Project

General

Profile

1 11 aaronmk
# Database access
2
3 1869 aaronmk
import copy
4 11 aaronmk
import re
5 3238 aaronmk
import time
6 865 aaronmk
import warnings
7 11 aaronmk
8 300 aaronmk
import exc
9 1909 aaronmk
import dicts
10 1893 aaronmk
import iters
11 1960 aaronmk
import lists
12 3241 aaronmk
import profiling
13 1889 aaronmk
from Proxy import Proxy
14 1872 aaronmk
import rand
15 2217 aaronmk
import sql_gen
16 862 aaronmk
import strings
17 131 aaronmk
import util
18 11 aaronmk
19 832 aaronmk
##### Exceptions
20
21 2804 aaronmk
def get_cur_query(cur, input_query=None):
22 2168 aaronmk
    raw_query = None
23
    if hasattr(cur, 'query'): raw_query = cur.query
24
    elif hasattr(cur, '_last_executed'): raw_query = cur._last_executed
25 2170 aaronmk
26
    if raw_query != None: return raw_query
27 2804 aaronmk
    else: return '[input] '+strings.ustr(input_query)
28 14 aaronmk
29 2170 aaronmk
def _add_cursor_info(e, *args, **kw_args):
30
    '''For params, see get_cur_query()'''
31 2771 aaronmk
    exc.add_msg(e, 'query: '+strings.ustr(get_cur_query(*args, **kw_args)))
32 135 aaronmk
33 300 aaronmk
class DbException(exc.ExceptionWithCause):
34 14 aaronmk
    def __init__(self, msg, cause=None, cur=None):
35 2145 aaronmk
        exc.ExceptionWithCause.__init__(self, msg, cause, cause_newline=True)
36 14 aaronmk
        if cur != None: _add_cursor_info(self, cur)
37
38 2143 aaronmk
class ExceptionWithName(DbException):
39
    def __init__(self, name, cause=None):
40 2484 aaronmk
        DbException.__init__(self, 'for name: '+strings.as_tt(str(name)), cause)
41 2143 aaronmk
        self.name = name
42 360 aaronmk
43 3109 aaronmk
class ExceptionWithValue(DbException):
44
    def __init__(self, value, cause=None):
45
        DbException.__init__(self, 'for value: '+strings.as_tt(repr(value)),
46
            cause)
47 2240 aaronmk
        self.value = value
48
49 2945 aaronmk
class ExceptionWithNameType(DbException):
50
    def __init__(self, type_, name, cause=None):
51
        DbException.__init__(self, 'for type: '+strings.as_tt(str(type_))
52
            +'; name: '+strings.as_tt(name), cause)
53
        self.type = type_
54
        self.name = name
55
56 2306 aaronmk
class ConstraintException(DbException):
57 3345 aaronmk
    def __init__(self, name, cond, cols, cause=None):
58
        msg = 'Violated '+strings.as_tt(name)+' constraint'
59
        if cond != None: msg += ' with condition '+cond
60
        if cols != []: msg += ' on columns: '+strings.as_tt(', '.join(cols))
61
        DbException.__init__(self, msg, cause)
62 2306 aaronmk
        self.name = name
63 3345 aaronmk
        self.cond = cond
64 468 aaronmk
        self.cols = cols
65 11 aaronmk
66 2523 aaronmk
class MissingCastException(DbException):
67 4139 aaronmk
    def __init__(self, type_, col=None, cause=None):
68
        msg = 'Missing cast to type '+strings.as_tt(type_)
69
        if col != None: msg += ' on column: '+strings.as_tt(col)
70
        DbException.__init__(self, msg, cause)
71 2523 aaronmk
        self.type = type_
72
        self.col = col
73
74 2143 aaronmk
class NameException(DbException): pass
75
76 2306 aaronmk
class DuplicateKeyException(ConstraintException): pass
77 13 aaronmk
78 2306 aaronmk
class NullValueException(ConstraintException): pass
79 13 aaronmk
80 3346 aaronmk
class CheckException(ConstraintException): pass
81
82 3109 aaronmk
class InvalidValueException(ExceptionWithValue): pass
83 2239 aaronmk
84 2945 aaronmk
class DuplicateException(ExceptionWithNameType): pass
85 2143 aaronmk
86 3419 aaronmk
class DoesNotExistException(ExceptionWithNameType): pass
87
88 89 aaronmk
class EmptyRowException(DbException): pass
89
90 865 aaronmk
##### Warnings
91
92
class DbWarning(UserWarning): pass
93
94 1930 aaronmk
##### Result retrieval
95
96
def col_names(cur): return (col[0] for col in cur.description)
97
98
def rows(cur): return iter(lambda: cur.fetchone(), None)
99
100
def consume_rows(cur):
101
    '''Used to fetch all rows so result will be cached'''
102
    iters.consume_iter(rows(cur))
103
104
def next_row(cur): return rows(cur).next()
105
106
def row(cur):
107
    row_ = next_row(cur)
108
    consume_rows(cur)
109
    return row_
110
111
def next_value(cur): return next_row(cur)[0]
112
113
def value(cur): return row(cur)[0]
114
115
def values(cur): return iters.func_iter(lambda: next_value(cur))
116
117
def value_or_none(cur):
118
    try: return value(cur)
119
    except StopIteration: return None
120
121 2762 aaronmk
##### Escaping
122 2101 aaronmk
123 2573 aaronmk
def esc_name_by_module(module, name):
124
    if module == 'psycopg2' or module == None: quote = '"'
125 2101 aaronmk
    elif module == 'MySQLdb': quote = '`'
126
    else: raise NotImplementedError("Can't escape name for "+module+' database')
127 2500 aaronmk
    return sql_gen.esc_name(name, quote)
128 2101 aaronmk
129
def esc_name_by_engine(engine, name, **kw_args):
130
    return esc_name_by_module(db_engines[engine][0], name, **kw_args)
131
132
def esc_name(db, name, **kw_args):
133
    return esc_name_by_module(util.root_module(db.db), name, **kw_args)
134
135
def qual_name(db, schema, table):
136
    def esc_name_(name): return esc_name(db, name)
137
    table = esc_name_(table)
138
    if schema != None: return esc_name_(schema)+'.'+table
139
    else: return table
140
141 1869 aaronmk
##### Database connections
142 1849 aaronmk
143 2097 aaronmk
db_config_names = ['engine', 'host', 'user', 'password', 'database', 'schemas']
144 1926 aaronmk
145 1869 aaronmk
db_engines = {
146
    'MySQL': ('MySQLdb', {'password': 'passwd', 'database': 'db'}),
147
    'PostgreSQL': ('psycopg2', {}),
148
}
149
150
DatabaseErrors_set = set([DbException])
151
DatabaseErrors = tuple(DatabaseErrors_set)
152
153
def _add_module(module):
154
    DatabaseErrors_set.add(module.DatabaseError)
155
    global DatabaseErrors
156
    DatabaseErrors = tuple(DatabaseErrors_set)
157
158
def db_config_str(db_config):
159
    return db_config['engine']+' database '+db_config['database']
160
161 2448 aaronmk
log_debug_none = lambda msg, level=2: None
162 1901 aaronmk
163 1849 aaronmk
class DbConn:
164 2923 aaronmk
    def __init__(self, db_config, autocommit=True, caching=True,
165 3183 aaronmk
        log_debug=log_debug_none, debug_temp=False, src=None):
166 2915 aaronmk
        '''
167
        @param debug_temp Whether temporary objects should instead be permanent.
168
            This assists in debugging the internal objects used by the program.
169 3183 aaronmk
        @param src In autocommit mode, will be included in a comment in every
170
            query, to help identify the data source in pg_stat_activity.
171 2915 aaronmk
        '''
172 1869 aaronmk
        self.db_config = db_config
173 2190 aaronmk
        self.autocommit = autocommit
174
        self.caching = caching
175 1901 aaronmk
        self.log_debug = log_debug
176 2193 aaronmk
        self.debug = log_debug != log_debug_none
177 2915 aaronmk
        self.debug_temp = debug_temp
178 3183 aaronmk
        self.src = src
179 3074 aaronmk
        self.autoanalyze = False
180 3269 aaronmk
        self.autoexplain = False
181
        self.profile_row_ct = None
182 1869 aaronmk
183 3124 aaronmk
        self._savepoint = 0
184 3120 aaronmk
        self._reset()
185 1869 aaronmk
186
    def __getattr__(self, name):
187
        if name == '__dict__': raise Exception('getting __dict__')
188
        if name == 'db': return self._db()
189
        else: raise AttributeError()
190
191
    def __getstate__(self):
192
        state = copy.copy(self.__dict__) # shallow copy
193 1915 aaronmk
        state['log_debug'] = None # don't pickle the debug callback
194 1869 aaronmk
        state['_DbConn__db'] = None # don't pickle the connection
195
        return state
196
197 3118 aaronmk
    def clear_cache(self): self.query_results = {}
198
199 3120 aaronmk
    def _reset(self):
200 3118 aaronmk
        self.clear_cache()
201 3124 aaronmk
        assert self._savepoint == 0
202 3118 aaronmk
        self._notices_seen = set()
203
        self.__db = None
204
205 2165 aaronmk
    def connected(self): return self.__db != None
206
207 3116 aaronmk
    def close(self):
208 3119 aaronmk
        if not self.connected(): return
209
210 3135 aaronmk
        # Record that the automatic transaction is now closed
211 3136 aaronmk
        self._savepoint -= 1
212 3135 aaronmk
213 3119 aaronmk
        self.db.close()
214 3120 aaronmk
        self._reset()
215 3116 aaronmk
216 3125 aaronmk
    def reconnect(self):
217
        # Do not do this in test mode as it would roll back everything
218
        if self.autocommit: self.close()
219
        # Connection will be reopened automatically on first query
220
221 1869 aaronmk
    def _db(self):
222
        if self.__db == None:
223
            # Process db_config
224
            db_config = self.db_config.copy() # don't modify input!
225 2097 aaronmk
            schemas = db_config.pop('schemas', None)
226 1869 aaronmk
            module_name, mappings = db_engines[db_config.pop('engine')]
227
            module = __import__(module_name)
228
            _add_module(module)
229
            for orig, new in mappings.iteritems():
230
                try: util.rename_key(db_config, orig, new)
231
                except KeyError: pass
232
233
            # Connect
234
            self.__db = module.connect(**db_config)
235
236 3161 aaronmk
            # Record that a transaction is already open
237
            self._savepoint += 1
238
239 1869 aaronmk
            # Configure connection
240 2906 aaronmk
            if hasattr(self.db, 'set_isolation_level'):
241
                import psycopg2.extensions
242
                self.db.set_isolation_level(
243
                    psycopg2.extensions.ISOLATION_LEVEL_READ_COMMITTED)
244 2101 aaronmk
            if schemas != None:
245 2893 aaronmk
                search_path = [self.esc_name(s) for s in schemas.split(',')]
246
                search_path.append(value(run_query(self, 'SHOW search_path',
247
                    log_level=4)))
248
                run_query(self, 'SET search_path TO '+(','.join(search_path)),
249
                    log_level=3)
250 1869 aaronmk
251
        return self.__db
252 1889 aaronmk
253 1891 aaronmk
    class DbCursor(Proxy):
254 1927 aaronmk
        def __init__(self, outer):
255 1891 aaronmk
            Proxy.__init__(self, outer.db.cursor())
256 2191 aaronmk
            self.outer = outer
257 1927 aaronmk
            self.query_results = outer.query_results
258 1894 aaronmk
            self.query_lookup = None
259 1891 aaronmk
            self.result = []
260 1889 aaronmk
261 2802 aaronmk
        def execute(self, query):
262 2764 aaronmk
            self._is_insert = query.startswith('INSERT')
263 2797 aaronmk
            self.query_lookup = query
264 2148 aaronmk
            try:
265 3162 aaronmk
                try: cur = self.inner.execute(query)
266 2802 aaronmk
                finally: self.query = get_cur_query(self.inner, query)
267 1904 aaronmk
            except Exception, e:
268
                self.result = e # cache the exception as the result
269
                self._cache_result()
270
                raise
271 3004 aaronmk
272
            # Always cache certain queries
273 3183 aaronmk
            query = sql_gen.lstrip(query)
274 3004 aaronmk
            if query.startswith('CREATE') or query.startswith('ALTER'):
275 3007 aaronmk
                # structural changes
276 3040 aaronmk
                # Rest of query must be unique in the face of name collisions,
277
                # so don't cache ADD COLUMN unless it has distinguishing comment
278
                if query.find('ADD COLUMN') < 0 or query.endswith('*/'):
279 3007 aaronmk
                    self._cache_result()
280 3004 aaronmk
            elif self.rowcount == 0 and query.startswith('SELECT'): # empty
281 2800 aaronmk
                consume_rows(self) # fetch all rows so result will be cached
282 3004 aaronmk
283 2762 aaronmk
            return cur
284 1894 aaronmk
285 1891 aaronmk
        def fetchone(self):
286
            row = self.inner.fetchone()
287 1899 aaronmk
            if row != None: self.result.append(row)
288
            # otherwise, fetched all rows
289 1904 aaronmk
            else: self._cache_result()
290
            return row
291
292
        def _cache_result(self):
293 2948 aaronmk
            # For inserts that return a result set, don't cache result set since
294
            # inserts are not idempotent. Other non-SELECT queries don't have
295
            # their result set read, so only exceptions will be cached (an
296
            # invalid query will always be invalid).
297 1930 aaronmk
            if self.query_results != None and (not self._is_insert
298 1906 aaronmk
                or isinstance(self.result, Exception)):
299
300 1894 aaronmk
                assert self.query_lookup != None
301 1916 aaronmk
                self.query_results[self.query_lookup] = self.CacheCursor(
302
                    util.dict_subset(dicts.AttrsDictView(self),
303
                    ['query', 'result', 'rowcount', 'description']))
304 1906 aaronmk
305 1916 aaronmk
        class CacheCursor:
306
            def __init__(self, cached_result): self.__dict__ = cached_result
307
308 1927 aaronmk
            def execute(self, *args, **kw_args):
309 1916 aaronmk
                if isinstance(self.result, Exception): raise self.result
310
                # otherwise, result is a rows list
311
                self.iter = iter(self.result)
312
313
            def fetchone(self):
314
                try: return self.iter.next()
315
                except StopIteration: return None
316 1891 aaronmk
317 2212 aaronmk
    def esc_value(self, value):
318 2663 aaronmk
        try: str_ = self.mogrify('%s', [value])
319
        except NotImplementedError, e:
320
            module = util.root_module(self.db)
321
            if module == 'MySQLdb':
322
                import _mysql
323
                str_ = _mysql.escape_string(value)
324
            else: raise e
325 2374 aaronmk
        return strings.to_unicode(str_)
326 2212 aaronmk
327 2347 aaronmk
    def esc_name(self, name): return esc_name(self, name) # calls global func
328
329 2814 aaronmk
    def std_code(self, str_):
330
        '''Standardizes SQL code.
331
        * Ensures that string literals are prefixed by `E`
332
        '''
333
        if str_.startswith("'"): str_ = 'E'+str_
334
        return str_
335
336 2665 aaronmk
    def can_mogrify(self):
337 2663 aaronmk
        module = util.root_module(self.db)
338 2665 aaronmk
        return module == 'psycopg2'
339 2663 aaronmk
340 2665 aaronmk
    def mogrify(self, query, params=None):
341
        if self.can_mogrify(): return self.db.cursor().mogrify(query, params)
342
        else: raise NotImplementedError("Can't mogrify query")
343
344 2671 aaronmk
    def print_notices(self):
345 2725 aaronmk
        if hasattr(self.db, 'notices'):
346
            for msg in self.db.notices:
347
                if msg not in self._notices_seen:
348
                    self._notices_seen.add(msg)
349
                    self.log_debug(msg, level=2)
350 2671 aaronmk
351 2793 aaronmk
    def run_query(self, query, cacheable=False, log_level=2,
352 2464 aaronmk
        debug_msg_ref=None):
353 2445 aaronmk
        '''
354 2464 aaronmk
        @param log_ignore_excs The log_level will be increased by 2 if the query
355
            throws one of these exceptions.
356 2664 aaronmk
        @param debug_msg_ref If specified, the log message will be returned in
357
            this instead of being output. This allows you to filter log messages
358
            depending on the result of the query.
359 2445 aaronmk
        '''
360 2167 aaronmk
        assert query != None
361
362 3183 aaronmk
        if self.autocommit and self.src != None:
363 3206 aaronmk
            query = sql_gen.esc_comment(self.src)+'\t'+query
364 3183 aaronmk
365 2047 aaronmk
        if not self.caching: cacheable = False
366 1903 aaronmk
        used_cache = False
367 2664 aaronmk
368 3242 aaronmk
        if self.debug:
369
            profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
370 1903 aaronmk
        try:
371 1927 aaronmk
            # Get cursor
372
            if cacheable:
373 3238 aaronmk
                try: cur = self.query_results[query]
374 1927 aaronmk
                except KeyError: cur = self.DbCursor(self)
375 3238 aaronmk
                else: used_cache = True
376 1927 aaronmk
            else: cur = self.db.cursor()
377
378
            # Run query
379 3238 aaronmk
            try: cur.execute(query)
380 3162 aaronmk
            except Exception, e:
381
                _add_cursor_info(e, self, query)
382
                raise
383 3238 aaronmk
            else: self.do_autocommit()
384 1903 aaronmk
        finally:
385 3242 aaronmk
            if self.debug:
386 3244 aaronmk
                profiler.stop(self.profile_row_ct)
387 3242 aaronmk
388
                ## Log or return query
389
390 3239 aaronmk
                query = str(get_cur_query(cur, query))
391 3281 aaronmk
                # Put the src comment on a separate line in the log file
392
                query = query.replace('\t', '\n', 1)
393 3239 aaronmk
394 3240 aaronmk
                msg = 'DB query: '
395 3239 aaronmk
396 3240 aaronmk
                if used_cache: msg += 'cache hit'
397
                elif cacheable: msg += 'cache miss'
398
                else: msg += 'non-cacheable'
399 3239 aaronmk
400 3241 aaronmk
                msg += ':\n'+profiler.msg()+'\n'+strings.as_code(query, 'SQL')
401 3240 aaronmk
402 3237 aaronmk
                if debug_msg_ref != None: debug_msg_ref[0] = msg
403
                else: self.log_debug(msg, log_level)
404 3245 aaronmk
405
                self.print_notices()
406 1903 aaronmk
407
        return cur
408 1914 aaronmk
409 2797 aaronmk
    def is_cached(self, query): return query in self.query_results
410 2139 aaronmk
411 2907 aaronmk
    def with_autocommit(self, func):
412 2801 aaronmk
        import psycopg2.extensions
413
414
        prev_isolation_level = self.db.isolation_level
415 2907 aaronmk
        self.db.set_isolation_level(
416
            psycopg2.extensions.ISOLATION_LEVEL_AUTOCOMMIT)
417 2683 aaronmk
        try: return func()
418 2801 aaronmk
        finally: self.db.set_isolation_level(prev_isolation_level)
419 2683 aaronmk
420 2139 aaronmk
    def with_savepoint(self, func):
421 3137 aaronmk
        top = self._savepoint == 0
422 2171 aaronmk
        savepoint = 'level_'+str(self._savepoint)
423 3137 aaronmk
424 3272 aaronmk
        if self.debug:
425 3273 aaronmk
            self.log_debug('Begin transaction', level=4)
426 3272 aaronmk
            profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
427
428 3160 aaronmk
        # Must happen before running queries so they don't get autocommitted
429
        self._savepoint += 1
430
431 3137 aaronmk
        if top: query = 'START TRANSACTION ISOLATION LEVEL READ COMMITTED'
432
        else: query = 'SAVEPOINT '+savepoint
433
        self.run_query(query, log_level=4)
434
        try:
435
            return func()
436
            if top: self.run_query('COMMIT', log_level=4)
437 2139 aaronmk
        except:
438 3137 aaronmk
            if top: query = 'ROLLBACK'
439
            else: query = 'ROLLBACK TO SAVEPOINT '+savepoint
440
            self.run_query(query, log_level=4)
441
442 2139 aaronmk
            raise
443 2930 aaronmk
        finally:
444
            # Always release savepoint, because after ROLLBACK TO SAVEPOINT,
445
            # "The savepoint remains valid and can be rolled back to again"
446
            # (http://www.postgresql.org/docs/8.3/static/sql-rollback-to.html).
447 3137 aaronmk
            if not top:
448
                self.run_query('RELEASE SAVEPOINT '+savepoint, log_level=4)
449 2930 aaronmk
450
            self._savepoint -= 1
451
            assert self._savepoint >= 0
452
453 3272 aaronmk
            if self.debug:
454
                profiler.stop(self.profile_row_ct)
455 3273 aaronmk
                self.log_debug('End transaction\n'+profiler.msg(), level=4)
456 3272 aaronmk
457 2930 aaronmk
            self.do_autocommit() # OK to do this after ROLLBACK TO SAVEPOINT
458 2191 aaronmk
459
    def do_autocommit(self):
460
        '''Autocommits if outside savepoint'''
461 3135 aaronmk
        assert self._savepoint >= 1
462
        if self.autocommit and self._savepoint == 1:
463 2924 aaronmk
            self.log_debug('Autocommitting', level=4)
464 2191 aaronmk
            self.db.commit()
465 2643 aaronmk
466 3155 aaronmk
    def col_info(self, col, cacheable=True):
467 2643 aaronmk
        table = sql_gen.Table('columns', 'information_schema')
468 3063 aaronmk
        type_ = sql_gen.Coalesce(sql_gen.Nullif(sql_gen.Col('data_type'),
469
            'USER-DEFINED'), sql_gen.Col('udt_name'))
470 3078 aaronmk
        cols = [type_, 'column_default',
471
            sql_gen.Cast('boolean', sql_gen.Col('is_nullable'))]
472 2643 aaronmk
473 3750 aaronmk
        conds = [('table_name', col.table.name),
474
            ('column_name', strings.ustr(col.name))]
475 2643 aaronmk
        schema = col.table.schema
476
        if schema != None: conds.append(('table_schema', schema))
477
478 3638 aaronmk
        cur = select(self, table, cols, conds, order_by='table_schema', limit=1,
479
            cacheable=cacheable, log_level=4) # TODO: order by search_path order
480
        try: type_, default, nullable = row(cur)
481 4114 aaronmk
        except StopIteration: raise sql_gen.NoUnderlyingTableException(col)
482 2819 aaronmk
        default = sql_gen.as_Code(default, self)
483
484
        return sql_gen.TypedCol(col.name, type_, default, nullable)
485 2917 aaronmk
486
    def TempFunction(self, name):
487
        if self.debug_temp: schema = None
488
        else: schema = 'pg_temp'
489
        return sql_gen.Function(name, schema)
490 1849 aaronmk
491 1869 aaronmk
connect = DbConn
492
493 832 aaronmk
##### Recoverable querying
494 15 aaronmk
495 2139 aaronmk
def with_savepoint(db, func): return db.with_savepoint(func)
496 11 aaronmk
497 2791 aaronmk
def run_query(db, query, recover=None, cacheable=False, log_level=2,
498
    log_ignore_excs=None, **kw_args):
499 2794 aaronmk
    '''For params, see DbConn.run_query()'''
500 830 aaronmk
    if recover == None: recover = False
501 2464 aaronmk
    if log_ignore_excs == None: log_ignore_excs = ()
502
    log_ignore_excs = tuple(log_ignore_excs)
503 3236 aaronmk
    debug_msg_ref = [None]
504 830 aaronmk
505 3267 aaronmk
    query = with_explain_comment(db, query)
506 3258 aaronmk
507 2148 aaronmk
    try:
508 2464 aaronmk
        try:
509 2794 aaronmk
            def run(): return db.run_query(query, cacheable, log_level,
510 2793 aaronmk
                debug_msg_ref, **kw_args)
511 2796 aaronmk
            if recover and not db.is_cached(query):
512 2464 aaronmk
                return with_savepoint(db, run)
513
            else: return run() # don't need savepoint if cached
514
        except Exception, e:
515 3095 aaronmk
            msg = strings.ustr(e.args[0])
516 4103 aaronmk
            msg = re.sub(r'^(?:PL/Python: )?ValueError: ', r'', msg)
517 2464 aaronmk
518 3095 aaronmk
            match = re.match(r'^duplicate key value violates unique constraint '
519 3338 aaronmk
                r'"(.+?)"', msg)
520 2464 aaronmk
            if match:
521 3338 aaronmk
                constraint, = match.groups()
522 3025 aaronmk
                cols = []
523
                if recover: # need auto-rollback to run index_cols()
524 3319 aaronmk
                    try: cols = index_cols(db, constraint)
525 3025 aaronmk
                    except NotImplementedError: pass
526 3345 aaronmk
                raise DuplicateKeyException(constraint, None, cols, e)
527 2464 aaronmk
528 3095 aaronmk
            match = re.match(r'^null value in column "(.+?)" violates not-null'
529 2464 aaronmk
                r' constraint', msg)
530 3345 aaronmk
            if match:
531
                col, = match.groups()
532
                raise NullValueException('NOT NULL', None, [col], e)
533 2464 aaronmk
534 3346 aaronmk
            match = re.match(r'^new row for relation "(.+?)" violates check '
535
                r'constraint "(.+?)"', msg)
536
            if match:
537
                table, constraint = match.groups()
538 3347 aaronmk
                constraint = sql_gen.Col(constraint, table)
539 3349 aaronmk
                cond = None
540
                if recover: # need auto-rollback to run constraint_cond()
541
                    try: cond = constraint_cond(db, constraint)
542
                    except NotImplementedError: pass
543
                raise CheckException(constraint.to_str(db), cond, [], e)
544 3346 aaronmk
545 3095 aaronmk
            match = re.match(r'^(?:invalid input (?:syntax|value)\b.*?'
546 3635 aaronmk
                r'|.+? out of range): "(.+?)"', msg)
547 2464 aaronmk
            if match:
548 3109 aaronmk
                value, = match.groups()
549
                raise InvalidValueException(strings.to_unicode(value), e)
550 2464 aaronmk
551 3095 aaronmk
            match = re.match(r'^column "(.+?)" is of type (.+?) but expression '
552 2523 aaronmk
                r'is of type', msg)
553
            if match:
554
                col, type_ = match.groups()
555
                raise MissingCastException(type_, col, e)
556
557 4141 aaronmk
            match = re.match(r'^could not determine polymorphic type because '
558
                r'input has type "unknown"', msg)
559
            if match: raise MissingCastException('text', None, e)
560
561 4485 aaronmk
            match = re.match(r'^.+? types .+? and .+? cannot be matched', msg)
562
            if match: raise MissingCastException('text', None, e)
563
564 3419 aaronmk
            typed_name_re = r'^(\S+) "(.+?)"'
565
566
            match = re.match(typed_name_re+r'.*? already exists', msg)
567 2945 aaronmk
            if match:
568
                type_, name = match.groups()
569
                raise DuplicateException(type_, name, e)
570 2464 aaronmk
571 4145 aaronmk
            match = re.match(r'more than one (\S+) named ""(.+?)""', msg)
572
            if match:
573
                type_, name = match.groups()
574
                raise DuplicateException(type_, name, e)
575
576 3419 aaronmk
            match = re.match(typed_name_re+r' does not exist', msg)
577
            if match:
578
                type_, name = match.groups()
579
                raise DoesNotExistException(type_, name, e)
580
581 2464 aaronmk
            raise # no specific exception raised
582
    except log_ignore_excs:
583
        log_level += 2
584
        raise
585
    finally:
586 3236 aaronmk
        if debug_msg_ref[0] != None: db.log_debug(debug_msg_ref[0], log_level)
587 830 aaronmk
588 832 aaronmk
##### Basic queries
589
590 3256 aaronmk
def is_explainable(query):
591
    # See <http://www.postgresql.org/docs/8.3/static/sql-explain.html#AEN57749>
592 3257 aaronmk
    return re.match(r'^(?:SELECT|INSERT|UPDATE|DELETE|VALUES|EXECUTE|DECLARE)\b'
593
        , query)
594 3256 aaronmk
595 3263 aaronmk
def explain(db, query, **kw_args):
596
    '''
597
    For params, see run_query().
598
    '''
599 3267 aaronmk
    kw_args.setdefault('log_level', 4)
600 3263 aaronmk
601 3750 aaronmk
    return strings.ustr(strings.join_lines(values(run_query(db,
602
        'EXPLAIN '+query, recover=True, cacheable=True, **kw_args))))
603 3256 aaronmk
        # not a higher log_level because it's useful to see what query is being
604
        # run before it's executed, which EXPLAIN effectively provides
605
606 3265 aaronmk
def has_comment(query): return query.endswith('*/')
607
608
def with_explain_comment(db, query, **kw_args):
609 3269 aaronmk
    if db.autoexplain and not has_comment(query) and is_explainable(query):
610 3265 aaronmk
        query += '\n'+sql_gen.esc_comment(' EXPLAIN:\n'
611
            +explain(db, query, **kw_args))
612
    return query
613
614 2153 aaronmk
def next_version(name):
615 2163 aaronmk
    version = 1 # first existing name was version 0
616 2586 aaronmk
    match = re.match(r'^(.*)#(\d+)$', name)
617 2153 aaronmk
    if match:
618 2586 aaronmk
        name, version = match.groups()
619
        version = int(version)+1
620 2932 aaronmk
    return sql_gen.concat(name, '#'+str(version))
621 2153 aaronmk
622 2899 aaronmk
def lock_table(db, table, mode):
623
    table = sql_gen.as_Table(table)
624
    run_query(db, 'LOCK TABLE '+table.to_str(db)+' IN '+mode+' MODE')
625
626 3303 aaronmk
def run_query_into(db, query, into=None, add_pkey_=False, **kw_args):
627 2085 aaronmk
    '''Outputs a query to a temp table.
628
    For params, see run_query().
629
    '''
630 2789 aaronmk
    if into == None: return run_query(db, query, **kw_args)
631 2790 aaronmk
632
    assert isinstance(into, sql_gen.Table)
633
634 2992 aaronmk
    into.is_temp = True
635 3008 aaronmk
    # "temporary tables cannot specify a schema name", so remove schema
636
    into.schema = None
637 2992 aaronmk
638 2790 aaronmk
    kw_args['recover'] = True
639 2945 aaronmk
    kw_args.setdefault('log_ignore_excs', (DuplicateException,))
640 2790 aaronmk
641 2916 aaronmk
    temp = not db.debug_temp # tables are permanent in debug_temp mode
642 2790 aaronmk
643
    # Create table
644
    while True:
645
        create_query = 'CREATE'
646
        if temp: create_query += ' TEMP'
647
        create_query += ' TABLE '+into.to_str(db)+' AS\n'+query
648 2385 aaronmk
649 2790 aaronmk
        try:
650
            cur = run_query(db, create_query, **kw_args)
651
                # CREATE TABLE AS sets rowcount to # rows in query
652
            break
653 2945 aaronmk
        except DuplicateException, e:
654 2790 aaronmk
            into.name = next_version(into.name)
655
            # try again with next version of name
656
657 3303 aaronmk
    if add_pkey_: add_pkey(db, into)
658 3075 aaronmk
659
    # According to the PostgreSQL doc, "The autovacuum daemon cannot access and
660
    # therefore cannot vacuum or analyze temporary tables. [...] if a temporary
661
    # table is going to be used in complex queries, it is wise to run ANALYZE on
662
    # the temporary table after it is populated."
663
    # (http://www.postgresql.org/docs/9.1/static/sql-createtable.html)
664
    # If into is not a temp table, ANALYZE is useful but not required.
665 3073 aaronmk
    analyze(db, into)
666 2790 aaronmk
667
    return cur
668 2085 aaronmk
669 2120 aaronmk
order_by_pkey = object() # tells mk_select() to order by the pkey
670
671 2199 aaronmk
distinct_on_all = object() # tells mk_select() to SELECT DISTINCT ON all columns
672
673 3420 aaronmk
def mk_select(db, tables=None, fields=None, conds=None, distinct_on=[],
674 3494 aaronmk
    limit=None, start=None, order_by=order_by_pkey, default_table=None,
675
    explain=True):
676 1981 aaronmk
    '''
677 2121 aaronmk
    @param tables The single table to select from, or a list of tables to join
678 2280 aaronmk
        together, with tables after the first being sql_gen.Join objects
679 1981 aaronmk
    @param fields Use None to select all fields in the table
680 2377 aaronmk
    @param conds WHERE conditions: [(compare_left_side, compare_right_side),...]
681 2379 aaronmk
        * container can be any iterable type
682 2399 aaronmk
        * compare_left_side: sql_gen.Code|str (for col name)
683
        * compare_right_side: sql_gen.ValueCond|literal value
684 2199 aaronmk
    @param distinct_on The columns to SELECT DISTINCT ON, or distinct_on_all to
685
        use all columns
686 2786 aaronmk
    @return query
687 1981 aaronmk
    '''
688 2315 aaronmk
    # Parse tables param
689 2964 aaronmk
    tables = lists.mk_seq(tables)
690 2141 aaronmk
    tables = list(tables) # don't modify input! (list() copies input)
691 2315 aaronmk
    table0 = sql_gen.as_Table(tables.pop(0)) # first table is separate
692 2121 aaronmk
693 2315 aaronmk
    # Parse other params
694 2376 aaronmk
    if conds == None: conds = []
695 2650 aaronmk
    elif dicts.is_dict(conds): conds = conds.items()
696 2379 aaronmk
    conds = list(conds) # don't modify input! (list() copies input)
697 3129 aaronmk
    assert limit == None or isinstance(limit, (int, long))
698
    assert start == None or isinstance(start, (int, long))
699 2315 aaronmk
    if order_by is order_by_pkey:
700 3421 aaronmk
        if table0 == None or distinct_on != []: order_by = None
701 2315 aaronmk
        else: order_by = pkey(db, table0, recover=True)
702 865 aaronmk
703 2315 aaronmk
    query = 'SELECT'
704 2056 aaronmk
705 2315 aaronmk
    def parse_col(col): return sql_gen.as_Col(col, default_table).to_str(db)
706 2056 aaronmk
707 2200 aaronmk
    # DISTINCT ON columns
708 2233 aaronmk
    if distinct_on != []:
709 2467 aaronmk
        query += '\nDISTINCT'
710 2254 aaronmk
        if distinct_on is not distinct_on_all:
711 2200 aaronmk
            query += ' ON ('+(', '.join(map(parse_col, distinct_on)))+')'
712
713
    # Columns
714 3185 aaronmk
    if query.find('\n') >= 0: whitespace = '\n'
715
    else: whitespace = ' '
716
    if fields == None: query += whitespace+'*'
717 2765 aaronmk
    else:
718
        assert fields != []
719 3185 aaronmk
        if len(fields) > 1: whitespace = '\n'
720
        query += whitespace+('\n, '.join(map(parse_col, fields)))
721 2200 aaronmk
722
    # Main table
723 3185 aaronmk
    if query.find('\n') >= 0 or len(tables) > 0: whitespace = '\n'
724
    else: whitespace = ' '
725 3420 aaronmk
    if table0 != None: query += whitespace+'FROM '+table0.to_str(db)
726 865 aaronmk
727 2122 aaronmk
    # Add joins
728 2271 aaronmk
    left_table = table0
729 2263 aaronmk
    for join_ in tables:
730
        table = join_.table
731 2238 aaronmk
732 2343 aaronmk
        # Parse special values
733
        if join_.type_ is sql_gen.filter_out: # filter no match
734 2376 aaronmk
            conds.append((sql_gen.Col(table_not_null_col(db, table), table),
735 2853 aaronmk
                sql_gen.CompareCond(None, '~=')))
736 2343 aaronmk
737 2467 aaronmk
        query += '\n'+join_.to_str(db, left_table)
738 2122 aaronmk
739
        left_table = table
740
741 865 aaronmk
    missing = True
742 2376 aaronmk
    if conds != []:
743 2576 aaronmk
        if len(conds) == 1: whitespace = ' '
744
        else: whitespace = '\n'
745 2578 aaronmk
        query += '\n'+sql_gen.combine_conds([sql_gen.ColValueCond(l, r)
746
            .to_str(db) for l, r in conds], 'WHERE')
747 2227 aaronmk
    if order_by != None:
748 2467 aaronmk
        query += '\nORDER BY '+sql_gen.as_Col(order_by, table0).to_str(db)
749 3297 aaronmk
    if limit != None: query += '\nLIMIT '+str(limit)
750 865 aaronmk
    if start != None:
751 2467 aaronmk
        if start != 0: query += '\nOFFSET '+str(start)
752 865 aaronmk
753 3494 aaronmk
    if explain: query = with_explain_comment(db, query)
754 3266 aaronmk
755 2786 aaronmk
    return query
756 11 aaronmk
757 2054 aaronmk
def select(db, *args, **kw_args):
758
    '''For params, see mk_select() and run_query()'''
759
    recover = kw_args.pop('recover', None)
760
    cacheable = kw_args.pop('cacheable', True)
761 2442 aaronmk
    log_level = kw_args.pop('log_level', 2)
762 2054 aaronmk
763 2791 aaronmk
    return run_query(db, mk_select(db, *args, **kw_args), recover, cacheable,
764
        log_level=log_level)
765 2054 aaronmk
766 2788 aaronmk
def mk_insert_select(db, table, cols=None, select_query=None, returning=None,
767 3181 aaronmk
    embeddable=False, ignore=False, src=None):
768 1960 aaronmk
    '''
769
    @param returning str|None An inserted column (such as pkey) to return
770 2070 aaronmk
    @param embeddable Whether the query should be embeddable as a nested SELECT.
771 2073 aaronmk
        Warning: If you set this and cacheable=True when the query is run, the
772
        query will be fully cached, not just if it raises an exception.
773 3009 aaronmk
    @param ignore Whether to ignore duplicate keys.
774 3181 aaronmk
    @param src Will be included in the name of any created function, to help
775
        identify the data source in pg_stat_activity.
776 1960 aaronmk
    '''
777 2754 aaronmk
    table = sql_gen.remove_table_rename(sql_gen.as_Table(table))
778 2318 aaronmk
    if cols == []: cols = None # no cols (all defaults) = unknown col names
779 3010 aaronmk
    if cols != None: cols = [sql_gen.to_name_only_col(c, table) for c in cols]
780 2063 aaronmk
    if select_query == None: select_query = 'DEFAULT VALUES'
781 2327 aaronmk
    if returning != None: returning = sql_gen.as_Col(returning, table)
782 2063 aaronmk
783 2497 aaronmk
    first_line = 'INSERT INTO '+table.to_str(db)
784 2063 aaronmk
785 3009 aaronmk
    def mk_insert(select_query):
786
        query = first_line
787 3014 aaronmk
        if cols != None:
788
            query += '\n('+(', '.join((c.to_str(db) for c in cols)))+')'
789 3009 aaronmk
        query += '\n'+select_query
790
791
        if returning != None:
792
            returning_name_col = sql_gen.to_name_only_col(returning)
793
            query += '\nRETURNING '+returning_name_col.to_str(db)
794
795
        return query
796 2063 aaronmk
797 3489 aaronmk
    return_type = sql_gen.CustomCode('unknown')
798
    if returning != None: return_type = sql_gen.ColType(returning)
799 3017 aaronmk
800 3009 aaronmk
    if ignore:
801 3017 aaronmk
        # Always return something to set the correct rowcount
802
        if returning == None: returning = sql_gen.NamedCol('NULL', None)
803
804 3009 aaronmk
        embeddable = True # must use function
805 3010 aaronmk
806 3450 aaronmk
        if cols == None: row = [sql_gen.Col(sql_gen.all_cols, 'row')]
807
        else: row = [sql_gen.Col(c.name, 'row') for c in cols]
808 3092 aaronmk
809 3484 aaronmk
        query = sql_gen.RowExcIgnore(sql_gen.RowType(table), select_query,
810 3497 aaronmk
            sql_gen.ReturnQuery(mk_insert(sql_gen.Values(row).to_str(db))),
811
            cols)
812 3009 aaronmk
    else: query = mk_insert(select_query)
813
814 2070 aaronmk
    if embeddable:
815
        # Create function
816 2513 aaronmk
        function_name = sql_gen.clean_name(first_line)
817 3181 aaronmk
        if src != None: function_name = src+': '+function_name
818 2189 aaronmk
        while True:
819
            try:
820 3451 aaronmk
                func = db.TempFunction(function_name)
821 3489 aaronmk
                def_ = sql_gen.FunctionDef(func, sql_gen.SetOf(return_type),
822
                    query)
823 2194 aaronmk
824 3443 aaronmk
                run_query(db, def_.to_str(db), recover=True, cacheable=True,
825 2945 aaronmk
                    log_ignore_excs=(DuplicateException,))
826 2189 aaronmk
                break # this version was successful
827 2945 aaronmk
            except DuplicateException, e:
828 2189 aaronmk
                function_name = next_version(function_name)
829
                # try again with next version of name
830 2070 aaronmk
831 2337 aaronmk
        # Return query that uses function
832 3009 aaronmk
        cols = None
833
        if returning != None: cols = [returning]
834 3451 aaronmk
        func_table = sql_gen.NamedTable('f', sql_gen.FunctionCall(func), cols)
835
            # AS clause requires function alias
836 3298 aaronmk
        return mk_select(db, func_table, order_by=None)
837 2070 aaronmk
838 2787 aaronmk
    return query
839 2066 aaronmk
840 3074 aaronmk
def insert_select(db, table, *args, **kw_args):
841 2085 aaronmk
    '''For params, see mk_insert_select() and run_query_into()
842 2386 aaronmk
    @param into sql_gen.Table with suggested name of temp table to put RETURNING
843
        values in
844 2072 aaronmk
    '''
845 3141 aaronmk
    returning = kw_args.get('returning', None)
846
    ignore = kw_args.get('ignore', False)
847
848 2386 aaronmk
    into = kw_args.pop('into', None)
849
    if into != None: kw_args['embeddable'] = True
850 2066 aaronmk
    recover = kw_args.pop('recover', None)
851 3141 aaronmk
    if ignore: recover = True
852 2066 aaronmk
    cacheable = kw_args.pop('cacheable', True)
853 2673 aaronmk
    log_level = kw_args.pop('log_level', 2)
854 2066 aaronmk
855 3141 aaronmk
    rowcount_only = ignore and returning == None # keep NULL rows on server
856
    if rowcount_only: into = sql_gen.Table('rowcount')
857
858 3074 aaronmk
    cur = run_query_into(db, mk_insert_select(db, table, *args, **kw_args),
859
        into, recover=recover, cacheable=cacheable, log_level=log_level)
860 3141 aaronmk
    if rowcount_only: empty_temp(db, into)
861 3074 aaronmk
    autoanalyze(db, table)
862
    return cur
863 2063 aaronmk
864 2738 aaronmk
default = sql_gen.default # tells insert() to use the default value for a column
865 2066 aaronmk
866 2063 aaronmk
def insert(db, table, row, *args, **kw_args):
867 2085 aaronmk
    '''For params, see insert_select()'''
868 1960 aaronmk
    if lists.is_seq(row): cols = None
869
    else:
870
        cols = row.keys()
871
        row = row.values()
872 2738 aaronmk
    row = list(row) # ensure that "== []" works
873 1960 aaronmk
874 2738 aaronmk
    if row == []: query = None
875
    else: query = sql_gen.Values(row).to_str(db)
876 1961 aaronmk
877 2788 aaronmk
    return insert_select(db, table, cols, query, *args, **kw_args)
878 11 aaronmk
879 3152 aaronmk
def mk_update(db, table, changes=None, cond=None, in_place=False,
880 3153 aaronmk
    cacheable_=True):
881 2402 aaronmk
    '''
882
    @param changes [(col, new_value),...]
883
        * container can be any iterable type
884
        * col: sql_gen.Code|str (for col name)
885
        * new_value: sql_gen.Code|literal value
886
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
887 3056 aaronmk
    @param in_place If set, locks the table and updates rows in place.
888
        This avoids creating dead rows in PostgreSQL.
889
        * cond must be None
890 3153 aaronmk
    @param cacheable_ Whether column structure information used to generate the
891 3152 aaronmk
        query can be cached
892 2402 aaronmk
    @return str query
893
    '''
894 3057 aaronmk
    table = sql_gen.as_Table(table)
895
    changes = [(sql_gen.to_name_only_col(c, table), sql_gen.as_Value(v))
896
        for c, v in changes]
897
898 3056 aaronmk
    if in_place:
899
        assert cond == None
900 3058 aaronmk
901 3065 aaronmk
        query = 'ALTER TABLE '+table.to_str(db)+'\n'
902
        query += ',\n'.join(('ALTER COLUMN '+c.to_str(db)+' TYPE '
903 3153 aaronmk
            +db.col_info(sql_gen.with_default_table(c, table), cacheable_).type
904 3065 aaronmk
            +'\nUSING '+v.to_str(db) for c, v in changes))
905 3058 aaronmk
    else:
906
        query = 'UPDATE '+table.to_str(db)+'\nSET\n'
907
        query += ',\n'.join((c.to_str(db)+' = '+v.to_str(db)
908
            for c, v in changes))
909
        if cond != None: query += '\nWHERE\n'+cond.to_str(db)
910 3056 aaronmk
911 3266 aaronmk
    query = with_explain_comment(db, query)
912
913 2402 aaronmk
    return query
914
915 3074 aaronmk
def update(db, table, *args, **kw_args):
916 2402 aaronmk
    '''For params, see mk_update() and run_query()'''
917
    recover = kw_args.pop('recover', None)
918 3043 aaronmk
    cacheable = kw_args.pop('cacheable', False)
919 3030 aaronmk
    log_level = kw_args.pop('log_level', 2)
920 2402 aaronmk
921 3074 aaronmk
    cur = run_query(db, mk_update(db, table, *args, **kw_args), recover,
922
        cacheable, log_level=log_level)
923
    autoanalyze(db, table)
924
    return cur
925 2402 aaronmk
926 3286 aaronmk
def mk_delete(db, table, cond=None):
927
    '''
928
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
929
    @return str query
930
    '''
931
    query = 'DELETE FROM '+table.to_str(db)
932
    if cond != None: query += '\nWHERE '+cond.to_str(db)
933
934
    query = with_explain_comment(db, query)
935
936
    return query
937
938
def delete(db, table, *args, **kw_args):
939
    '''For params, see mk_delete() and run_query()'''
940
    recover = kw_args.pop('recover', None)
941 3295 aaronmk
    cacheable = kw_args.pop('cacheable', True)
942 3286 aaronmk
    log_level = kw_args.pop('log_level', 2)
943
944
    cur = run_query(db, mk_delete(db, table, *args, **kw_args), recover,
945
        cacheable, log_level=log_level)
946
    autoanalyze(db, table)
947
    return cur
948
949 135 aaronmk
def last_insert_id(db):
950 1849 aaronmk
    module = util.root_module(db.db)
951 135 aaronmk
    if module == 'psycopg2': return value(run_query(db, 'SELECT lastval()'))
952
    elif module == 'MySQLdb': return db.insert_id()
953
    else: return None
954 13 aaronmk
955 3490 aaronmk
def define_func(db, def_):
956
    func = def_.function
957
    while True:
958
        try:
959
            run_query(db, def_.to_str(db), recover=True, cacheable=True,
960
                log_ignore_excs=(DuplicateException,))
961
            break # successful
962
        except DuplicateException:
963 3495 aaronmk
            func.name = next_version(func.name)
964 3490 aaronmk
            # try again with next version of name
965
966 2394 aaronmk
def mk_flatten_mapping(db, into, cols, preserve=[], as_items=False):
967 2383 aaronmk
    '''Creates a mapping from original column names (which may have collisions)
968 2415 aaronmk
    to names that will be distinct among the columns' tables.
969 2383 aaronmk
    This is meant to be used for several tables that are being joined together.
970 2415 aaronmk
    @param cols The columns to combine. Duplicates will be removed.
971
    @param into The table for the new columns.
972 2394 aaronmk
    @param preserve [sql_gen.Col...] Columns not to rename. Note that these
973
        columns will be included in the mapping even if they are not in cols.
974
        The tables of the provided Col objects will be changed to into, so make
975
        copies of them if you want to keep the original tables.
976
    @param as_items Whether to return a list of dict items instead of a dict
977 2383 aaronmk
    @return dict(orig_col=new_col, ...)
978
        * orig_col: sql_gen.Col(orig_col_name, orig_table)
979 2392 aaronmk
        * new_col: sql_gen.Col(orig_col_name, into)
980
        * All mappings use the into table so its name can easily be
981 2383 aaronmk
          changed for all columns at once
982
    '''
983 2415 aaronmk
    cols = lists.uniqify(cols)
984
985 2394 aaronmk
    items = []
986 2389 aaronmk
    for col in preserve:
987 2390 aaronmk
        orig_col = copy.copy(col)
988 2392 aaronmk
        col.table = into
989 2394 aaronmk
        items.append((orig_col, col))
990
    preserve = set(preserve)
991
    for col in cols:
992 2716 aaronmk
        if col not in preserve:
993 3750 aaronmk
            items.append((col, sql_gen.Col(strings.ustr(col), into, col.srcs)))
994 2394 aaronmk
995
    if not as_items: items = dict(items)
996
    return items
997 2383 aaronmk
998 2393 aaronmk
def flatten(db, into, joins, cols, limit=None, start=None, **kw_args):
999 2391 aaronmk
    '''For params, see mk_flatten_mapping()
1000
    @return See return value of mk_flatten_mapping()
1001
    '''
1002 2394 aaronmk
    items = mk_flatten_mapping(db, into, cols, as_items=True, **kw_args)
1003
    cols = [sql_gen.NamedCol(new.name, old) for old, new in items]
1004 3296 aaronmk
    run_query_into(db, mk_select(db, joins, cols, order_by=None, limit=limit,
1005 3305 aaronmk
        start=start), into=into, add_pkey_=True)
1006 3708 aaronmk
        # don't cache because the temp table will usually be truncated after use
1007 2394 aaronmk
    return dict(items)
1008 2391 aaronmk
1009 3079 aaronmk
##### Database structure introspection
1010 2414 aaronmk
1011 3321 aaronmk
#### Expressions
1012
1013 3353 aaronmk
bool_re = r'(?:true|false)'
1014
1015
def simplify_expr(expr):
1016
    expr = expr.replace('(NULL IS NULL)', 'true')
1017
    expr = expr.replace('(NULL IS NOT NULL)', 'false')
1018
    expr = re.sub(r' OR '+bool_re, r'', expr)
1019
    expr = re.sub(bool_re+r' OR ', r'', expr)
1020
    while True:
1021
        expr, n = re.subn(r'\((\([^()]*\))\)', r'\1', expr)
1022
        if n == 0: break
1023
    return expr
1024
1025 3321 aaronmk
name_re = r'(?:\w+|(?:"[^"]*")+)'
1026
1027
def parse_expr_col(str_):
1028
    match = re.match(r'^\('+name_re+r'\(('+name_re+r').*\)\)$', str_)
1029
    if match: str_ = match.group(1)
1030
    return sql_gen.unesc_name(str_)
1031
1032 3351 aaronmk
def map_expr(db, expr, mapping, in_cols_found=None):
1033
    '''Replaces output columns with input columns in an expression.
1034
    @param in_cols_found If set, will be filled in with the expr's (input) cols
1035
    '''
1036
    for out, in_ in mapping.iteritems():
1037
        orig_expr = expr
1038
        out = sql_gen.to_name_only_col(out)
1039
        in_str = sql_gen.to_name_only_col(sql_gen.remove_col_rename(in_)
1040
            ).to_str(db)
1041
1042
        # Replace out both with and without quotes
1043
        expr = expr.replace(out.to_str(db), in_str)
1044 4490 aaronmk
        expr = re.sub(r'(?<!\.)\b'+out.name+r'\b(?!\.)', in_str, expr)
1045 3351 aaronmk
1046
        if in_cols_found != None and expr != orig_expr: # replaced something
1047
            in_cols_found.append(in_)
1048 3353 aaronmk
1049
    return simplify_expr(expr)
1050 3351 aaronmk
1051 3079 aaronmk
#### Tables
1052
1053
def tables(db, schema_like='public', table_like='%', exact=False):
1054
    if exact: compare = '='
1055
    else: compare = 'LIKE'
1056
1057
    module = util.root_module(db.db)
1058
    if module == 'psycopg2':
1059
        conds = [('schemaname', sql_gen.CompareCond(schema_like, compare)),
1060
            ('tablename', sql_gen.CompareCond(table_like, compare))]
1061
        return values(select(db, 'pg_tables', ['tablename'], conds,
1062
            order_by='tablename', log_level=4))
1063
    elif module == 'MySQLdb':
1064
        return values(run_query(db, 'SHOW TABLES LIKE '+db.esc_value(table_like)
1065
            , cacheable=True, log_level=4))
1066
    else: raise NotImplementedError("Can't list tables for "+module+' database')
1067
1068
def table_exists(db, table):
1069
    table = sql_gen.as_Table(table)
1070
    return list(tables(db, table.schema, table.name, exact=True)) != []
1071
1072 2426 aaronmk
def table_row_count(db, table, recover=None):
1073 2786 aaronmk
    return value(run_query(db, mk_select(db, table, [sql_gen.row_count],
1074 3298 aaronmk
        order_by=None), recover=recover, log_level=3))
1075 2426 aaronmk
1076 2414 aaronmk
def table_cols(db, table, recover=None):
1077
    return list(col_names(select(db, table, limit=0, order_by=None,
1078 2443 aaronmk
        recover=recover, log_level=4)))
1079 2414 aaronmk
1080 4261 aaronmk
pkey_col = 'row_num'
1081
1082 2291 aaronmk
def pkey(db, table, recover=None):
1083 4261 aaronmk
    '''Uses pkey_col, or if not found, the first column in the table.'''
1084
    cols = table_cols(db, table, recover)
1085
    if pkey_col in cols: return pkey_col
1086
    else: return cols[0]
1087 832 aaronmk
1088 2559 aaronmk
not_null_col = 'not_null_col'
1089 2340 aaronmk
1090
def table_not_null_col(db, table, recover=None):
1091
    '''Name assumed to be the value of not_null_col. If not found, uses pkey.'''
1092
    if not_null_col in table_cols(db, table, recover): return not_null_col
1093
    else: return pkey(db, table, recover)
1094
1095 3348 aaronmk
def constraint_cond(db, constraint):
1096
    module = util.root_module(db.db)
1097
    if module == 'psycopg2':
1098
        table_str = sql_gen.Literal(constraint.table.to_str(db))
1099
        name_str = sql_gen.Literal(constraint.name)
1100
        return value(run_query(db, '''\
1101
SELECT consrc
1102
FROM pg_constraint
1103
WHERE
1104
conrelid = '''+table_str.to_str(db)+'''::regclass
1105
AND conname = '''+name_str.to_str(db)+'''
1106
'''
1107
            , cacheable=True, log_level=4))
1108
    else: raise NotImplementedError("Can't list index columns for "+module+
1109
        ' database')
1110
1111 3319 aaronmk
def index_cols(db, index):
1112 853 aaronmk
    '''Can also use this for UNIQUE constraints, because a UNIQUE index is
1113
    automatically created. When you don't know whether something is a UNIQUE
1114
    constraint or a UNIQUE index, use this function.'''
1115 3322 aaronmk
    index = sql_gen.as_Table(index)
1116 1909 aaronmk
    module = util.root_module(db.db)
1117
    if module == 'psycopg2':
1118 3322 aaronmk
        qual_index = sql_gen.Literal(index.to_str(db))
1119
        return map(parse_expr_col, values(run_query(db, '''\
1120
SELECT pg_get_indexdef(indexrelid, generate_series(1, indnatts), true)
1121
FROM pg_index
1122
WHERE indexrelid = '''+qual_index.to_str(db)+'''::regclass
1123 2782 aaronmk
'''
1124
            , cacheable=True, log_level=4)))
1125 1909 aaronmk
    else: raise NotImplementedError("Can't list index columns for "+module+
1126
        ' database')
1127 853 aaronmk
1128 3079 aaronmk
#### Functions
1129
1130
def function_exists(db, function):
1131 3423 aaronmk
    qual_function = sql_gen.Literal(function.to_str(db))
1132
    try:
1133 3425 aaronmk
        select(db, fields=[sql_gen.Cast('regproc', qual_function)],
1134
            recover=True, cacheable=True, log_level=4)
1135 3423 aaronmk
    except DoesNotExistException: return False
1136 4146 aaronmk
    except DuplicateException: return True # overloaded function
1137 3423 aaronmk
    else: return True
1138 3079 aaronmk
1139
##### Structural changes
1140
1141
#### Columns
1142
1143
def add_col(db, table, col, comment=None, **kw_args):
1144
    '''
1145
    @param col TypedCol Name may be versioned, so be sure to propagate any
1146
        renaming back to any source column for the TypedCol.
1147
    @param comment None|str SQL comment used to distinguish columns of the same
1148
        name from each other when they contain different data, to allow the
1149
        ADD COLUMN query to be cached. If not set, query will not be cached.
1150
    '''
1151
    assert isinstance(col, sql_gen.TypedCol)
1152
1153
    while True:
1154
        str_ = 'ALTER TABLE '+table.to_str(db)+' ADD COLUMN '+col.to_str(db)
1155
        if comment != None: str_ += ' '+sql_gen.esc_comment(comment)
1156
1157
        try:
1158
            run_query(db, str_, recover=True, cacheable=True, **kw_args)
1159
            break
1160
        except DuplicateException:
1161
            col.name = next_version(col.name)
1162
            # try again with next version of name
1163
1164
def add_not_null(db, col):
1165
    table = col.table
1166
    col = sql_gen.to_name_only_col(col)
1167
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ALTER COLUMN '
1168
        +col.to_str(db)+' SET NOT NULL', cacheable=True, log_level=3)
1169
1170 4443 aaronmk
def drop_not_null(db, col):
1171
    table = col.table
1172
    col = sql_gen.to_name_only_col(col)
1173
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ALTER COLUMN '
1174
        +col.to_str(db)+' DROP NOT NULL', cacheable=True, log_level=3)
1175
1176 2096 aaronmk
row_num_col = '_row_num'
1177
1178 3079 aaronmk
row_num_typed_col = sql_gen.TypedCol(row_num_col, 'serial', nullable=False,
1179
    constraints='PRIMARY KEY')
1180
1181
def add_row_num(db, table):
1182
    '''Adds a row number column to a table. Its name is in row_num_col. It will
1183
    be the primary key.'''
1184
    add_col(db, table, row_num_typed_col, log_level=3)
1185
1186
#### Indexes
1187
1188
def add_pkey(db, table, cols=None, recover=None):
1189
    '''Adds a primary key.
1190
    @param cols [sql_gen.Col,...] The columns in the primary key.
1191
        Defaults to the first column in the table.
1192
    @pre The table must not already have a primary key.
1193
    '''
1194
    table = sql_gen.as_Table(table)
1195
    if cols == None: cols = [pkey(db, table, recover)]
1196
    col_strs = [sql_gen.to_name_only_col(v).to_str(db) for v in cols]
1197
1198
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ADD PRIMARY KEY ('
1199
        +(', '.join(col_strs))+')', recover=True, cacheable=True, log_level=3,
1200
        log_ignore_excs=(DuplicateException,))
1201
1202 2998 aaronmk
def add_index(db, exprs, table=None, unique=False, ensure_not_null_=True):
1203 2688 aaronmk
    '''Adds an index on column(s) or expression(s) if it doesn't already exist.
1204 3356 aaronmk
    Currently, only function calls and literal values are supported expressions.
1205 2998 aaronmk
    @param ensure_not_null_ If set, translates NULL values to sentinel values.
1206 2847 aaronmk
        This allows indexes to be used for comparisons where NULLs are equal.
1207 2538 aaronmk
    '''
1208 2964 aaronmk
    exprs = lists.mk_seq(exprs)
1209 2538 aaronmk
1210 2688 aaronmk
    # Parse exprs
1211
    old_exprs = exprs[:]
1212
    exprs = []
1213
    cols = []
1214
    for i, expr in enumerate(old_exprs):
1215 2823 aaronmk
        expr = sql_gen.as_Col(expr, table)
1216 2688 aaronmk
1217 2823 aaronmk
        # Handle nullable columns
1218 2998 aaronmk
        if ensure_not_null_:
1219 3164 aaronmk
            try: expr = sql_gen.ensure_not_null(db, expr)
1220 2860 aaronmk
            except KeyError: pass # unknown type, so just create plain index
1221 2823 aaronmk
1222 2688 aaronmk
        # Extract col
1223 3002 aaronmk
        expr = copy.deepcopy(expr) # don't modify input!
1224 3356 aaronmk
        col = expr
1225
        if isinstance(expr, sql_gen.FunctionCall): col = expr.args[0]
1226
        expr = sql_gen.cast_literal(expr)
1227
        if not isinstance(expr, (sql_gen.Expr, sql_gen.Col)):
1228 2688 aaronmk
            expr = sql_gen.Expr(expr)
1229 3356 aaronmk
1230 2688 aaronmk
1231
        # Extract table
1232
        if table == None:
1233
            assert sql_gen.is_table_col(col)
1234
            table = col.table
1235
1236 3356 aaronmk
        if isinstance(col, sql_gen.Col): col.table = None
1237 2688 aaronmk
1238
        exprs.append(expr)
1239
        cols.append(col)
1240 2408 aaronmk
1241 2688 aaronmk
    table = sql_gen.as_Table(table)
1242
1243 3005 aaronmk
    # Add index
1244 3148 aaronmk
    str_ = 'CREATE'
1245
    if unique: str_ += ' UNIQUE'
1246
    str_ += ' INDEX ON '+table.to_str(db)+' ('+(
1247
        ', '.join((v.to_str(db) for v in exprs)))+')'
1248
    run_query(db, str_, recover=True, cacheable=True, log_level=3)
1249 2408 aaronmk
1250 3083 aaronmk
already_indexed = object() # tells add_indexes() the pkey has already been added
1251
1252
def add_indexes(db, table, has_pkey=True):
1253
    '''Adds an index on all columns in a table.
1254
    @param has_pkey bool|already_indexed Whether a pkey instead of a regular
1255
        index should be added on the first column.
1256
        * If already_indexed, the pkey is assumed to have already been added
1257
    '''
1258
    cols = table_cols(db, table)
1259
    if has_pkey:
1260
        if has_pkey is not already_indexed: add_pkey(db, table)
1261
        cols = cols[1:]
1262
    for col in cols: add_index(db, col, table)
1263
1264 3079 aaronmk
#### Tables
1265 2772 aaronmk
1266 3079 aaronmk
### Maintenance
1267 2772 aaronmk
1268 3079 aaronmk
def analyze(db, table):
1269
    table = sql_gen.as_Table(table)
1270
    run_query(db, 'ANALYZE '+table.to_str(db), log_level=3)
1271 2934 aaronmk
1272 3079 aaronmk
def autoanalyze(db, table):
1273
    if db.autoanalyze: analyze(db, table)
1274 2935 aaronmk
1275 3079 aaronmk
def vacuum(db, table):
1276
    table = sql_gen.as_Table(table)
1277
    db.with_autocommit(lambda: run_query(db, 'VACUUM ANALYZE '+table.to_str(db),
1278
        log_level=3))
1279 2086 aaronmk
1280 3079 aaronmk
### Lifecycle
1281
1282 3247 aaronmk
def drop(db, type_, name):
1283
    name = sql_gen.as_Name(name)
1284
    run_query(db, 'DROP '+type_+' IF EXISTS '+name.to_str(db)+' CASCADE')
1285 2889 aaronmk
1286 3247 aaronmk
def drop_table(db, table): drop(db, 'TABLE', table)
1287
1288 3082 aaronmk
def create_table(db, table, cols=[], has_pkey=True, col_indexes=True,
1289
    like=None):
1290 2675 aaronmk
    '''Creates a table.
1291 2681 aaronmk
    @param cols [sql_gen.TypedCol,...] The column names and types
1292
    @param has_pkey If set, the first column becomes the primary key.
1293 2760 aaronmk
    @param col_indexes bool|[ref]
1294
        * If True, indexes will be added on all non-pkey columns.
1295
        * If a list reference, [0] will be set to a function to do this.
1296
          This can be used to delay index creation until the table is populated.
1297 2675 aaronmk
    '''
1298
    table = sql_gen.as_Table(table)
1299
1300 3082 aaronmk
    if like != None:
1301
        cols = [sql_gen.CustomCode('LIKE '+like.to_str(db)+' INCLUDING ALL')
1302
            ]+cols
1303 2681 aaronmk
    if has_pkey:
1304
        cols[0] = pkey = copy.copy(cols[0]) # don't modify input!
1305 2872 aaronmk
        pkey.constraints = 'PRIMARY KEY'
1306 2681 aaronmk
1307 3085 aaronmk
    temp = table.is_temp and not db.debug_temp
1308
        # temp tables permanent in debug_temp mode
1309 2760 aaronmk
1310 3085 aaronmk
    # Create table
1311 3383 aaronmk
    def create():
1312 3085 aaronmk
        str_ = 'CREATE'
1313
        if temp: str_ += ' TEMP'
1314
        str_ += ' TABLE '+table.to_str(db)+' (\n'
1315
        str_ += '\n, '.join(c.to_str(db) for c in cols)
1316 3126 aaronmk
        str_ += '\n);'
1317 3085 aaronmk
1318 3383 aaronmk
        run_query(db, str_, recover=True, cacheable=True, log_level=2,
1319
            log_ignore_excs=(DuplicateException,))
1320
    if table.is_temp:
1321
        while True:
1322
            try:
1323
                create()
1324
                break
1325
            except DuplicateException:
1326
                table.name = next_version(table.name)
1327
                # try again with next version of name
1328
    else: create()
1329 3085 aaronmk
1330 2760 aaronmk
    # Add indexes
1331 2773 aaronmk
    if has_pkey: has_pkey = already_indexed
1332
    def add_indexes_(): add_indexes(db, table, has_pkey)
1333
    if isinstance(col_indexes, list): col_indexes[0] = add_indexes_ # defer
1334
    elif col_indexes: add_indexes_() # add now
1335 2675 aaronmk
1336 3084 aaronmk
def copy_table_struct(db, src, dest):
1337
    '''Creates a structure-only copy of a table. (Does not copy data.)'''
1338 3085 aaronmk
    create_table(db, dest, has_pkey=False, col_indexes=False, like=src)
1339 3084 aaronmk
1340 3079 aaronmk
### Data
1341 2684 aaronmk
1342 2970 aaronmk
def truncate(db, table, schema='public', **kw_args):
1343
    '''For params, see run_query()'''
1344 2777 aaronmk
    table = sql_gen.as_Table(table, schema)
1345 2970 aaronmk
    return run_query(db, 'TRUNCATE '+table.to_str(db)+' CASCADE', **kw_args)
1346 2732 aaronmk
1347 2965 aaronmk
def empty_temp(db, tables):
1348
    tables = lists.mk_seq(tables)
1349 2971 aaronmk
    for table in tables: truncate(db, table, log_level=3)
1350 2965 aaronmk
1351 1968 aaronmk
def empty_db(db, schema='public', **kw_args):
1352
    '''For kw_args, see tables()'''
1353
    for table in tables(db, schema, **kw_args): truncate(db, table, schema)
1354 3094 aaronmk
1355
def distinct_table(db, table, distinct_on):
1356
    '''Creates a copy of a temp table which is distinct on the given columns.
1357 3099 aaronmk
    The old and new tables will both get an index on these columns, to
1358
    facilitate merge joins.
1359 3097 aaronmk
    @param distinct_on If empty, creates a table with one row. This is useful if
1360
        your distinct_on columns are all literal values.
1361 3099 aaronmk
    @return The new table.
1362 3094 aaronmk
    '''
1363 3099 aaronmk
    new_table = sql_gen.suffixed_table(table, '_distinct')
1364 3411 aaronmk
    distinct_on = filter(sql_gen.is_table_col, distinct_on)
1365 3094 aaronmk
1366 3099 aaronmk
    copy_table_struct(db, table, new_table)
1367 3097 aaronmk
1368
    limit = None
1369
    if distinct_on == []: limit = 1 # one sample row
1370 3099 aaronmk
    else:
1371
        add_index(db, distinct_on, new_table, unique=True)
1372
        add_index(db, distinct_on, table) # for join optimization
1373 3097 aaronmk
1374 3313 aaronmk
    insert_select(db, new_table, None, mk_select(db, table, order_by=None,
1375
        limit=limit), ignore=True)
1376 3099 aaronmk
    analyze(db, new_table)
1377 3094 aaronmk
1378 3099 aaronmk
    return new_table