70
        msg = 'Missing cast to type '+strings.as_tt(type_)
71
        if col != None: msg += ' on column: '+strings.as_tt(col)
72
        DbException.__init__(self, msg, cause)
73
        self.type = type_
74
        self.col = col
75

    
76
class NameException(DbException): pass
77

    
78
class DuplicateKeyException(ConstraintException): pass
79

    
80
class NullValueException(ConstraintException): pass
81

    
82
class CheckException(ConstraintException): pass
83

    
84
class InvalidValueException(ExceptionWithValue): pass
85

    
86
class DuplicateException(ExceptionWithNameType): pass
87

    
88
class DoesNotExistException(ExceptionWithNameType): pass
89

    
90
class EmptyRowException(DbException): pass
91

    
92
##### Warnings
93

    
94
class DbWarning(UserWarning): pass
95

    
96
##### Result retrieval
97

    
98
def col_names(cur): return (col[0] for col in cur.description)
99

    
100
def rows(cur): return iter(lambda: cur.fetchone(), None)
101

    
102
def consume_rows(cur):
103
    '''Used to fetch all rows so result will be cached'''
104
    iters.consume_iter(rows(cur))
105

    
106
def next_row(cur): return rows(cur).next()
107

    
108
def row(cur):
109
    row_ = next_row(cur)
110
    consume_rows(cur)
111
    return row_
112

    
113
def next_value(cur): return next_row(cur)[0]
114

    
115
def value(cur): return row(cur)[0]
116

    
117
def values(cur): return iters.func_iter(lambda: next_value(cur))
118

    
119
def value_or_none(cur):
120
    try: return value(cur)
121
    except StopIteration: return None
122

    
123
##### Escaping
124

    
125
def esc_name_by_module(module, name):
126
    if module == 'psycopg2' or module == None: quote = '"'
127
    elif module == 'MySQLdb': quote = '`'
128
    else: raise NotImplementedError("Can't escape name for "+module+' database')
129
    return sql_gen.esc_name(name, quote)
130

    
131
def esc_name_by_engine(engine, name, **kw_args):
132
    return esc_name_by_module(db_engines[engine][0], name, **kw_args)
133

    
134
def esc_name(db, name, **kw_args):
135
    return esc_name_by_module(util.root_module(db.db), name, **kw_args)
136

    
137
def qual_name(db, schema, table):
138
    def esc_name_(name): return esc_name(db, name)
139
    table = esc_name_(table)
140
    if schema != None: return esc_name_(schema)+'.'+table
141
    else: return table
142

    
143
##### Database connections
144

    
145
db_config_names = ['engine', 'host', 'user', 'password', 'database', 'schemas']
146

    
147
db_engines = {
148
    'MySQL': ('MySQLdb', {'password': 'passwd', 'database': 'db'}),
149
    'PostgreSQL': ('psycopg2', {}),
150
}
151

    
152
DatabaseErrors_set = set([DbException])
153
DatabaseErrors = tuple(DatabaseErrors_set)
154

    
155
def _add_module(module):
156
    DatabaseErrors_set.add(module.DatabaseError)
157
    global DatabaseErrors
158
    DatabaseErrors = tuple(DatabaseErrors_set)
159

    
160
def db_config_str(db_config):
161
    return db_config['engine']+' database '+db_config['database']
162

    
163
log_debug_none = lambda msg, level=2: None
164

    
165
class DbConn:
166
    def __init__(self, db_config, autocommit=True, caching=True,
167
        log_debug=log_debug_none, debug_temp=False, src=None):
168
        '''
169
        @param debug_temp Whether temporary objects should instead be permanent.
170
            This assists in debugging the internal objects used by the program.
171
        @param src In autocommit mode, will be included in a comment in every
172
            query, to help identify the data source in pg_stat_activity.
173
        '''
174
        self.db_config = db_config
175
        self.autocommit = autocommit
176
        self.caching = caching
177
        self.log_debug = log_debug
178
        self.debug = log_debug != log_debug_none
179
        self.debug_temp = debug_temp
180
        self.src = src
181
        self.autoanalyze = False
182
        self.autoexplain = False
183
        self.profile_row_ct = None
184
        
185
        self._savepoint = 0
186
        self._reset()
187
    
188
    def __getattr__(self, name):
189
        if name == '__dict__': raise Exception('getting __dict__')
190
        if name == 'db': return self._db()
191
        else: raise AttributeError()
192
    
193
    def __getstate__(self):
194
        state = copy.copy(self.__dict__) # shallow copy
195
        state['log_debug'] = None # don't pickle the debug callback
196
        state['_DbConn__db'] = None # don't pickle the connection
197
        return state
198
    
199
    def clear_cache(self): self.query_results = {}
200
    
201
    def _reset(self):
202
        self.clear_cache()
203
        assert self._savepoint == 0
204
        self._notices_seen = set()
205
        self.__db = None
206
    
207
    def connected(self): return self.__db != None
208
    
209
    def close(self):
210
        if not self.connected(): return
211
        
212
        # Record that the automatic transaction is now closed
213
        self._savepoint -= 1
214
        
215
        self.db.close()
216
        self._reset()
217
    
218
    def reconnect(self):
219
        # Do not do this in test mode as it would roll back everything
220
        if self.autocommit: self.close()
221
        # Connection will be reopened automatically on first query
222
    
223
    def _db(self):
224
        if self.__db == None:
225
            # Process db_config
226
            db_config = self.db_config.copy() # don't modify input!
227
            schemas = db_config.pop('schemas', None)
228
            module_name, mappings = db_engines[db_config.pop('engine')]
229
            module = __import__(module_name)
230
            _add_module(module)
231
            for orig, new in mappings.iteritems():
232
                try: util.rename_key(db_config, orig, new)
233
                except KeyError: pass
234
            
235
            # Connect
236
            self.__db = module.connect(**db_config)
237
            
238
            # Record that a transaction is already open
239
            self._savepoint += 1
240
            
241
            # Configure connection
242
            if hasattr(self.db, 'set_isolation_level'):
243
                import psycopg2.extensions
244
                self.db.set_isolation_level(
245
                    psycopg2.extensions.ISOLATION_LEVEL_READ_COMMITTED)
246
            if schemas != None:
247
                search_path = [self.esc_name(s) for s in schemas.split(',')]
248
                search_path.append(value(run_query(self, 'SHOW search_path',
249
                    log_level=4)))
250
                run_query(self, 'SET search_path TO '+(','.join(search_path)),
251
                    log_level=3)
252
        
253
        return self.__db
254
    
255
    class DbCursor(Proxy):
256
        def __init__(self, outer):
257
            Proxy.__init__(self, outer.db.cursor())
258
            self.outer = outer
259
            self.query_results = outer.query_results
260
            self.query_lookup = None
261
            self.result = []
262
        
263
        def execute(self, query):
264
            self._is_insert = query.startswith('INSERT')
265
            self.query_lookup = query
266
            try:
267
                try: cur = self.inner.execute(query)
268
                finally: self.query = get_cur_query(self.inner, query)
269
            except Exception, e:
270
                self.result = e # cache the exception as the result
271
                self._cache_result()
272
                raise
273
            
274
            # Always cache certain queries
275
            query = sql_gen.lstrip(query)
276
            if query.startswith('CREATE') or query.startswith('ALTER'):
277
                # structural changes
278
                # Rest of query must be unique in the face of name collisions,
279
                # so don't cache ADD COLUMN unless it has distinguishing comment
280
                if query.find('ADD COLUMN') < 0 or query.endswith('*/'):
281
                    self._cache_result()
282
            elif self.rowcount == 0 and query.startswith('SELECT'): # empty
283
                consume_rows(self) # fetch all rows so result will be cached
284
            
285
            return cur
286
        
287
        def fetchone(self):
288
            row = self.inner.fetchone()
289
            if row != None: self.result.append(row)
290
            # otherwise, fetched all rows
291
            else: self._cache_result()
292
            return row
293
        
294
        def _cache_result(self):
295
            # For inserts that return a result set, don't cache result set since
296
            # inserts are not idempotent. Other non-SELECT queries don't have
297
            # their result set read, so only exceptions will be cached (an
298
            # invalid query will always be invalid).
299
            if self.query_results != None and (not self._is_insert
300
                or isinstance(self.result, Exception)):
301
                
302
                assert self.query_lookup != None
303
                self.query_results[self.query_lookup] = self.CacheCursor(
304
                    util.dict_subset(dicts.AttrsDictView(self),
305
                    ['query', 'result', 'rowcount', 'description']))
306
        
307
        class CacheCursor:
308
            def __init__(self, cached_result): self.__dict__ = cached_result
309
            
310
            def execute(self, *args, **kw_args):
311
                if isinstance(self.result, Exception): raise self.result
312
                # otherwise, result is a rows list
313
                self.iter = iter(self.result)
314
            
315
            def fetchone(self):
316
                try: return self.iter.next()
317
                except StopIteration: return None
318
    
319
    def esc_value(self, value):
320
        try: str_ = self.mogrify('%s', [value])
321
        except NotImplementedError, e:
322
            module = util.root_module(self.db)
323
            if module == 'MySQLdb':
324
                import _mysql
325
                str_ = _mysql.escape_string(value)
326
            else: raise e
327
        return strings.to_unicode(str_)
328
    
329
    def esc_name(self, name): return esc_name(self, name) # calls global func
330
    
331
    def std_code(self, str_):
332
        '''Standardizes SQL code.
333
        * Ensures that string literals are prefixed by `E`
334
        '''
335
        if str_.startswith("'"): str_ = 'E'+str_
336
        return str_
337
    
338
    def can_mogrify(self):
339
        module = util.root_module(self.db)
340
        return module == 'psycopg2'
341
    
342
    def mogrify(self, query, params=None):
343
        if self.can_mogrify(): return self.db.cursor().mogrify(query, params)
344
        else: raise NotImplementedError("Can't mogrify query")
345
    
346
    def print_notices(self):
347
        if hasattr(self.db, 'notices'):
348
            for msg in self.db.notices:
349
                if msg not in self._notices_seen:
350
                    self._notices_seen.add(msg)
351
                    self.log_debug(msg, level=2)
352
    
353
    def run_query(self, query, cacheable=False, log_level=2,
354
        debug_msg_ref=None):
355
        '''
356
        @param log_ignore_excs The log_level will be increased by 2 if the query
357
            throws one of these exceptions.
358
        @param debug_msg_ref If specified, the log message will be returned in
359
            this instead of being output. This allows you to filter log messages
360
            depending on the result of the query.
361
        '''
362
        assert query != None
363
        
364
        if self.autocommit and self.src != None:
365
            query = sql_gen.esc_comment(self.src)+'\t'+query
366
        
367
        if not self.caching: cacheable = False
368
        used_cache = False
369
        
370
        if self.debug:
371
            profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
372
        try:
373
            # Get cursor
374
            if cacheable:
375
                try: cur = self.query_results[query]
376
                except KeyError: cur = self.DbCursor(self)
377
                else: used_cache = True
378
            else: cur = self.db.cursor()
379
            
380
            # Run query
381
            try: cur.execute(query)
382
            except Exception, e:
383
                _add_cursor_info(e, self, query)
384
                raise
385
            else: self.do_autocommit()
386
        finally:
387
            if self.debug:
388
                profiler.stop(self.profile_row_ct)
389
                
390
                ## Log or return query
391
                
392
                query = strings.ustr(get_cur_query(cur, query))
393
                # Put the src comment on a separate line in the log file
394
                query = query.replace('\t', '\n', 1)
395
                
396
                msg = 'DB query: '
397
                
398
                if used_cache: msg += 'cache hit'
399
                elif cacheable: msg += 'cache miss'
400
                else: msg += 'non-cacheable'
401
                
402
                msg += ':\n'+profiler.msg()+'\n'+strings.as_code(query, 'SQL')
403
                
404
                if debug_msg_ref != None: debug_msg_ref[0] = msg
405
                else: self.log_debug(msg, log_level)
406
                
407
                self.print_notices()
408
        
409
        return cur
410
    
411
    def is_cached(self, query): return query in self.query_results
412
    
413
    def with_autocommit(self, func):
414
        import psycopg2.extensions
415
        
416
        prev_isolation_level = self.db.isolation_level
417
        self.db.set_isolation_level(
418
            psycopg2.extensions.ISOLATION_LEVEL_AUTOCOMMIT)
419
        try: return func()
420
        finally: self.db.set_isolation_level(prev_isolation_level)
421
    
422
    def with_savepoint(self, func):
423
        top = self._savepoint == 0
424
        savepoint = 'level_'+str(self._savepoint)
425
        
426
        if self.debug:
427
            self.log_debug('Begin transaction', level=4)
428
            profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
429
        
430
        # Must happen before running queries so they don't get autocommitted
431
        self._savepoint += 1
432
        
433
        if top: query = 'START TRANSACTION ISOLATION LEVEL READ COMMITTED'
434
        else: query = 'SAVEPOINT '+savepoint
435
        self.run_query(query, log_level=4)
436
        try:
437
            return func()
438
            if top: self.run_query('COMMIT', log_level=4)
439
        except:
440
            if top: query = 'ROLLBACK'
441
            else: query = 'ROLLBACK TO SAVEPOINT '+savepoint
442
            self.run_query(query, log_level=4)
443
            
444
            raise
445
        finally:
446
            # Always release savepoint, because after ROLLBACK TO SAVEPOINT,
447
            # "The savepoint remains valid and can be rolled back to again"
448
            # (http://www.postgresql.org/docs/8.3/static/sql-rollback-to.html).
449
            if not top:
450
                self.run_query('RELEASE SAVEPOINT '+savepoint, log_level=4)
451
            
452
            self._savepoint -= 1
453
            assert self._savepoint >= 0
454
            
455
            if self.debug:
456
                profiler.stop(self.profile_row_ct)
457
                self.log_debug('End transaction\n'+profiler.msg(), level=4)
458
            
459
            self.do_autocommit() # OK to do this after ROLLBACK TO SAVEPOINT
460
    
461
    def do_autocommit(self):
462
        '''Autocommits if outside savepoint'''
463
        assert self._savepoint >= 1
464
        if self.autocommit and self._savepoint == 1:
465
            self.log_debug('Autocommitting', level=4)
466
            self.db.commit()
467
    
468
    def col_info(self, col, cacheable=True):
469
        table = sql_gen.Table('columns', 'information_schema')
470
        cols = [sql_gen.Col('data_type'), sql_gen.Col('udt_name'),
471
            'column_default', sql_gen.Cast('boolean',
472
            sql_gen.Col('is_nullable'))]
473
        
474
        conds = [('table_name', col.table.name),
475
            ('column_name', strings.ustr(col.name))]
476
        schema = col.table.schema
477
        if schema != None: conds.append(('table_schema', schema))
478
        
479
        cur = select(self, table, cols, conds, order_by='table_schema', limit=1,
480
            cacheable=cacheable, log_level=4) # TODO: order by search_path order
481
        try: type_, extra_type, default, nullable = row(cur)
482
        except StopIteration: raise sql_gen.NoUnderlyingTableException(col)
483
        default = sql_gen.as_Code(default, self)
484
        if type_ == 'USER-DEFINED': type_ = extra_type
485
        elif type_ == 'ARRAY':
486
            type_ = sql_gen.ArrayType(strings.remove_prefix('_', extra_type,
487
                require=True))
488
        
489
        return sql_gen.TypedCol(col.name, type_, default, nullable)
490
    
491
    def TempFunction(self, name):
492
        if self.debug_temp: schema = None
493
        else: schema = 'pg_temp'
494
        return sql_gen.Function(name, schema)
495

    
496
connect = DbConn
497

    
498
##### Recoverable querying
499

    
500
def with_savepoint(db, func): return db.with_savepoint(func)
501

    
502
def run_query(db, query, recover=None, cacheable=False, log_level=2,
503
    log_ignore_excs=None, **kw_args):
504
    '''For params, see DbConn.run_query()'''
505
    if recover == None: recover = False
506
    if log_ignore_excs == None: log_ignore_excs = ()
507
    log_ignore_excs = tuple(log_ignore_excs)
508
    debug_msg_ref = [None]
509
    
510
    query = with_explain_comment(db, query)
511
    
512
    try:
513
        try:
514
            def run(): return db.run_query(query, cacheable, log_level,
515
                debug_msg_ref, **kw_args)
516
            if recover and not db.is_cached(query):
517
                return with_savepoint(db, run)
518
            else: return run() # don't need savepoint if cached
519
        except Exception, e:
520
            msg = strings.ustr(e.args[0])
521
            msg = re.sub(r'^(?:PL/Python: )?ValueError: ', r'', msg)
522
            
523
            match = re.match(r'^duplicate key value violates unique constraint '
524
                r'"(.+?)"', msg)
525
            if match:
526
                constraint, = match.groups()
527
                cols = []
528
                if recover: # need auto-rollback to run index_cols()
529
                    try: cols = index_cols(db, constraint)
530
                    except NotImplementedError: pass
531
                raise DuplicateKeyException(constraint, None, cols, e)
532
            
533
            match = re.match(r'^null value in column "(.+?)" violates not-null'
534
                r' constraint', msg)
535
            if match:
536
                col, = match.groups()
537
                raise NullValueException('NOT NULL', None, [col], e)
538
            
539
            match = re.match(r'^new row for relation "(.+?)" violates check '
540
                r'constraint "(.+?)"', msg)
541
            if match:
542
                table, constraint = match.groups()
543
                constraint = sql_gen.Col(constraint, table)
544
                cond = None
545
                if recover: # need auto-rollback to run constraint_cond()
546
                    try: cond = constraint_cond(db, constraint)
547
                    except NotImplementedError: pass
548
                raise CheckException(constraint.to_str(db), cond, [], e)
549
            
550
            match = re.match(r'^(?:invalid input (?:syntax|value)\b.*?'
551
                r'|.+? out of range): "(.+?)"', msg)
552
            if match:
553
                value, = match.groups()
554
                raise InvalidValueException(strings.to_unicode(value), e)
555
            
556
            match = re.match(r'^column "(.+?)" is of type (.+?) but expression '
557
                r'is of type', msg)
558
            if match:
559
                col, type_ = match.groups()
560
                raise MissingCastException(type_, col, e)
561
            
562
            match = re.match(r'^could not determine polymorphic type because '
563
                r'input has type "unknown"', msg)
564
            if match: raise MissingCastException('text', None, e)
565
            
566
            match = re.match(r'^.+? types .+? and .+? cannot be matched', msg)
567
            if match: raise MissingCastException('text', None, e)
568
            
569
            typed_name_re = r'^(\S+) "(.+?)"(?: of relation ".+?")?'
570
            
571
            match = re.match(typed_name_re+r'.*? already exists', msg)
572
            if match:
573
                type_, name = match.groups()
574
                raise DuplicateException(type_, name, e)
575
            
576
            match = re.match(r'more than one (\S+) named ""(.+?)""', msg)
577
            if match:
578
                type_, name = match.groups()
579
                raise DuplicateException(type_, name, e)
580
            
581
            match = re.match(typed_name_re+r' does not exist', msg)
582
            if match:
583
                type_, name = match.groups()
584
                raise DoesNotExistException(type_, name, e)
585
            
586
            raise # no specific exception raised
587
    except log_ignore_excs:
588
        log_level += 2
589
        raise
590
    finally:
591
        if debug_msg_ref[0] != None: db.log_debug(debug_msg_ref[0], log_level)
592

    
593
##### Basic queries
594

    
595
def is_explainable(query):
596
    # See <http://www.postgresql.org/docs/8.3/static/sql-explain.html#AEN57749>
597
    return re.match(r'^(?:SELECT|INSERT|UPDATE|DELETE|VALUES|EXECUTE|DECLARE)\b'
598
        , query)
599

    
600
def explain(db, query, **kw_args):
601
    '''
602
    For params, see run_query().
603
    '''
604
    kw_args.setdefault('log_level', 4)
605
    
606
    return strings.ustr(strings.join_lines(values(run_query(db,
607
        'EXPLAIN '+query, recover=True, cacheable=True, **kw_args))))
608
        # not a higher log_level because it's useful to see what query is being
609
        # run before it's executed, which EXPLAIN effectively provides
610

    
611
def has_comment(query): return query.endswith('*/')
612

    
613
def with_explain_comment(db, query, **kw_args):
614
    if db.autoexplain and not has_comment(query) and is_explainable(query):
615
        query += '\n'+sql_gen.esc_comment(' EXPLAIN:\n'
616
            +explain(db, query, **kw_args))
617
    return query
618

    
619
def next_version(name):
620
    version = 1 # first existing name was version 0
621
    match = re.match(r'^(.*)#(\d+)$', name)
622
    if match:
623
        name, version = match.groups()
624
        version = int(version)+1
625
    return sql_gen.concat(name, '#'+str(version))
626

    
627
def lock_table(db, table, mode):
628
    table = sql_gen.as_Table(table)
629
    run_query(db, 'LOCK TABLE '+table.to_str(db)+' IN '+mode+' MODE')
630

    
631
def run_query_into(db, query, into=None, add_pkey_=False, **kw_args):
632
    '''Outputs a query to a temp table.
633
    For params, see run_query().
634
    '''
635
    if into == None: return run_query(db, query, **kw_args)
636
    
637
    assert isinstance(into, sql_gen.Table)
638
    
639
    into.is_temp = True
640
    # "temporary tables cannot specify a schema name", so remove schema
641
    into.schema = None
642
    
643
    kw_args['recover'] = True
644
    kw_args.setdefault('log_ignore_excs', (DuplicateException,))
645
    
646
    temp = not db.debug_temp # tables are permanent in debug_temp mode
647
    
648
    # Create table
649
    while True:
650
        create_query = 'CREATE'
651
        if temp: create_query += ' TEMP'
652
        create_query += ' TABLE '+into.to_str(db)+' AS\n'+query
653
        
654
        try:
655
            cur = run_query(db, create_query, **kw_args)
656
                # CREATE TABLE AS sets rowcount to # rows in query
657
            break
658
        except DuplicateException, e:
659
            into.name = next_version(into.name)
660
            # try again with next version of name
661
    
662
    if add_pkey_: add_pkey(db, into)
663
    
664
    # According to the PostgreSQL doc, "The autovacuum daemon cannot access and
665
    # therefore cannot vacuum or analyze temporary tables. [...] if a temporary
666
    # table is going to be used in complex queries, it is wise to run ANALYZE on
667
    # the temporary table after it is populated."
668
    # (http://www.postgresql.org/docs/9.1/static/sql-createtable.html)
669
    # If into is not a temp table, ANALYZE is useful but not required.
670
    analyze(db, into)
671
    
672
    return cur
673

    
674
order_by_pkey = object() # tells mk_select() to order by the pkey
675

    
676
distinct_on_all = object() # tells mk_select() to SELECT DISTINCT ON all columns
677

    
678
def mk_select(db, tables=None, fields=None, conds=None, distinct_on=[],
679
    limit=None, start=None, order_by=order_by_pkey, default_table=None,
680
    explain=True):
681
    '''
682
    @param tables The single table to select from, or a list of tables to join
683
        together, with tables after the first being sql_gen.Join objects
684
    @param fields Use None to select all fields in the table
685
    @param conds WHERE conditions: [(compare_left_side, compare_right_side),...]
686
        * container can be any iterable type
687
        * compare_left_side: sql_gen.Code|str (for col name)
688
        * compare_right_side: sql_gen.ValueCond|literal value
689
    @param distinct_on The columns to SELECT DISTINCT ON, or distinct_on_all to
690
        use all columns
691
    @return query
692
    '''
693
    # Parse tables param
694
    tables = lists.mk_seq(tables)
695
    tables = list(tables) # don't modify input! (list() copies input)
696
    table0 = sql_gen.as_Table(tables.pop(0)) # first table is separate
697
    
698
    # Parse other params
699
    if conds == None: conds = []
700
    elif dicts.is_dict(conds): conds = conds.items()
701
    conds = list(conds) # don't modify input! (list() copies input)
702
    assert limit == None or isinstance(limit, (int, long))
703
    assert start == None or isinstance(start, (int, long))
704
    if order_by is order_by_pkey:
705
        if table0 == None or distinct_on != []: order_by = None
706
        else: order_by = pkey(db, table0, recover=True)
707
    
708
    query = 'SELECT'
709
    
710
    def parse_col(col): return sql_gen.as_Col(col, default_table).to_str(db)
711
    
712
    # DISTINCT ON columns
713
    if distinct_on != []:
714
        query += '\nDISTINCT'
715
        if distinct_on is not distinct_on_all:
716
            query += ' ON ('+(', '.join(map(parse_col, distinct_on)))+')'
717
    
718
    # Columns
719
    if query.find('\n') >= 0: whitespace = '\n'
720
    else: whitespace = ' '
721
    if fields == None: query += whitespace+'*'
722
    else:
723
        assert fields != []
724
        if len(fields) > 1: whitespace = '\n'
725
        query += whitespace+('\n, '.join(map(parse_col, fields)))
726
    
727
    # Main table
728
    if query.find('\n') >= 0 or len(tables) > 0: whitespace = '\n'
729
    else: whitespace = ' '
730
    if table0 != None: query += whitespace+'FROM '+table0.to_str(db)
731
    
732
    # Add joins
733
    left_table = table0
734
    for join_ in tables:
735
        table = join_.table
736
        
737
        # Parse special values
738
        if join_.type_ is sql_gen.filter_out: # filter no match
739
            conds.append((sql_gen.Col(table_not_null_col(db, table), table),
740
                sql_gen.CompareCond(None, '~=')))
741
        
742
        query += '\n'+join_.to_str(db, left_table)
743
        
744
        left_table = table
745
    
746
    missing = True
747
    if conds != []:
748
        if len(conds) == 1: whitespace = ' '
749
        else: whitespace = '\n'
750
        query += '\n'+sql_gen.combine_conds([sql_gen.ColValueCond(l, r)
751
            .to_str(db) for l, r in conds], 'WHERE')
752
    if order_by != None:
753
        query += '\nORDER BY '+sql_gen.as_Col(order_by, table0).to_str(db)
754
    if limit != None: query += '\nLIMIT '+str(limit)
755
    if start != None:
756
        if start != 0: query += '\nOFFSET '+str(start)
757
    
758
    if explain: query = with_explain_comment(db, query)
759
    
760
    return query
761

    
762
def select(db, *args, **kw_args):
763
    '''For params, see mk_select() and run_query()'''
764
    recover = kw_args.pop('recover', None)
765
    cacheable = kw_args.pop('cacheable', True)
766
    log_level = kw_args.pop('log_level', 2)
767
    
768
    return run_query(db, mk_select(db, *args, **kw_args), recover, cacheable,
769
        log_level=log_level)
770

    
771
def mk_insert_select(db, table, cols=None, select_query=None, returning=None,
772
    embeddable=False, ignore=False, src=None):
773
    '''
774
    @param returning str|None An inserted column (such as pkey) to return
775
    @param embeddable Whether the query should be embeddable as a nested SELECT.
776
        Warning: If you set this and cacheable=True when the query is run, the
777
        query will be fully cached, not just if it raises an exception.
778
    @param ignore Whether to ignore duplicate keys.
779
    @param src Will be included in the name of any created function, to help
780
        identify the data source in pg_stat_activity.
781
    '''
782
    table = sql_gen.remove_table_rename(sql_gen.as_Table(table))
783
    if cols == []: cols = None # no cols (all defaults) = unknown col names
784
    if cols != None: cols = [sql_gen.to_name_only_col(c, table) for c in cols]
785
    if select_query == None: select_query = 'DEFAULT VALUES'
786
    if returning != None: returning = sql_gen.as_Col(returning, table)
787
    
788
    first_line = 'INSERT INTO '+table.to_str(db)
789
    
790
    def mk_insert(select_query):
791
        query = first_line
792
        if cols != None:
793
            query += '\n('+(', '.join((c.to_str(db) for c in cols)))+')'
794
        query += '\n'+select_query
795
        
796
        if returning != None:
797
            returning_name_col = sql_gen.to_name_only_col(returning)
798
            query += '\nRETURNING '+returning_name_col.to_str(db)
799
        
800
        return query
801
    
802
    return_type = sql_gen.CustomCode('unknown')
803
    if returning != None: return_type = sql_gen.ColType(returning)
804
    
805
    if ignore:
806
        # Always return something to set the correct rowcount
807
        if returning == None: returning = sql_gen.NamedCol('NULL', None)
808
        
809
        embeddable = True # must use function
810
        
811
        if cols == None: row = [sql_gen.Col(sql_gen.all_cols, 'row')]
812
        else: row = [sql_gen.Col(c.name, 'row') for c in cols]
813
        
814
        query = sql_gen.RowExcIgnore(sql_gen.RowType(table), select_query,
815
            sql_gen.ReturnQuery(mk_insert(sql_gen.Values(row).to_str(db))),
816
            cols)
817
    else: query = mk_insert(select_query)
818
    
819
    if embeddable:
820
        # Create function
821
        function_name = sql_gen.clean_name(first_line)
822
        if src != None: function_name = src+': '+function_name
823
        while True:
824
            try:
825
                func = db.TempFunction(function_name)
826
                def_ = sql_gen.FunctionDef(func, sql_gen.SetOf(return_type),
827
                    query)
828
                
829
                run_query(db, def_.to_str(db), recover=True, cacheable=True,
830
                    log_ignore_excs=(DuplicateException,))
831
                break # this version was successful
832
            except DuplicateException, e:
833
                function_name = next_version(function_name)
834
                # try again with next version of name
835
        
836
        # Return query that uses function
837
        cols = None
838
        if returning != None: cols = [returning]
839
        func_table = sql_gen.NamedTable('f', sql_gen.FunctionCall(func), cols)
840
            # AS clause requires function alias
841
        return mk_select(db, func_table, order_by=None)
842
    
843
    return query
844

    
845
def insert_select(db, table, *args, **kw_args):
846
    '''For params, see mk_insert_select() and run_query_into()
847
    @param into sql_gen.Table with suggested name of temp table to put RETURNING
848
        values in
849
    '''
850
    returning = kw_args.get('returning', None)
851
    ignore = kw_args.get('ignore', False)
852
    
853
    into = kw_args.pop('into', None)
854
    if into != None: kw_args['embeddable'] = True
855
    recover = kw_args.pop('recover', None)
856
    if ignore: recover = True
857
    cacheable = kw_args.pop('cacheable', True)
858
    log_level = kw_args.pop('log_level', 2)
859
    
860
    rowcount_only = ignore and returning == None # keep NULL rows on server
861
    if rowcount_only: into = sql_gen.Table('rowcount')
862
    
863
    cur = run_query_into(db, mk_insert_select(db, table, *args, **kw_args),
864
        into, recover=recover, cacheable=cacheable, log_level=log_level)
865
    if rowcount_only: empty_temp(db, into)
866
    autoanalyze(db, table)
867
    return cur
868

    
869
default = sql_gen.default # tells insert() to use the default value for a column
870

    
871
def insert(db, table, row, *args, **kw_args):
872
    '''For params, see insert_select()'''
873
    ignore = kw_args.pop('ignore', False)
874
    if ignore: kw_args.setdefault('recover', True)
875
    
876
    if lists.is_seq(row): cols = None
877
    else:
878
        cols = row.keys()
879
        row = row.values()
880
    row = list(row) # ensure that "== []" works
881
    
882
    if row == []: query = None
883
    else: query = sql_gen.Values(row).to_str(db)
884
    
885
    try: return insert_select(db, table, cols, query, *args, **kw_args)
886
    except (DuplicateKeyException, NullValueException):
887
        if not ignore: raise
888
        return None
889

    
890
def mk_update(db, table, changes=None, cond=None, in_place=False,
891
    cacheable_=True):
892
    '''
893
    @param changes [(col, new_value),...]
894
        * container can be any iterable type
895
        * col: sql_gen.Code|str (for col name)
896
        * new_value: sql_gen.Code|literal value
897
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
898
    @param in_place If set, locks the table and updates rows in place.
899
        This avoids creating dead rows in PostgreSQL.
900
        * cond must be None
901
    @param cacheable_ Whether column structure information used to generate the
902
        query can be cached
903
    @return str query
904
    '''
905
    table = sql_gen.as_Table(table)
906
    changes = [(sql_gen.to_name_only_col(c, table), sql_gen.as_Value(v))
907
        for c, v in changes]
908
    
909
    if in_place:
910
        assert cond == None
911
        
912
        query = 'ALTER TABLE '+table.to_str(db)+'\n'
913
        query += ',\n'.join(('ALTER COLUMN '+c.to_str(db)+' TYPE '
914
            +db.col_info(sql_gen.with_default_table(c, table), cacheable_).type
915
            +'\nUSING '+v.to_str(db) for c, v in changes))
916
    else:
917
        query = 'UPDATE '+table.to_str(db)+'\nSET\n'
918
        query += ',\n'.join((c.to_str(db)+' = '+v.to_str(db)
919
            for c, v in changes))
920
        if cond != None: query += '\nWHERE\n'+cond.to_str(db)
921
    
922
    query = with_explain_comment(db, query)
923
    
924
    return query
925

    
926
def update(db, table, *args, **kw_args):
927
    '''For params, see mk_update() and run_query()'''
928
    recover = kw_args.pop('recover', None)
929
    cacheable = kw_args.pop('cacheable', False)
930
    log_level = kw_args.pop('log_level', 2)
931
    
932
    cur = run_query(db, mk_update(db, table, *args, **kw_args), recover,
933
        cacheable, log_level=log_level)
934
    autoanalyze(db, table)
935
    return cur
936

    
937
def mk_delete(db, table, cond=None):
938
    '''
939
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
940
    @return str query
941
    '''
942
    query = 'DELETE FROM '+table.to_str(db)
943
    if cond != None: query += '\nWHERE '+cond.to_str(db)
944
    
945
    query = with_explain_comment(db, query)
946
    
947
    return query
948

    
949
def delete(db, table, *args, **kw_args):
950
    '''For params, see mk_delete() and run_query()'''
951
    recover = kw_args.pop('recover', None)
952
    cacheable = kw_args.pop('cacheable', True)
953
    log_level = kw_args.pop('log_level', 2)
954
    
955
    cur = run_query(db, mk_delete(db, table, *args, **kw_args), recover,
956
        cacheable, log_level=log_level)
957
    autoanalyze(db, table)
958
    return cur
959

    
960
def last_insert_id(db):
961
    module = util.root_module(db.db)
962
    if module == 'psycopg2': return value(run_query(db, 'SELECT lastval()'))
963
    elif module == 'MySQLdb': return db.insert_id()
964
    else: return None
965

    
966
def define_func(db, def_):
967
    func = def_.function
968
    while True:
969
        try:
970
            run_query(db, def_.to_str(db), recover=True, cacheable=True,
971
                log_ignore_excs=(DuplicateException,))
972
            break # successful
973
        except DuplicateException:
974
            func.name = next_version(func.name)
975
            # try again with next version of name
976

    
977
def mk_flatten_mapping(db, into, cols, preserve=[], as_items=False):
978
    '''Creates a mapping from original column names (which may have collisions)
979
    to names that will be distinct among the columns' tables.
980
    This is meant to be used for several tables that are being joined together.
981
    @param cols The columns to combine. Duplicates will be removed.
982
    @param into The table for the new columns.
983
    @param preserve [sql_gen.Col...] Columns not to rename. Note that these
984
        columns will be included in the mapping even if they are not in cols.
985
        The tables of the provided Col objects will be changed to into, so make
986
        copies of them if you want to keep the original tables.
987
    @param as_items Whether to return a list of dict items instead of a dict
988
    @return dict(orig_col=new_col, ...)
989
        * orig_col: sql_gen.Col(orig_col_name, orig_table)
990
        * new_col: sql_gen.Col(orig_col_name, into)
991
        * All mappings use the into table so its name can easily be
992
          changed for all columns at once
993
    '''
994
    cols = lists.uniqify(cols)
995
    
996
    items = []
997
    for col in preserve:
998
        orig_col = copy.copy(col)
999
        col.table = into
1000
        items.append((orig_col, col))
1001
    preserve = set(preserve)
1002
    for col in cols:
1003
        if col not in preserve:
1004
            items.append((col, sql_gen.Col(strings.ustr(col), into, col.srcs)))
1005
    
1006
    if not as_items: items = dict(items)
1007
    return items
1008

    
1009
def flatten(db, into, joins, cols, limit=None, start=None, **kw_args):
1010
    '''For params, see mk_flatten_mapping()
1011
    @return See return value of mk_flatten_mapping()
1012
    '''
1013
    items = mk_flatten_mapping(db, into, cols, as_items=True, **kw_args)
1014
    cols = [sql_gen.NamedCol(new.name, old) for old, new in items]
1015
    run_query_into(db, mk_select(db, joins, cols, order_by=None, limit=limit,
1016
        start=start), into=into, add_pkey_=True)
1017
        # don't cache because the temp table will usually be truncated after use
1018
    return dict(items)
1019

    
1020
##### Database structure introspection
1021

    
1022
#### Expressions
1023

    
1024
def paren_re(re_): return r'(?:'+re_+r'|\('+re_+r'\))'
1025

    
1026
true_re = paren_re(r'true')
1027
false_re = paren_re(r'false')
1028
bool_re = r'(?:'+true_re+r'|'+false_re+r')'
1029

    
1030
def logic_op_re(op, value_re):
1031
    op_re = ' '+op+' '
1032
    return '(?:'+op_re+value_re+'|'+value_re+op_re+')'
1033

    
1034
def simplify_parens(expr):
1035
    return regexp.sub_nested(r'\((true|false|\([^()]*\))\)', r'\1', expr)
1036

    
1037
def simplify_expr(expr):
1038
    expr = expr.replace('(NULL IS NULL)', 'true')
1039
    expr = expr.replace('(NULL IS NOT NULL)', 'false')
1040
    expr = re.sub(logic_op_re('OR', bool_re), r'', expr)
1041
    expr = simplify_parens(expr)
1042
    return expr
1043

    
1044
name_re = r'(?:\w+|(?:"[^"]*")+)'
1045

    
1046
def parse_expr_col(str_):
1047
    match = re.match(r'^\('+name_re+r'\(('+name_re+r').*\)\)$', str_)
1048
    if match: str_ = match.group(1)
1049
    return sql_gen.unesc_name(str_)
1050

    
1051
def map_expr(db, expr, mapping, in_cols_found=None):
1052
    '''Replaces output columns with input columns in an expression.
1053
    @param in_cols_found If set, will be filled in with the expr's (input) cols
1054
    '''
1055
    for out, in_ in mapping.iteritems():
1056
        orig_expr = expr
1057
        out = sql_gen.to_name_only_col(out)
1058
        in_str = sql_gen.to_name_only_col(sql_gen.remove_col_rename(in_)
1059
            ).to_str(db)
1060
        
1061
        # Replace out both with and without quotes
1062
        expr = expr.replace(out.to_str(db), in_str)
1063
        expr = re.sub(r'(?<!["\'\.\[])\b'+out.name+r'\b(?!["\'\.=\]])', in_str,
1064
            expr)
1065
        
1066
        if in_cols_found != None and expr != orig_expr: # replaced something
1067
            in_cols_found.append(in_)
1068
    
1069
    return simplify_expr(expr)
1070

    
1071
#### Tables
1072

    
1073
def tables(db, schema_like='public', table_like='%', exact=False,
1074
    cacheable=True):
1075
    if exact: compare = '='
1076
    else: compare = 'LIKE'
1077
    
1078
    module = util.root_module(db.db)
1079
    if module == 'psycopg2':
1080
        conds = [('schemaname', sql_gen.CompareCond(schema_like, compare)),
1081
            ('tablename', sql_gen.CompareCond(table_like, compare))]
1082
        return values(select(db, 'pg_tables', ['tablename'], conds,
1083
            order_by='tablename', cacheable=cacheable, log_level=4))
1084
    elif module == 'MySQLdb':
1085
        return values(run_query(db, 'SHOW TABLES LIKE '+db.esc_value(table_like)
1086
            , cacheable=True, log_level=4))
1087
    else: raise NotImplementedError("Can't list tables for "+module+' database')
1088

    
1089
def table_exists(db, table, cacheable=True):
1090
    table = sql_gen.as_Table(table)
1091
    return list(tables(db, table.schema, table.name, True, cacheable)) != []
1092

    
1093
def table_row_count(db, table, recover=None):
1094
    return value(run_query(db, mk_select(db, table, [sql_gen.row_count],
1095
        order_by=None), recover=recover, log_level=3))
1096

    
1097
def table_col_names(db, table, recover=None):
1098
    return list(col_names(select(db, table, limit=0, order_by=None,
1099
        recover=recover, log_level=4)))
1100

    
1101
pkey_col = 'row_num'
1102

    
1103
def pkey(db, table, recover=None):
1104
    '''If no pkey, returns the first column in the table.'''
1105
    table = sql_gen.as_Table(table)
1106
    
1107
    join_cols = ['table_schema', 'table_name', 'constraint_schema',
1108
        'constraint_name']
1109
    tables = [sql_gen.Table('key_column_usage', 'information_schema'),
1110
        sql_gen.Join(sql_gen.Table('table_constraints', 'information_schema'),
1111
            dict(((c, sql_gen.join_same_not_null) for c in join_cols)))]
1112
    cols = [sql_gen.Col('column_name')]
1113
    
1114
    conds = [('constraint_type', 'PRIMARY KEY'), ('table_name', table.name)]
1115
    schema = table.schema
1116
    if schema != None: conds.append(('table_schema', schema))
1117
    order_by = 'position_in_unique_constraint'
1118
    
1119
    try: return value(select(db, tables, cols, conds, order_by=order_by,
1120
        limit=1, log_level=4))
1121
    except StopIteration: return table_col_names(db, table, recover)[0]
1122

    
1123
def pkey_col_(db, table, *args, **kw_args):
1124
    return sql_gen.Col(pkey(db, table, *args, **kw_args), table)
1125

    
1126
not_null_col = 'not_null_col'
1127

    
1128
def table_not_null_col(db, table, recover=None):
1129
    '''Name assumed to be the value of not_null_col. If not found, uses pkey.'''
1130
    if not_null_col in table_col_names(db, table, recover): return not_null_col
1131
    else: return pkey(db, table, recover)
1132

    
1133
def constraint_cond(db, constraint):
1134
    module = util.root_module(db.db)
1135
    if module == 'psycopg2':
1136
        table_str = sql_gen.Literal(constraint.table.to_str(db))
1137
        name_str = sql_gen.Literal(constraint.name)
1138
        return value(run_query(db, '''\
1139
SELECT consrc
1140
FROM pg_constraint
1141
WHERE
1142
conrelid = '''+table_str.to_str(db)+'''::regclass
1143
AND conname = '''+name_str.to_str(db)+'''
1144
'''
1145
            , cacheable=True, log_level=4))
1146
    else: raise NotImplementedError("Can't list index columns for "+module+
1147
        ' database')
1148

    
1149
def index_cols(db, index):
1150
    '''Can also use this for UNIQUE constraints, because a UNIQUE index is
1151
    automatically created. When you don't know whether something is a UNIQUE
1152
    constraint or a UNIQUE index, use this function.'''
1153
    index = sql_gen.as_Table(index)
1154
    module = util.root_module(db.db)
1155
    if module == 'psycopg2':
1156
        qual_index = sql_gen.Literal(index.to_str(db))
1157
        return map(parse_expr_col, values(run_query(db, '''\
1158
SELECT pg_get_indexdef(indexrelid, generate_series(1, indnatts), true)
1159
FROM pg_index
1160
WHERE indexrelid = '''+qual_index.to_str(db)+'''::regclass
1161
'''
1162
            , cacheable=True, log_level=4)))
1163
    else: raise NotImplementedError("Can't list index columns for "+module+
1164
        ' database')
1165

    
1166
#### Functions
1167

    
1168
def function_exists(db, function):
1169
    qual_function = sql_gen.Literal(function.to_str(db))
1170
    try:
1171
        select(db, fields=[sql_gen.Cast('regproc', qual_function)],
1172
            recover=True, cacheable=True, log_level=4)
1173
    except DoesNotExistException: return False
1174
    except DuplicateException: return True # overloaded function
1175
    else: return True
1176

    
1177
##### Structural changes
1178

    
1179
#### Columns
1180

    
1181
def add_col(db, table, col, comment=None, if_not_exists=False, **kw_args):
1182
    '''
1183
    @param col TypedCol Name may be versioned, so be sure to propagate any
1184
        renaming back to any source column for the TypedCol.
1185
    @param comment None|str SQL comment used to distinguish columns of the same
1186
        name from each other when they contain different data, to allow the
1187
        ADD COLUMN query to be cached. If not set, query will not be cached.
1188
    '''
1189
    assert isinstance(col, sql_gen.TypedCol)
1190
    
1191
    while True:
1192
        str_ = 'ALTER TABLE '+table.to_str(db)+' ADD COLUMN '+col.to_str(db)
1193
        if comment != None: str_ += ' '+sql_gen.esc_comment(comment)
1194
        
1195
        try:
1196
            run_query(db, str_, recover=True, cacheable=True, **kw_args)
1197
            break
1198
        except DuplicateException:
1199
            if if_not_exists: raise
1200
            col.name = next_version(col.name)
1201
            # try again with next version of name
1202

    
1203
def add_not_null(db, col):
1204
    table = col.table
1205
    col = sql_gen.to_name_only_col(col)
1206
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ALTER COLUMN '
1207
        +col.to_str(db)+' SET NOT NULL', cacheable=True, log_level=3)
1208

    
1209
def drop_not_null(db, col):
1210
    table = col.table
1211
    col = sql_gen.to_name_only_col(col)
1212
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ALTER COLUMN '
1213
        +col.to_str(db)+' DROP NOT NULL', cacheable=True, log_level=3)
1214

    
1215
row_num_col = '_row_num'
1216

    
1217
row_num_col_def = sql_gen.TypedCol('', 'serial', nullable=False,
1218
    constraints='PRIMARY KEY')
1219

    
1220
def add_row_num(db, table, name=row_num_col):
1221
    '''Adds a row number column to a table. Its definition is in
1222
    row_num_col_def. It will be the primary key.'''
1223
    col_def = copy.copy(row_num_col_def)
1224
    col_def.name = name
1225
    add_col(db, table, col_def, comment='', if_not_exists=True, log_level=3)
1226

    
1227
#### Indexes
1228

    
1229
def add_pkey(db, table, cols=None, recover=None):
1230
    '''Adds a primary key.
1231
    @param cols [sql_gen.Col,...] The columns in the primary key.
1232
        Defaults to the first column in the table.
1233
    @pre The table must not already have a primary key.
1234
    '''
1235
    table = sql_gen.as_Table(table)
1236
    if cols == None: cols = [pkey(db, table, recover)]
1237
    col_strs = [sql_gen.to_name_only_col(v).to_str(db) for v in cols]
1238
    
1239
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ADD PRIMARY KEY ('
1240
        +(', '.join(col_strs))+')', recover=True, cacheable=True, log_level=3,
1241
        log_ignore_excs=(DuplicateException,))
1242

    
1243
def add_index(db, exprs, table=None, unique=False, ensure_not_null_=True):
1244
    '''Adds an index on column(s) or expression(s) if it doesn't already exist.
1245
    Currently, only function calls and literal values are supported expressions.
1246
    @param ensure_not_null_ If set, translates NULL values to sentinel values.
1247
        This allows indexes to be used for comparisons where NULLs are equal.
1248
    '''
1249
    exprs = lists.mk_seq(exprs)
1250
    
1251
    # Parse exprs
1252
    old_exprs = exprs[:]
1253
    exprs = []
1254
    cols = []
1255
    for i, expr in enumerate(old_exprs):
1256
        expr = sql_gen.as_Col(expr, table)
1257
        
1258
        # Handle nullable columns
1259
        if ensure_not_null_:
1260
            try: expr = sql_gen.ensure_not_null(db, expr)
1261
            except KeyError: pass # unknown type, so just create plain index
1262
        
1263
        # Extract col
1264
        expr = copy.deepcopy(expr) # don't modify input!
1265
        col = expr
1266
        if isinstance(expr, sql_gen.FunctionCall): col = expr.args[0]
1267
        expr = sql_gen.cast_literal(expr)
1268
        if not isinstance(expr, (sql_gen.Expr, sql_gen.Col)):
1269
            expr = sql_gen.Expr(expr)
1270
            
1271
        
1272
        # Extract table
1273
        if table == None:
1274
            assert sql_gen.is_table_col(col)
1275
            table = col.table
1276
        
1277
        if isinstance(col, sql_gen.Col): col.table = None
1278
        
1279
        exprs.append(expr)
1280
        cols.append(col)
1281
    
1282
    table = sql_gen.as_Table(table)
1283
    
1284
    # Add index
1285
    str_ = 'CREATE'
1286
    if unique: str_ += ' UNIQUE'
1287
    str_ += ' INDEX ON '+table.to_str(db)+' ('+(
1288
        ', '.join((v.to_str(db) for v in exprs)))+')'
1289
    run_query(db, str_, recover=True, cacheable=True, log_level=3)
1290

    
1291
already_indexed = object() # tells add_indexes() the pkey has already been added
1292

    
1293
def add_indexes(db, table, has_pkey=True):
1294
    '''Adds an index on all columns in a table.
1295
    @param has_pkey bool|already_indexed Whether a pkey instead of a regular
1296
        index should be added on the first column.
1297
        * If already_indexed, the pkey is assumed to have already been added
1298
    '''
1299
    cols = table_col_names(db, table)
1300
    if has_pkey:
1301
        if has_pkey is not already_indexed: add_pkey(db, table)
1302
        cols = cols[1:]
1303
    for col in cols: add_index(db, col, table)
1304

    
1305
#### Tables
1306

    
1307
### Maintenance
1308

    
1309
def analyze(db, table):
1310
    table = sql_gen.as_Table(table)
1311
    run_query(db, 'ANALYZE '+table.to_str(db), log_level=3)
1312

    
1313
def autoanalyze(db, table):
1314
    if db.autoanalyze: analyze(db, table)
1315

    
1316
def vacuum(db, table):
1317
    table = sql_gen.as_Table(table)
1318
    db.with_autocommit(lambda: run_query(db, 'VACUUM ANALYZE '+table.to_str(db),
1319
        log_level=3))
1320

    
1321
### Lifecycle
1322

    
1323
def drop(db, type_, name):
1324
    name = sql_gen.as_Name(name)
1325
    run_query(db, 'DROP '+type_+' IF EXISTS '+name.to_str(db)+' CASCADE')
1326

    
1327
def drop_table(db, table): drop(db, 'TABLE', table)
1328

    
1329
def create_table(db, table, cols=[], has_pkey=True, col_indexes=True,
1330
    like=None):
1331
    '''Creates a table.
1332
    @param cols [sql_gen.TypedCol,...] The column names and types
1333
    @param has_pkey If set, the first column becomes the primary key.
1334
    @param col_indexes bool|[ref]
1335
        * If True, indexes will be added on all non-pkey columns.
1336
        * If a list reference, [0] will be set to a function to do this.
1337
          This can be used to delay index creation until the table is populated.
1338
    '''
1339
    table = sql_gen.as_Table(table)
1340
    
1341
    if like != None:
1342
        cols = [sql_gen.CustomCode('LIKE '+like.to_str(db)+' INCLUDING ALL')
1343
            ]+cols
1344
    if has_pkey:
1345
        cols[0] = pkey = copy.copy(cols[0]) # don't modify input!
1346
        pkey.constraints = 'PRIMARY KEY'
1347
    
1348
    temp = table.is_temp and not db.debug_temp
1349
        # temp tables permanent in debug_temp mode
1350
    
1351
    # Create table
1352
    def create():
1353
        str_ = 'CREATE'
1354
        if temp: str_ += ' TEMP'
1355
        str_ += ' TABLE '+table.to_str(db)+' (\n'
1356
        str_ += '\n, '.join(c.to_str(db) for c in cols)
1357
        str_ += '\n);'
1358
        
1359
        run_query(db, str_, recover=True, cacheable=True, log_level=2,
1360
            log_ignore_excs=(DuplicateException,))
1361
    if table.is_temp:
1362
        while True:
1363
            try:
1364
                create()
1365
                break
1366
            except DuplicateException:
1367
                table.name = next_version(table.name)
1368
                # try again with next version of name
1369
    else: create()
1370
    
1371
    # Add indexes
1372
    if has_pkey: has_pkey = already_indexed
1373
    def add_indexes_(): add_indexes(db, table, has_pkey)
1374
    if isinstance(col_indexes, list): col_indexes[0] = add_indexes_ # defer
1375
    elif col_indexes: add_indexes_() # add now
1376

    
1377
def copy_table_struct(db, src, dest):
1378
    '''Creates a structure-only copy of a table. (Does not copy data.)'''
1379
    create_table(db, dest, has_pkey=False, col_indexes=False, like=src)
1380

    
1381
### Data
1382

    
1383
def truncate(db, table, schema='public', **kw_args):
1384
    '''For params, see run_query()'''
1385
    table = sql_gen.as_Table(table, schema)
1386
    return run_query(db, 'TRUNCATE '+table.to_str(db)+' CASCADE', **kw_args)
1387

    
1388
def empty_temp(db, tables):
1389
    tables = lists.mk_seq(tables)
1390
    for table in tables: truncate(db, table, log_level=3)
1391

    
1392
def empty_db(db, schema='public', **kw_args):
1393
    '''For kw_args, see tables()'''
1394
    for table in tables(db, schema, **kw_args): truncate(db, table, schema)
1395

    
1396
def distinct_table(db, table, distinct_on):
1397
    '''Creates a copy of a temp table which is distinct on the given columns.
1398
    The old and new tables will both get an index on these columns, to
1399
    facilitate merge joins.
1400
    @param distinct_on If empty, creates a table with one row. This is useful if
1401
        your distinct_on columns are all literal values.
1402
    @return The new table.
1403
    '''
1404
    new_table = sql_gen.suffixed_table(table, '_distinct')
1405
    distinct_on = filter(sql_gen.is_table_col, distinct_on)
1406
    
1407
    copy_table_struct(db, table, new_table)
1408
    
1409
    limit = None
1410
    if distinct_on == []: limit = 1 # one sample row
1411
    else:
1412
        add_index(db, distinct_on, new_table, unique=True)
1413
        add_index(db, distinct_on, table) # for join optimization
1414
    
1415
    insert_select(db, new_table, None, mk_select(db, table, order_by=None,
1416
        limit=limit), ignore=True)
1417
    analyze(db, new_table)
1418
    
1419
    return new_table
(28-28/42)