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
    def __init__(self, name, cols, cause=None):
58 2484 aaronmk
        DbException.__init__(self, 'Violated '+strings.as_tt(name)
59
            +' constraint on columns: '+strings.as_tt(', '.join(cols)), cause)
60 2306 aaronmk
        self.name = name
61 468 aaronmk
        self.cols = cols
62 11 aaronmk
63 2523 aaronmk
class MissingCastException(DbException):
64
    def __init__(self, type_, col, cause=None):
65
        DbException.__init__(self, 'Missing cast to type '+strings.as_tt(type_)
66
            +' on column: '+strings.as_tt(col), cause)
67
        self.type = type_
68
        self.col = col
69
70 2143 aaronmk
class NameException(DbException): pass
71
72 2306 aaronmk
class DuplicateKeyException(ConstraintException): pass
73 13 aaronmk
74 2306 aaronmk
class NullValueException(ConstraintException): pass
75 13 aaronmk
76 3109 aaronmk
class InvalidValueException(ExceptionWithValue): pass
77 2239 aaronmk
78 2945 aaronmk
class DuplicateException(ExceptionWithNameType): pass
79 2143 aaronmk
80 89 aaronmk
class EmptyRowException(DbException): pass
81
82 865 aaronmk
##### Warnings
83
84
class DbWarning(UserWarning): pass
85
86 1930 aaronmk
##### Result retrieval
87
88
def col_names(cur): return (col[0] for col in cur.description)
89
90
def rows(cur): return iter(lambda: cur.fetchone(), None)
91
92
def consume_rows(cur):
93
    '''Used to fetch all rows so result will be cached'''
94
    iters.consume_iter(rows(cur))
95
96
def next_row(cur): return rows(cur).next()
97
98
def row(cur):
99
    row_ = next_row(cur)
100
    consume_rows(cur)
101
    return row_
102
103
def next_value(cur): return next_row(cur)[0]
104
105
def value(cur): return row(cur)[0]
106
107
def values(cur): return iters.func_iter(lambda: next_value(cur))
108
109
def value_or_none(cur):
110
    try: return value(cur)
111
    except StopIteration: return None
112
113 2762 aaronmk
##### Escaping
114 2101 aaronmk
115 2573 aaronmk
def esc_name_by_module(module, name):
116
    if module == 'psycopg2' or module == None: quote = '"'
117 2101 aaronmk
    elif module == 'MySQLdb': quote = '`'
118
    else: raise NotImplementedError("Can't escape name for "+module+' database')
119 2500 aaronmk
    return sql_gen.esc_name(name, quote)
120 2101 aaronmk
121
def esc_name_by_engine(engine, name, **kw_args):
122
    return esc_name_by_module(db_engines[engine][0], name, **kw_args)
123
124
def esc_name(db, name, **kw_args):
125
    return esc_name_by_module(util.root_module(db.db), name, **kw_args)
126
127
def qual_name(db, schema, table):
128
    def esc_name_(name): return esc_name(db, name)
129
    table = esc_name_(table)
130
    if schema != None: return esc_name_(schema)+'.'+table
131
    else: return table
132
133 1869 aaronmk
##### Database connections
134 1849 aaronmk
135 2097 aaronmk
db_config_names = ['engine', 'host', 'user', 'password', 'database', 'schemas']
136 1926 aaronmk
137 1869 aaronmk
db_engines = {
138
    'MySQL': ('MySQLdb', {'password': 'passwd', 'database': 'db'}),
139
    'PostgreSQL': ('psycopg2', {}),
140
}
141
142
DatabaseErrors_set = set([DbException])
143
DatabaseErrors = tuple(DatabaseErrors_set)
144
145
def _add_module(module):
146
    DatabaseErrors_set.add(module.DatabaseError)
147
    global DatabaseErrors
148
    DatabaseErrors = tuple(DatabaseErrors_set)
149
150
def db_config_str(db_config):
151
    return db_config['engine']+' database '+db_config['database']
152
153 2448 aaronmk
log_debug_none = lambda msg, level=2: None
154 1901 aaronmk
155 1849 aaronmk
class DbConn:
156 2923 aaronmk
    def __init__(self, db_config, autocommit=True, caching=True,
157 3183 aaronmk
        log_debug=log_debug_none, debug_temp=False, src=None):
158 2915 aaronmk
        '''
159
        @param debug_temp Whether temporary objects should instead be permanent.
160
            This assists in debugging the internal objects used by the program.
161 3183 aaronmk
        @param src In autocommit mode, will be included in a comment in every
162
            query, to help identify the data source in pg_stat_activity.
163 2915 aaronmk
        '''
164 1869 aaronmk
        self.db_config = db_config
165 2190 aaronmk
        self.autocommit = autocommit
166
        self.caching = caching
167 1901 aaronmk
        self.log_debug = log_debug
168 2193 aaronmk
        self.debug = log_debug != log_debug_none
169 2915 aaronmk
        self.debug_temp = debug_temp
170 3183 aaronmk
        self.src = src
171 3074 aaronmk
        self.autoanalyze = False
172 3269 aaronmk
        self.autoexplain = False
173
        self.profile_row_ct = None
174 1869 aaronmk
175 3124 aaronmk
        self._savepoint = 0
176 3120 aaronmk
        self._reset()
177 1869 aaronmk
178
    def __getattr__(self, name):
179
        if name == '__dict__': raise Exception('getting __dict__')
180
        if name == 'db': return self._db()
181
        else: raise AttributeError()
182
183
    def __getstate__(self):
184
        state = copy.copy(self.__dict__) # shallow copy
185 1915 aaronmk
        state['log_debug'] = None # don't pickle the debug callback
186 1869 aaronmk
        state['_DbConn__db'] = None # don't pickle the connection
187
        return state
188
189 3118 aaronmk
    def clear_cache(self): self.query_results = {}
190
191 3120 aaronmk
    def _reset(self):
192 3118 aaronmk
        self.clear_cache()
193 3124 aaronmk
        assert self._savepoint == 0
194 3118 aaronmk
        self._notices_seen = set()
195
        self.__db = None
196
197 2165 aaronmk
    def connected(self): return self.__db != None
198
199 3116 aaronmk
    def close(self):
200 3119 aaronmk
        if not self.connected(): return
201
202 3135 aaronmk
        # Record that the automatic transaction is now closed
203 3136 aaronmk
        self._savepoint -= 1
204 3135 aaronmk
205 3119 aaronmk
        self.db.close()
206 3120 aaronmk
        self._reset()
207 3116 aaronmk
208 3125 aaronmk
    def reconnect(self):
209
        # Do not do this in test mode as it would roll back everything
210
        if self.autocommit: self.close()
211
        # Connection will be reopened automatically on first query
212
213 1869 aaronmk
    def _db(self):
214
        if self.__db == None:
215
            # Process db_config
216
            db_config = self.db_config.copy() # don't modify input!
217 2097 aaronmk
            schemas = db_config.pop('schemas', None)
218 1869 aaronmk
            module_name, mappings = db_engines[db_config.pop('engine')]
219
            module = __import__(module_name)
220
            _add_module(module)
221
            for orig, new in mappings.iteritems():
222
                try: util.rename_key(db_config, orig, new)
223
                except KeyError: pass
224
225
            # Connect
226
            self.__db = module.connect(**db_config)
227
228 3161 aaronmk
            # Record that a transaction is already open
229
            self._savepoint += 1
230
231 1869 aaronmk
            # Configure connection
232 2906 aaronmk
            if hasattr(self.db, 'set_isolation_level'):
233
                import psycopg2.extensions
234
                self.db.set_isolation_level(
235
                    psycopg2.extensions.ISOLATION_LEVEL_READ_COMMITTED)
236 2101 aaronmk
            if schemas != None:
237 2893 aaronmk
                search_path = [self.esc_name(s) for s in schemas.split(',')]
238
                search_path.append(value(run_query(self, 'SHOW search_path',
239
                    log_level=4)))
240
                run_query(self, 'SET search_path TO '+(','.join(search_path)),
241
                    log_level=3)
242 1869 aaronmk
243
        return self.__db
244 1889 aaronmk
245 1891 aaronmk
    class DbCursor(Proxy):
246 1927 aaronmk
        def __init__(self, outer):
247 1891 aaronmk
            Proxy.__init__(self, outer.db.cursor())
248 2191 aaronmk
            self.outer = outer
249 1927 aaronmk
            self.query_results = outer.query_results
250 1894 aaronmk
            self.query_lookup = None
251 1891 aaronmk
            self.result = []
252 1889 aaronmk
253 2802 aaronmk
        def execute(self, query):
254 2764 aaronmk
            self._is_insert = query.startswith('INSERT')
255 2797 aaronmk
            self.query_lookup = query
256 2148 aaronmk
            try:
257 3162 aaronmk
                try: cur = self.inner.execute(query)
258 2802 aaronmk
                finally: self.query = get_cur_query(self.inner, query)
259 1904 aaronmk
            except Exception, e:
260
                self.result = e # cache the exception as the result
261
                self._cache_result()
262
                raise
263 3004 aaronmk
264
            # Always cache certain queries
265 3183 aaronmk
            query = sql_gen.lstrip(query)
266 3004 aaronmk
            if query.startswith('CREATE') or query.startswith('ALTER'):
267 3007 aaronmk
                # structural changes
268 3040 aaronmk
                # Rest of query must be unique in the face of name collisions,
269
                # so don't cache ADD COLUMN unless it has distinguishing comment
270
                if query.find('ADD COLUMN') < 0 or query.endswith('*/'):
271 3007 aaronmk
                    self._cache_result()
272 3004 aaronmk
            elif self.rowcount == 0 and query.startswith('SELECT'): # empty
273 2800 aaronmk
                consume_rows(self) # fetch all rows so result will be cached
274 3004 aaronmk
275 2762 aaronmk
            return cur
276 1894 aaronmk
277 1891 aaronmk
        def fetchone(self):
278
            row = self.inner.fetchone()
279 1899 aaronmk
            if row != None: self.result.append(row)
280
            # otherwise, fetched all rows
281 1904 aaronmk
            else: self._cache_result()
282
            return row
283
284
        def _cache_result(self):
285 2948 aaronmk
            # For inserts that return a result set, don't cache result set since
286
            # inserts are not idempotent. Other non-SELECT queries don't have
287
            # their result set read, so only exceptions will be cached (an
288
            # invalid query will always be invalid).
289 1930 aaronmk
            if self.query_results != None and (not self._is_insert
290 1906 aaronmk
                or isinstance(self.result, Exception)):
291
292 1894 aaronmk
                assert self.query_lookup != None
293 1916 aaronmk
                self.query_results[self.query_lookup] = self.CacheCursor(
294
                    util.dict_subset(dicts.AttrsDictView(self),
295
                    ['query', 'result', 'rowcount', 'description']))
296 1906 aaronmk
297 1916 aaronmk
        class CacheCursor:
298
            def __init__(self, cached_result): self.__dict__ = cached_result
299
300 1927 aaronmk
            def execute(self, *args, **kw_args):
301 1916 aaronmk
                if isinstance(self.result, Exception): raise self.result
302
                # otherwise, result is a rows list
303
                self.iter = iter(self.result)
304
305
            def fetchone(self):
306
                try: return self.iter.next()
307
                except StopIteration: return None
308 1891 aaronmk
309 2212 aaronmk
    def esc_value(self, value):
310 2663 aaronmk
        try: str_ = self.mogrify('%s', [value])
311
        except NotImplementedError, e:
312
            module = util.root_module(self.db)
313
            if module == 'MySQLdb':
314
                import _mysql
315
                str_ = _mysql.escape_string(value)
316
            else: raise e
317 2374 aaronmk
        return strings.to_unicode(str_)
318 2212 aaronmk
319 2347 aaronmk
    def esc_name(self, name): return esc_name(self, name) # calls global func
320
321 2814 aaronmk
    def std_code(self, str_):
322
        '''Standardizes SQL code.
323
        * Ensures that string literals are prefixed by `E`
324
        '''
325
        if str_.startswith("'"): str_ = 'E'+str_
326
        return str_
327
328 2665 aaronmk
    def can_mogrify(self):
329 2663 aaronmk
        module = util.root_module(self.db)
330 2665 aaronmk
        return module == 'psycopg2'
331 2663 aaronmk
332 2665 aaronmk
    def mogrify(self, query, params=None):
333
        if self.can_mogrify(): return self.db.cursor().mogrify(query, params)
334
        else: raise NotImplementedError("Can't mogrify query")
335
336 2671 aaronmk
    def print_notices(self):
337 2725 aaronmk
        if hasattr(self.db, 'notices'):
338
            for msg in self.db.notices:
339
                if msg not in self._notices_seen:
340
                    self._notices_seen.add(msg)
341
                    self.log_debug(msg, level=2)
342 2671 aaronmk
343 2793 aaronmk
    def run_query(self, query, cacheable=False, log_level=2,
344 2464 aaronmk
        debug_msg_ref=None):
345 2445 aaronmk
        '''
346 2464 aaronmk
        @param log_ignore_excs The log_level will be increased by 2 if the query
347
            throws one of these exceptions.
348 2664 aaronmk
        @param debug_msg_ref If specified, the log message will be returned in
349
            this instead of being output. This allows you to filter log messages
350
            depending on the result of the query.
351 2445 aaronmk
        '''
352 2167 aaronmk
        assert query != None
353
354 3183 aaronmk
        if self.autocommit and self.src != None:
355 3206 aaronmk
            query = sql_gen.esc_comment(self.src)+'\t'+query
356 3183 aaronmk
357 2047 aaronmk
        if not self.caching: cacheable = False
358 1903 aaronmk
        used_cache = False
359 2664 aaronmk
360 3242 aaronmk
        if self.debug:
361
            profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
362 1903 aaronmk
        try:
363 1927 aaronmk
            # Get cursor
364
            if cacheable:
365 3238 aaronmk
                try: cur = self.query_results[query]
366 1927 aaronmk
                except KeyError: cur = self.DbCursor(self)
367 3238 aaronmk
                else: used_cache = True
368 1927 aaronmk
            else: cur = self.db.cursor()
369
370
            # Run query
371 3238 aaronmk
            try: cur.execute(query)
372 3162 aaronmk
            except Exception, e:
373
                _add_cursor_info(e, self, query)
374
                raise
375 3238 aaronmk
            else: self.do_autocommit()
376 1903 aaronmk
        finally:
377 3242 aaronmk
            if self.debug:
378 3244 aaronmk
                profiler.stop(self.profile_row_ct)
379 3242 aaronmk
380
                ## Log or return query
381
382 3239 aaronmk
                query = str(get_cur_query(cur, query))
383
                # So that the src comment is hidden when cating the file, and
384
                # put on a separate line when viewed in a text editor.
385
                query = query.replace('\t', '\r', 1)
386
387 3240 aaronmk
                msg = 'DB query: '
388 3239 aaronmk
389 3240 aaronmk
                if used_cache: msg += 'cache hit'
390
                elif cacheable: msg += 'cache miss'
391
                else: msg += 'non-cacheable'
392 3239 aaronmk
393 3241 aaronmk
                msg += ':\n'+profiler.msg()+'\n'+strings.as_code(query, 'SQL')
394 3240 aaronmk
395 3237 aaronmk
                if debug_msg_ref != None: debug_msg_ref[0] = msg
396
                else: self.log_debug(msg, log_level)
397 3245 aaronmk
398
                self.print_notices()
399 1903 aaronmk
400
        return cur
401 1914 aaronmk
402 2797 aaronmk
    def is_cached(self, query): return query in self.query_results
403 2139 aaronmk
404 2907 aaronmk
    def with_autocommit(self, func):
405 2801 aaronmk
        import psycopg2.extensions
406
407
        prev_isolation_level = self.db.isolation_level
408 2907 aaronmk
        self.db.set_isolation_level(
409
            psycopg2.extensions.ISOLATION_LEVEL_AUTOCOMMIT)
410 2683 aaronmk
        try: return func()
411 2801 aaronmk
        finally: self.db.set_isolation_level(prev_isolation_level)
412 2683 aaronmk
413 2139 aaronmk
    def with_savepoint(self, func):
414 3137 aaronmk
        top = self._savepoint == 0
415 2171 aaronmk
        savepoint = 'level_'+str(self._savepoint)
416 3137 aaronmk
417 3272 aaronmk
        if self.debug:
418 3273 aaronmk
            self.log_debug('Begin transaction', level=4)
419 3272 aaronmk
            profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
420
421 3160 aaronmk
        # Must happen before running queries so they don't get autocommitted
422
        self._savepoint += 1
423
424 3137 aaronmk
        if top: query = 'START TRANSACTION ISOLATION LEVEL READ COMMITTED'
425
        else: query = 'SAVEPOINT '+savepoint
426
        self.run_query(query, log_level=4)
427
        try:
428
            return func()
429
            if top: self.run_query('COMMIT', log_level=4)
430 2139 aaronmk
        except:
431 3137 aaronmk
            if top: query = 'ROLLBACK'
432
            else: query = 'ROLLBACK TO SAVEPOINT '+savepoint
433
            self.run_query(query, log_level=4)
434
435 2139 aaronmk
            raise
436 2930 aaronmk
        finally:
437
            # Always release savepoint, because after ROLLBACK TO SAVEPOINT,
438
            # "The savepoint remains valid and can be rolled back to again"
439
            # (http://www.postgresql.org/docs/8.3/static/sql-rollback-to.html).
440 3137 aaronmk
            if not top:
441
                self.run_query('RELEASE SAVEPOINT '+savepoint, log_level=4)
442 2930 aaronmk
443
            self._savepoint -= 1
444
            assert self._savepoint >= 0
445
446 3272 aaronmk
            if self.debug:
447
                profiler.stop(self.profile_row_ct)
448 3273 aaronmk
                self.log_debug('End transaction\n'+profiler.msg(), level=4)
449 3272 aaronmk
450 2930 aaronmk
            self.do_autocommit() # OK to do this after ROLLBACK TO SAVEPOINT
451 2191 aaronmk
452
    def do_autocommit(self):
453
        '''Autocommits if outside savepoint'''
454 3135 aaronmk
        assert self._savepoint >= 1
455
        if self.autocommit and self._savepoint == 1:
456 2924 aaronmk
            self.log_debug('Autocommitting', level=4)
457 2191 aaronmk
            self.db.commit()
458 2643 aaronmk
459 3155 aaronmk
    def col_info(self, col, cacheable=True):
460 2643 aaronmk
        table = sql_gen.Table('columns', 'information_schema')
461 3063 aaronmk
        type_ = sql_gen.Coalesce(sql_gen.Nullif(sql_gen.Col('data_type'),
462
            'USER-DEFINED'), sql_gen.Col('udt_name'))
463 3078 aaronmk
        cols = [type_, 'column_default',
464
            sql_gen.Cast('boolean', sql_gen.Col('is_nullable'))]
465 2643 aaronmk
466
        conds = [('table_name', col.table.name), ('column_name', col.name)]
467
        schema = col.table.schema
468
        if schema != None: conds.append(('table_schema', schema))
469
470 2819 aaronmk
        type_, default, nullable = row(select(self, table, cols, conds,
471 3150 aaronmk
            order_by='table_schema', limit=1, cacheable=cacheable, log_level=4))
472 2643 aaronmk
            # TODO: order_by search_path schema order
473 2819 aaronmk
        default = sql_gen.as_Code(default, self)
474
475
        return sql_gen.TypedCol(col.name, type_, default, nullable)
476 2917 aaronmk
477
    def TempFunction(self, name):
478
        if self.debug_temp: schema = None
479
        else: schema = 'pg_temp'
480
        return sql_gen.Function(name, schema)
481 1849 aaronmk
482 1869 aaronmk
connect = DbConn
483
484 832 aaronmk
##### Recoverable querying
485 15 aaronmk
486 2139 aaronmk
def with_savepoint(db, func): return db.with_savepoint(func)
487 11 aaronmk
488 2791 aaronmk
def run_query(db, query, recover=None, cacheable=False, log_level=2,
489
    log_ignore_excs=None, **kw_args):
490 2794 aaronmk
    '''For params, see DbConn.run_query()'''
491 830 aaronmk
    if recover == None: recover = False
492 2464 aaronmk
    if log_ignore_excs == None: log_ignore_excs = ()
493
    log_ignore_excs = tuple(log_ignore_excs)
494 3236 aaronmk
    debug_msg_ref = [None]
495 830 aaronmk
496 3267 aaronmk
    query = with_explain_comment(db, query)
497 3258 aaronmk
498 2148 aaronmk
    try:
499 2464 aaronmk
        try:
500 2794 aaronmk
            def run(): return db.run_query(query, cacheable, log_level,
501 2793 aaronmk
                debug_msg_ref, **kw_args)
502 2796 aaronmk
            if recover and not db.is_cached(query):
503 2464 aaronmk
                return with_savepoint(db, run)
504
            else: return run() # don't need savepoint if cached
505
        except Exception, e:
506 3095 aaronmk
            msg = strings.ustr(e.args[0])
507 2464 aaronmk
508 3095 aaronmk
            match = re.match(r'^duplicate key value violates unique constraint '
509 3096 aaronmk
                r'"((_?[^\W_]+(?=[._]))?.+?)"', msg)
510 2464 aaronmk
            if match:
511
                constraint, table = match.groups()
512 3025 aaronmk
                cols = []
513
                if recover: # need auto-rollback to run index_cols()
514
                    try: cols = index_cols(db, table, constraint)
515
                    except NotImplementedError: pass
516
                raise DuplicateKeyException(constraint, cols, e)
517 2464 aaronmk
518 3095 aaronmk
            match = re.match(r'^null value in column "(.+?)" violates not-null'
519 2464 aaronmk
                r' constraint', msg)
520
            if match: raise NullValueException('NOT NULL', [match.group(1)], e)
521
522 3095 aaronmk
            match = re.match(r'^(?:invalid input (?:syntax|value)\b.*?'
523 3109 aaronmk
                r'|.+? field value out of range): "(.+?)"', msg)
524 2464 aaronmk
            if match:
525 3109 aaronmk
                value, = match.groups()
526
                raise InvalidValueException(strings.to_unicode(value), e)
527 2464 aaronmk
528 3095 aaronmk
            match = re.match(r'^column "(.+?)" is of type (.+?) but expression '
529 2523 aaronmk
                r'is of type', msg)
530
            if match:
531
                col, type_ = match.groups()
532
                raise MissingCastException(type_, col, e)
533
534 3095 aaronmk
            match = re.match(r'^(\S+) "(.+?)".*? already exists', msg)
535 2945 aaronmk
            if match:
536
                type_, name = match.groups()
537
                raise DuplicateException(type_, name, e)
538 2464 aaronmk
539
            raise # no specific exception raised
540
    except log_ignore_excs:
541
        log_level += 2
542
        raise
543
    finally:
544 3236 aaronmk
        if debug_msg_ref[0] != None: db.log_debug(debug_msg_ref[0], log_level)
545 830 aaronmk
546 832 aaronmk
##### Basic queries
547
548 3256 aaronmk
def is_explainable(query):
549
    # See <http://www.postgresql.org/docs/8.3/static/sql-explain.html#AEN57749>
550 3257 aaronmk
    return re.match(r'^(?:SELECT|INSERT|UPDATE|DELETE|VALUES|EXECUTE|DECLARE)\b'
551
        , query)
552 3256 aaronmk
553 3263 aaronmk
def explain(db, query, **kw_args):
554
    '''
555
    For params, see run_query().
556
    '''
557 3267 aaronmk
    kw_args.setdefault('log_level', 4)
558 3263 aaronmk
559 3256 aaronmk
    return strings.join_lines(values(run_query(db, 'EXPLAIN '+query,
560 3263 aaronmk
        recover=True, cacheable=True, **kw_args)))
561 3256 aaronmk
        # not a higher log_level because it's useful to see what query is being
562
        # run before it's executed, which EXPLAIN effectively provides
563
564 3265 aaronmk
def has_comment(query): return query.endswith('*/')
565
566
def with_explain_comment(db, query, **kw_args):
567 3269 aaronmk
    if db.autoexplain and not has_comment(query) and is_explainable(query):
568 3265 aaronmk
        query += '\n'+sql_gen.esc_comment(' EXPLAIN:\n'
569
            +explain(db, query, **kw_args))
570
    return query
571
572 2153 aaronmk
def next_version(name):
573 2163 aaronmk
    version = 1 # first existing name was version 0
574 2586 aaronmk
    match = re.match(r'^(.*)#(\d+)$', name)
575 2153 aaronmk
    if match:
576 2586 aaronmk
        name, version = match.groups()
577
        version = int(version)+1
578 2932 aaronmk
    return sql_gen.concat(name, '#'+str(version))
579 2153 aaronmk
580 2899 aaronmk
def lock_table(db, table, mode):
581
    table = sql_gen.as_Table(table)
582
    run_query(db, 'LOCK TABLE '+table.to_str(db)+' IN '+mode+' MODE')
583
584 2789 aaronmk
def run_query_into(db, query, into=None, add_indexes_=False, **kw_args):
585 2085 aaronmk
    '''Outputs a query to a temp table.
586
    For params, see run_query().
587
    '''
588 2789 aaronmk
    if into == None: return run_query(db, query, **kw_args)
589 2790 aaronmk
590
    assert isinstance(into, sql_gen.Table)
591
592 2992 aaronmk
    into.is_temp = True
593 3008 aaronmk
    # "temporary tables cannot specify a schema name", so remove schema
594
    into.schema = None
595 2992 aaronmk
596 2790 aaronmk
    kw_args['recover'] = True
597 2945 aaronmk
    kw_args.setdefault('log_ignore_excs', (DuplicateException,))
598 2790 aaronmk
599 2916 aaronmk
    temp = not db.debug_temp # tables are permanent in debug_temp mode
600 2790 aaronmk
601
    # Create table
602
    while True:
603
        create_query = 'CREATE'
604
        if temp: create_query += ' TEMP'
605
        create_query += ' TABLE '+into.to_str(db)+' AS\n'+query
606 2385 aaronmk
607 2790 aaronmk
        try:
608
            cur = run_query(db, create_query, **kw_args)
609
                # CREATE TABLE AS sets rowcount to # rows in query
610
            break
611 2945 aaronmk
        except DuplicateException, e:
612 2790 aaronmk
            into.name = next_version(into.name)
613
            # try again with next version of name
614
615
    if add_indexes_: add_indexes(db, into)
616 3075 aaronmk
617
    # According to the PostgreSQL doc, "The autovacuum daemon cannot access and
618
    # therefore cannot vacuum or analyze temporary tables. [...] if a temporary
619
    # table is going to be used in complex queries, it is wise to run ANALYZE on
620
    # the temporary table after it is populated."
621
    # (http://www.postgresql.org/docs/9.1/static/sql-createtable.html)
622
    # If into is not a temp table, ANALYZE is useful but not required.
623 3073 aaronmk
    analyze(db, into)
624 2790 aaronmk
625
    return cur
626 2085 aaronmk
627 2120 aaronmk
order_by_pkey = object() # tells mk_select() to order by the pkey
628
629 2199 aaronmk
distinct_on_all = object() # tells mk_select() to SELECT DISTINCT ON all columns
630
631 2233 aaronmk
def mk_select(db, tables, fields=None, conds=None, distinct_on=[], limit=None,
632 2293 aaronmk
    start=None, order_by=order_by_pkey, default_table=None):
633 1981 aaronmk
    '''
634 2121 aaronmk
    @param tables The single table to select from, or a list of tables to join
635 2280 aaronmk
        together, with tables after the first being sql_gen.Join objects
636 1981 aaronmk
    @param fields Use None to select all fields in the table
637 2377 aaronmk
    @param conds WHERE conditions: [(compare_left_side, compare_right_side),...]
638 2379 aaronmk
        * container can be any iterable type
639 2399 aaronmk
        * compare_left_side: sql_gen.Code|str (for col name)
640
        * compare_right_side: sql_gen.ValueCond|literal value
641 2199 aaronmk
    @param distinct_on The columns to SELECT DISTINCT ON, or distinct_on_all to
642
        use all columns
643 2786 aaronmk
    @return query
644 1981 aaronmk
    '''
645 2315 aaronmk
    # Parse tables param
646 2964 aaronmk
    tables = lists.mk_seq(tables)
647 2141 aaronmk
    tables = list(tables) # don't modify input! (list() copies input)
648 2315 aaronmk
    table0 = sql_gen.as_Table(tables.pop(0)) # first table is separate
649 2121 aaronmk
650 2315 aaronmk
    # Parse other params
651 2376 aaronmk
    if conds == None: conds = []
652 2650 aaronmk
    elif dicts.is_dict(conds): conds = conds.items()
653 2379 aaronmk
    conds = list(conds) # don't modify input! (list() copies input)
654 3129 aaronmk
    assert limit == None or isinstance(limit, (int, long))
655
    assert start == None or isinstance(start, (int, long))
656 2315 aaronmk
    if order_by is order_by_pkey:
657
        if distinct_on != []: order_by = None
658
        else: order_by = pkey(db, table0, recover=True)
659 865 aaronmk
660 2315 aaronmk
    query = 'SELECT'
661 2056 aaronmk
662 2315 aaronmk
    def parse_col(col): return sql_gen.as_Col(col, default_table).to_str(db)
663 2056 aaronmk
664 2200 aaronmk
    # DISTINCT ON columns
665 2233 aaronmk
    if distinct_on != []:
666 2467 aaronmk
        query += '\nDISTINCT'
667 2254 aaronmk
        if distinct_on is not distinct_on_all:
668 2200 aaronmk
            query += ' ON ('+(', '.join(map(parse_col, distinct_on)))+')'
669
670
    # Columns
671 3185 aaronmk
    if query.find('\n') >= 0: whitespace = '\n'
672
    else: whitespace = ' '
673
    if fields == None: query += whitespace+'*'
674 2765 aaronmk
    else:
675
        assert fields != []
676 3185 aaronmk
        if len(fields) > 1: whitespace = '\n'
677
        query += whitespace+('\n, '.join(map(parse_col, fields)))
678 2200 aaronmk
679
    # Main table
680 3185 aaronmk
    if query.find('\n') >= 0 or len(tables) > 0: whitespace = '\n'
681
    else: whitespace = ' '
682
    query += whitespace+'FROM '+table0.to_str(db)
683 865 aaronmk
684 2122 aaronmk
    # Add joins
685 2271 aaronmk
    left_table = table0
686 2263 aaronmk
    for join_ in tables:
687
        table = join_.table
688 2238 aaronmk
689 2343 aaronmk
        # Parse special values
690
        if join_.type_ is sql_gen.filter_out: # filter no match
691 2376 aaronmk
            conds.append((sql_gen.Col(table_not_null_col(db, table), table),
692 2853 aaronmk
                sql_gen.CompareCond(None, '~=')))
693 2343 aaronmk
694 2467 aaronmk
        query += '\n'+join_.to_str(db, left_table)
695 2122 aaronmk
696
        left_table = table
697
698 865 aaronmk
    missing = True
699 2376 aaronmk
    if conds != []:
700 2576 aaronmk
        if len(conds) == 1: whitespace = ' '
701
        else: whitespace = '\n'
702 2578 aaronmk
        query += '\n'+sql_gen.combine_conds([sql_gen.ColValueCond(l, r)
703
            .to_str(db) for l, r in conds], 'WHERE')
704 865 aaronmk
        missing = False
705 2227 aaronmk
    if order_by != None:
706 2467 aaronmk
        query += '\nORDER BY '+sql_gen.as_Col(order_by, table0).to_str(db)
707
    if limit != None: query += '\nLIMIT '+str(limit); missing = False
708 865 aaronmk
    if start != None:
709 2467 aaronmk
        if start != 0: query += '\nOFFSET '+str(start)
710 865 aaronmk
        missing = False
711
    if missing: warnings.warn(DbWarning(
712
        'SELECT statement missing a WHERE, LIMIT, or OFFSET clause: '+query))
713
714 3266 aaronmk
    query = with_explain_comment(db, query)
715
716 2786 aaronmk
    return query
717 11 aaronmk
718 2054 aaronmk
def select(db, *args, **kw_args):
719
    '''For params, see mk_select() and run_query()'''
720
    recover = kw_args.pop('recover', None)
721
    cacheable = kw_args.pop('cacheable', True)
722 2442 aaronmk
    log_level = kw_args.pop('log_level', 2)
723 2054 aaronmk
724 2791 aaronmk
    return run_query(db, mk_select(db, *args, **kw_args), recover, cacheable,
725
        log_level=log_level)
726 2054 aaronmk
727 2788 aaronmk
def mk_insert_select(db, table, cols=None, select_query=None, returning=None,
728 3181 aaronmk
    embeddable=False, ignore=False, src=None):
729 1960 aaronmk
    '''
730
    @param returning str|None An inserted column (such as pkey) to return
731 2070 aaronmk
    @param embeddable Whether the query should be embeddable as a nested SELECT.
732 2073 aaronmk
        Warning: If you set this and cacheable=True when the query is run, the
733
        query will be fully cached, not just if it raises an exception.
734 3009 aaronmk
    @param ignore Whether to ignore duplicate keys.
735 3181 aaronmk
    @param src Will be included in the name of any created function, to help
736
        identify the data source in pg_stat_activity.
737 1960 aaronmk
    '''
738 2754 aaronmk
    table = sql_gen.remove_table_rename(sql_gen.as_Table(table))
739 2318 aaronmk
    if cols == []: cols = None # no cols (all defaults) = unknown col names
740 3010 aaronmk
    if cols != None: cols = [sql_gen.to_name_only_col(c, table) for c in cols]
741 2063 aaronmk
    if select_query == None: select_query = 'DEFAULT VALUES'
742 2327 aaronmk
    if returning != None: returning = sql_gen.as_Col(returning, table)
743 2063 aaronmk
744 2497 aaronmk
    first_line = 'INSERT INTO '+table.to_str(db)
745 2063 aaronmk
746 3009 aaronmk
    def mk_insert(select_query):
747
        query = first_line
748 3014 aaronmk
        if cols != None:
749
            query += '\n('+(', '.join((c.to_str(db) for c in cols)))+')'
750 3009 aaronmk
        query += '\n'+select_query
751
752
        if returning != None:
753
            returning_name_col = sql_gen.to_name_only_col(returning)
754
            query += '\nRETURNING '+returning_name_col.to_str(db)
755
756
        return query
757 2063 aaronmk
758 3017 aaronmk
    return_type = 'unknown'
759
    if returning != None: return_type = returning.to_str(db)+'%TYPE'
760
761 3009 aaronmk
    lang = 'sql'
762
    if ignore:
763 3017 aaronmk
        # Always return something to set the correct rowcount
764
        if returning == None: returning = sql_gen.NamedCol('NULL', None)
765
766 3009 aaronmk
        embeddable = True # must use function
767
        lang = 'plpgsql'
768 3010 aaronmk
769 3092 aaronmk
        if cols == None:
770
            row = [sql_gen.Col(sql_gen.all_cols, 'row')]
771
            row_vars = [sql_gen.Table('row')]
772
        else:
773
            row_vars = row = [sql_gen.Col(c.name, 'row') for c in cols]
774
775 3009 aaronmk
        query = '''\
776 3010 aaronmk
DECLARE
777 3014 aaronmk
    row '''+table.to_str(db)+'''%ROWTYPE;
778 3009 aaronmk
BEGIN
779 3019 aaronmk
    /* Need an EXCEPTION block for each individual row because "When an error is
780
    caught by an EXCEPTION clause, [...] all changes to persistent database
781
    state within the block are rolled back."
782
    This is unfortunate because "A block containing an EXCEPTION clause is
783
    significantly more expensive to enter and exit than a block without one."
784 3015 aaronmk
    (http://www.postgresql.org/docs/8.3/static/plpgsql-control-structures.html\
785
#PLPGSQL-ERROR-TRAPPING)
786
    */
787 3092 aaronmk
    FOR '''+(', '.join((v.to_str(db) for v in row_vars)))+''' IN
788 3034 aaronmk
'''+select_query+'''
789
    LOOP
790 3015 aaronmk
        BEGIN
791 3019 aaronmk
            RETURN QUERY
792 3014 aaronmk
'''+mk_insert(sql_gen.Values(row).to_str(db))+'''
793 3010 aaronmk
;
794 3015 aaronmk
        EXCEPTION
795 3019 aaronmk
            WHEN unique_violation THEN NULL; -- continue to next row
796 3015 aaronmk
        END;
797 3010 aaronmk
    END LOOP;
798
END;\
799 3009 aaronmk
'''
800
    else: query = mk_insert(select_query)
801
802 2070 aaronmk
    if embeddable:
803
        # Create function
804 2513 aaronmk
        function_name = sql_gen.clean_name(first_line)
805 3181 aaronmk
        if src != None: function_name = src+': '+function_name
806 2189 aaronmk
        while True:
807
            try:
808 2918 aaronmk
                function = db.TempFunction(function_name)
809 2194 aaronmk
810 2189 aaronmk
                function_query = '''\
811 2698 aaronmk
CREATE FUNCTION '''+function.to_str(db)+'''()
812 3017 aaronmk
RETURNS SETOF '''+return_type+'''
813 3009 aaronmk
LANGUAGE '''+lang+'''
814 2467 aaronmk
AS $$
815 3009 aaronmk
'''+query+'''
816 2467 aaronmk
$$;
817 2070 aaronmk
'''
818 2446 aaronmk
                run_query(db, function_query, recover=True, cacheable=True,
819 2945 aaronmk
                    log_ignore_excs=(DuplicateException,))
820 2189 aaronmk
                break # this version was successful
821 2945 aaronmk
            except DuplicateException, e:
822 2189 aaronmk
                function_name = next_version(function_name)
823
                # try again with next version of name
824 2070 aaronmk
825 2337 aaronmk
        # Return query that uses function
826 3009 aaronmk
        cols = None
827
        if returning != None: cols = [returning]
828 2698 aaronmk
        func_table = sql_gen.NamedTable('f', sql_gen.FunctionCall(function),
829 3009 aaronmk
            cols) # AS clause requires function alias
830 2787 aaronmk
        return mk_select(db, func_table, start=0, order_by=None)
831 2070 aaronmk
832 2787 aaronmk
    return query
833 2066 aaronmk
834 3074 aaronmk
def insert_select(db, table, *args, **kw_args):
835 2085 aaronmk
    '''For params, see mk_insert_select() and run_query_into()
836 2386 aaronmk
    @param into sql_gen.Table with suggested name of temp table to put RETURNING
837
        values in
838 2072 aaronmk
    '''
839 3141 aaronmk
    returning = kw_args.get('returning', None)
840
    ignore = kw_args.get('ignore', False)
841
842 2386 aaronmk
    into = kw_args.pop('into', None)
843
    if into != None: kw_args['embeddable'] = True
844 2066 aaronmk
    recover = kw_args.pop('recover', None)
845 3141 aaronmk
    if ignore: recover = True
846 2066 aaronmk
    cacheable = kw_args.pop('cacheable', True)
847 2673 aaronmk
    log_level = kw_args.pop('log_level', 2)
848 2066 aaronmk
849 3141 aaronmk
    rowcount_only = ignore and returning == None # keep NULL rows on server
850
    if rowcount_only: into = sql_gen.Table('rowcount')
851
852 3074 aaronmk
    cur = run_query_into(db, mk_insert_select(db, table, *args, **kw_args),
853
        into, recover=recover, cacheable=cacheable, log_level=log_level)
854 3141 aaronmk
    if rowcount_only: empty_temp(db, into)
855 3074 aaronmk
    autoanalyze(db, table)
856
    return cur
857 2063 aaronmk
858 2738 aaronmk
default = sql_gen.default # tells insert() to use the default value for a column
859 2066 aaronmk
860 2063 aaronmk
def insert(db, table, row, *args, **kw_args):
861 2085 aaronmk
    '''For params, see insert_select()'''
862 1960 aaronmk
    if lists.is_seq(row): cols = None
863
    else:
864
        cols = row.keys()
865
        row = row.values()
866 2738 aaronmk
    row = list(row) # ensure that "== []" works
867 1960 aaronmk
868 2738 aaronmk
    if row == []: query = None
869
    else: query = sql_gen.Values(row).to_str(db)
870 1961 aaronmk
871 2788 aaronmk
    return insert_select(db, table, cols, query, *args, **kw_args)
872 11 aaronmk
873 3152 aaronmk
def mk_update(db, table, changes=None, cond=None, in_place=False,
874 3153 aaronmk
    cacheable_=True):
875 2402 aaronmk
    '''
876
    @param changes [(col, new_value),...]
877
        * container can be any iterable type
878
        * col: sql_gen.Code|str (for col name)
879
        * new_value: sql_gen.Code|literal value
880
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
881 3056 aaronmk
    @param in_place If set, locks the table and updates rows in place.
882
        This avoids creating dead rows in PostgreSQL.
883
        * cond must be None
884 3153 aaronmk
    @param cacheable_ Whether column structure information used to generate the
885 3152 aaronmk
        query can be cached
886 2402 aaronmk
    @return str query
887
    '''
888 3057 aaronmk
    table = sql_gen.as_Table(table)
889
    changes = [(sql_gen.to_name_only_col(c, table), sql_gen.as_Value(v))
890
        for c, v in changes]
891
892 3056 aaronmk
    if in_place:
893
        assert cond == None
894 3058 aaronmk
895 3065 aaronmk
        query = 'ALTER TABLE '+table.to_str(db)+'\n'
896
        query += ',\n'.join(('ALTER COLUMN '+c.to_str(db)+' TYPE '
897 3153 aaronmk
            +db.col_info(sql_gen.with_default_table(c, table), cacheable_).type
898 3065 aaronmk
            +'\nUSING '+v.to_str(db) for c, v in changes))
899 3058 aaronmk
    else:
900
        query = 'UPDATE '+table.to_str(db)+'\nSET\n'
901
        query += ',\n'.join((c.to_str(db)+' = '+v.to_str(db)
902
            for c, v in changes))
903
        if cond != None: query += '\nWHERE\n'+cond.to_str(db)
904 3056 aaronmk
905 3266 aaronmk
    query = with_explain_comment(db, query)
906
907 2402 aaronmk
    return query
908
909 3074 aaronmk
def update(db, table, *args, **kw_args):
910 2402 aaronmk
    '''For params, see mk_update() and run_query()'''
911
    recover = kw_args.pop('recover', None)
912 3043 aaronmk
    cacheable = kw_args.pop('cacheable', False)
913 3030 aaronmk
    log_level = kw_args.pop('log_level', 2)
914 2402 aaronmk
915 3074 aaronmk
    cur = run_query(db, mk_update(db, table, *args, **kw_args), recover,
916
        cacheable, log_level=log_level)
917
    autoanalyze(db, table)
918
    return cur
919 2402 aaronmk
920 135 aaronmk
def last_insert_id(db):
921 1849 aaronmk
    module = util.root_module(db.db)
922 135 aaronmk
    if module == 'psycopg2': return value(run_query(db, 'SELECT lastval()'))
923
    elif module == 'MySQLdb': return db.insert_id()
924
    else: return None
925 13 aaronmk
926 2394 aaronmk
def mk_flatten_mapping(db, into, cols, preserve=[], as_items=False):
927 2383 aaronmk
    '''Creates a mapping from original column names (which may have collisions)
928 2415 aaronmk
    to names that will be distinct among the columns' tables.
929 2383 aaronmk
    This is meant to be used for several tables that are being joined together.
930 2415 aaronmk
    @param cols The columns to combine. Duplicates will be removed.
931
    @param into The table for the new columns.
932 2394 aaronmk
    @param preserve [sql_gen.Col...] Columns not to rename. Note that these
933
        columns will be included in the mapping even if they are not in cols.
934
        The tables of the provided Col objects will be changed to into, so make
935
        copies of them if you want to keep the original tables.
936
    @param as_items Whether to return a list of dict items instead of a dict
937 2383 aaronmk
    @return dict(orig_col=new_col, ...)
938
        * orig_col: sql_gen.Col(orig_col_name, orig_table)
939 2392 aaronmk
        * new_col: sql_gen.Col(orig_col_name, into)
940
        * All mappings use the into table so its name can easily be
941 2383 aaronmk
          changed for all columns at once
942
    '''
943 2415 aaronmk
    cols = lists.uniqify(cols)
944
945 2394 aaronmk
    items = []
946 2389 aaronmk
    for col in preserve:
947 2390 aaronmk
        orig_col = copy.copy(col)
948 2392 aaronmk
        col.table = into
949 2394 aaronmk
        items.append((orig_col, col))
950
    preserve = set(preserve)
951
    for col in cols:
952 2716 aaronmk
        if col not in preserve:
953
            items.append((col, sql_gen.Col(str(col), into, col.srcs)))
954 2394 aaronmk
955
    if not as_items: items = dict(items)
956
    return items
957 2383 aaronmk
958 2393 aaronmk
def flatten(db, into, joins, cols, limit=None, start=None, **kw_args):
959 2391 aaronmk
    '''For params, see mk_flatten_mapping()
960
    @return See return value of mk_flatten_mapping()
961
    '''
962 2394 aaronmk
    items = mk_flatten_mapping(db, into, cols, as_items=True, **kw_args)
963
    cols = [sql_gen.NamedCol(new.name, old) for old, new in items]
964 2786 aaronmk
    run_query_into(db, mk_select(db, joins, cols, limit=limit, start=start),
965 3166 aaronmk
        into=into)
966 2394 aaronmk
    return dict(items)
967 2391 aaronmk
968 3079 aaronmk
##### Database structure introspection
969 2414 aaronmk
970 3079 aaronmk
#### Tables
971
972
def tables(db, schema_like='public', table_like='%', exact=False):
973
    if exact: compare = '='
974
    else: compare = 'LIKE'
975
976
    module = util.root_module(db.db)
977
    if module == 'psycopg2':
978
        conds = [('schemaname', sql_gen.CompareCond(schema_like, compare)),
979
            ('tablename', sql_gen.CompareCond(table_like, compare))]
980
        return values(select(db, 'pg_tables', ['tablename'], conds,
981
            order_by='tablename', log_level=4))
982
    elif module == 'MySQLdb':
983
        return values(run_query(db, 'SHOW TABLES LIKE '+db.esc_value(table_like)
984
            , cacheable=True, log_level=4))
985
    else: raise NotImplementedError("Can't list tables for "+module+' database')
986
987
def table_exists(db, table):
988
    table = sql_gen.as_Table(table)
989
    return list(tables(db, table.schema, table.name, exact=True)) != []
990
991 2426 aaronmk
def table_row_count(db, table, recover=None):
992 2786 aaronmk
    return value(run_query(db, mk_select(db, table, [sql_gen.row_count],
993 2443 aaronmk
        order_by=None, start=0), recover=recover, log_level=3))
994 2426 aaronmk
995 2414 aaronmk
def table_cols(db, table, recover=None):
996
    return list(col_names(select(db, table, limit=0, order_by=None,
997 2443 aaronmk
        recover=recover, log_level=4)))
998 2414 aaronmk
999 2291 aaronmk
def pkey(db, table, recover=None):
1000 832 aaronmk
    '''Assumed to be first column in table'''
1001 2339 aaronmk
    return table_cols(db, table, recover)[0]
1002 832 aaronmk
1003 2559 aaronmk
not_null_col = 'not_null_col'
1004 2340 aaronmk
1005
def table_not_null_col(db, table, recover=None):
1006
    '''Name assumed to be the value of not_null_col. If not found, uses pkey.'''
1007
    if not_null_col in table_cols(db, table, recover): return not_null_col
1008
    else: return pkey(db, table, recover)
1009
1010 853 aaronmk
def index_cols(db, table, index):
1011
    '''Can also use this for UNIQUE constraints, because a UNIQUE index is
1012
    automatically created. When you don't know whether something is a UNIQUE
1013
    constraint or a UNIQUE index, use this function.'''
1014 1909 aaronmk
    module = util.root_module(db.db)
1015
    if module == 'psycopg2':
1016
        return list(values(run_query(db, '''\
1017 853 aaronmk
SELECT attname
1018 866 aaronmk
FROM
1019
(
1020
        SELECT attnum, attname
1021
        FROM pg_index
1022
        JOIN pg_class index ON index.oid = indexrelid
1023
        JOIN pg_class table_ ON table_.oid = indrelid
1024
        JOIN pg_attribute ON attrelid = indrelid AND attnum = ANY (indkey)
1025
        WHERE
1026 2782 aaronmk
            table_.relname = '''+db.esc_value(table)+'''
1027
            AND index.relname = '''+db.esc_value(index)+'''
1028 866 aaronmk
    UNION
1029
        SELECT attnum, attname
1030
        FROM
1031
        (
1032
            SELECT
1033
                indrelid
1034
                , (regexp_matches(indexprs, E':varattno (\\\\d+)', 'g'))[1]::int
1035
                    AS indkey
1036
            FROM pg_index
1037
            JOIN pg_class index ON index.oid = indexrelid
1038
            JOIN pg_class table_ ON table_.oid = indrelid
1039
            WHERE
1040 2782 aaronmk
                table_.relname = '''+db.esc_value(table)+'''
1041
                AND index.relname = '''+db.esc_value(index)+'''
1042 866 aaronmk
        ) s
1043
        JOIN pg_attribute ON attrelid = indrelid AND attnum = indkey
1044
) s
1045 853 aaronmk
ORDER BY attnum
1046 2782 aaronmk
'''
1047
            , cacheable=True, log_level=4)))
1048 1909 aaronmk
    else: raise NotImplementedError("Can't list index columns for "+module+
1049
        ' database')
1050 853 aaronmk
1051 464 aaronmk
def constraint_cols(db, table, constraint):
1052 1849 aaronmk
    module = util.root_module(db.db)
1053 464 aaronmk
    if module == 'psycopg2':
1054
        return list(values(run_query(db, '''\
1055
SELECT attname
1056
FROM pg_constraint
1057
JOIN pg_class ON pg_class.oid = conrelid
1058
JOIN pg_attribute ON attrelid = conrelid AND attnum = ANY (conkey)
1059
WHERE
1060 2783 aaronmk
    relname = '''+db.esc_value(table)+'''
1061
    AND conname = '''+db.esc_value(constraint)+'''
1062 464 aaronmk
ORDER BY attnum
1063 2783 aaronmk
'''
1064
            )))
1065 464 aaronmk
    else: raise NotImplementedError("Can't list constraint columns for "+module+
1066
        ' database')
1067
1068 3079 aaronmk
#### Functions
1069
1070
def function_exists(db, function):
1071
    function = sql_gen.as_Function(function)
1072
1073
    info_table = sql_gen.Table('routines', 'information_schema')
1074
    conds = [('routine_name', function.name)]
1075
    schema = function.schema
1076
    if schema != None: conds.append(('routine_schema', schema))
1077
    # Exclude trigger functions, since they cannot be called directly
1078
    conds.append(('data_type', sql_gen.CompareCond('trigger', '!=')))
1079
1080
    return list(values(select(db, info_table, ['routine_name'], conds,
1081
        order_by='routine_schema', limit=1, log_level=4))) != []
1082
        # TODO: order_by search_path schema order
1083
1084
##### Structural changes
1085
1086
#### Columns
1087
1088
def add_col(db, table, col, comment=None, **kw_args):
1089
    '''
1090
    @param col TypedCol Name may be versioned, so be sure to propagate any
1091
        renaming back to any source column for the TypedCol.
1092
    @param comment None|str SQL comment used to distinguish columns of the same
1093
        name from each other when they contain different data, to allow the
1094
        ADD COLUMN query to be cached. If not set, query will not be cached.
1095
    '''
1096
    assert isinstance(col, sql_gen.TypedCol)
1097
1098
    while True:
1099
        str_ = 'ALTER TABLE '+table.to_str(db)+' ADD COLUMN '+col.to_str(db)
1100
        if comment != None: str_ += ' '+sql_gen.esc_comment(comment)
1101
1102
        try:
1103
            run_query(db, str_, recover=True, cacheable=True, **kw_args)
1104
            break
1105
        except DuplicateException:
1106
            col.name = next_version(col.name)
1107
            # try again with next version of name
1108
1109
def add_not_null(db, col):
1110
    table = col.table
1111
    col = sql_gen.to_name_only_col(col)
1112
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ALTER COLUMN '
1113
        +col.to_str(db)+' SET NOT NULL', cacheable=True, log_level=3)
1114
1115 2096 aaronmk
row_num_col = '_row_num'
1116
1117 3079 aaronmk
row_num_typed_col = sql_gen.TypedCol(row_num_col, 'serial', nullable=False,
1118
    constraints='PRIMARY KEY')
1119
1120
def add_row_num(db, table):
1121
    '''Adds a row number column to a table. Its name is in row_num_col. It will
1122
    be the primary key.'''
1123
    add_col(db, table, row_num_typed_col, log_level=3)
1124
1125
#### Indexes
1126
1127
def add_pkey(db, table, cols=None, recover=None):
1128
    '''Adds a primary key.
1129
    @param cols [sql_gen.Col,...] The columns in the primary key.
1130
        Defaults to the first column in the table.
1131
    @pre The table must not already have a primary key.
1132
    '''
1133
    table = sql_gen.as_Table(table)
1134
    if cols == None: cols = [pkey(db, table, recover)]
1135
    col_strs = [sql_gen.to_name_only_col(v).to_str(db) for v in cols]
1136
1137
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ADD PRIMARY KEY ('
1138
        +(', '.join(col_strs))+')', recover=True, cacheable=True, log_level=3,
1139
        log_ignore_excs=(DuplicateException,))
1140
1141 2998 aaronmk
def add_index(db, exprs, table=None, unique=False, ensure_not_null_=True):
1142 2688 aaronmk
    '''Adds an index on column(s) or expression(s) if it doesn't already exist.
1143 2538 aaronmk
    Currently, only function calls are supported as expressions.
1144 2998 aaronmk
    @param ensure_not_null_ If set, translates NULL values to sentinel values.
1145 2847 aaronmk
        This allows indexes to be used for comparisons where NULLs are equal.
1146 2538 aaronmk
    '''
1147 2964 aaronmk
    exprs = lists.mk_seq(exprs)
1148 2538 aaronmk
1149 2688 aaronmk
    # Parse exprs
1150
    old_exprs = exprs[:]
1151
    exprs = []
1152
    cols = []
1153
    for i, expr in enumerate(old_exprs):
1154 2823 aaronmk
        expr = sql_gen.as_Col(expr, table)
1155 2688 aaronmk
1156 2823 aaronmk
        # Handle nullable columns
1157 2998 aaronmk
        if ensure_not_null_:
1158 3164 aaronmk
            try: expr = sql_gen.ensure_not_null(db, expr)
1159 2860 aaronmk
            except KeyError: pass # unknown type, so just create plain index
1160 2823 aaronmk
1161 2688 aaronmk
        # Extract col
1162 3002 aaronmk
        expr = copy.deepcopy(expr) # don't modify input!
1163 2688 aaronmk
        if isinstance(expr, sql_gen.FunctionCall):
1164
            col = expr.args[0]
1165
            expr = sql_gen.Expr(expr)
1166
        else: col = expr
1167 2823 aaronmk
        assert isinstance(col, sql_gen.Col)
1168 2688 aaronmk
1169
        # Extract table
1170
        if table == None:
1171
            assert sql_gen.is_table_col(col)
1172
            table = col.table
1173
1174
        col.table = None
1175
1176
        exprs.append(expr)
1177
        cols.append(col)
1178 2408 aaronmk
1179 2688 aaronmk
    table = sql_gen.as_Table(table)
1180
1181 3005 aaronmk
    # Add index
1182 3148 aaronmk
    str_ = 'CREATE'
1183
    if unique: str_ += ' UNIQUE'
1184
    str_ += ' INDEX ON '+table.to_str(db)+' ('+(
1185
        ', '.join((v.to_str(db) for v in exprs)))+')'
1186
    run_query(db, str_, recover=True, cacheable=True, log_level=3)
1187 2408 aaronmk
1188 3083 aaronmk
already_indexed = object() # tells add_indexes() the pkey has already been added
1189
1190
def add_indexes(db, table, has_pkey=True):
1191
    '''Adds an index on all columns in a table.
1192
    @param has_pkey bool|already_indexed Whether a pkey instead of a regular
1193
        index should be added on the first column.
1194
        * If already_indexed, the pkey is assumed to have already been added
1195
    '''
1196
    cols = table_cols(db, table)
1197
    if has_pkey:
1198
        if has_pkey is not already_indexed: add_pkey(db, table)
1199
        cols = cols[1:]
1200
    for col in cols: add_index(db, col, table)
1201
1202 3079 aaronmk
#### Tables
1203 2772 aaronmk
1204 3079 aaronmk
### Maintenance
1205 2772 aaronmk
1206 3079 aaronmk
def analyze(db, table):
1207
    table = sql_gen.as_Table(table)
1208
    run_query(db, 'ANALYZE '+table.to_str(db), log_level=3)
1209 2934 aaronmk
1210 3079 aaronmk
def autoanalyze(db, table):
1211
    if db.autoanalyze: analyze(db, table)
1212 2935 aaronmk
1213 3079 aaronmk
def vacuum(db, table):
1214
    table = sql_gen.as_Table(table)
1215
    db.with_autocommit(lambda: run_query(db, 'VACUUM ANALYZE '+table.to_str(db),
1216
        log_level=3))
1217 2086 aaronmk
1218 3079 aaronmk
### Lifecycle
1219
1220 3247 aaronmk
def drop(db, type_, name):
1221
    name = sql_gen.as_Name(name)
1222
    run_query(db, 'DROP '+type_+' IF EXISTS '+name.to_str(db)+' CASCADE')
1223 2889 aaronmk
1224 3247 aaronmk
def drop_table(db, table): drop(db, 'TABLE', table)
1225
1226 3082 aaronmk
def create_table(db, table, cols=[], has_pkey=True, col_indexes=True,
1227
    like=None):
1228 2675 aaronmk
    '''Creates a table.
1229 2681 aaronmk
    @param cols [sql_gen.TypedCol,...] The column names and types
1230
    @param has_pkey If set, the first column becomes the primary key.
1231 2760 aaronmk
    @param col_indexes bool|[ref]
1232
        * If True, indexes will be added on all non-pkey columns.
1233
        * If a list reference, [0] will be set to a function to do this.
1234
          This can be used to delay index creation until the table is populated.
1235 2675 aaronmk
    '''
1236
    table = sql_gen.as_Table(table)
1237
1238 3082 aaronmk
    if like != None:
1239
        cols = [sql_gen.CustomCode('LIKE '+like.to_str(db)+' INCLUDING ALL')
1240
            ]+cols
1241 2681 aaronmk
    if has_pkey:
1242
        cols[0] = pkey = copy.copy(cols[0]) # don't modify input!
1243 2872 aaronmk
        pkey.constraints = 'PRIMARY KEY'
1244 2681 aaronmk
1245 3085 aaronmk
    temp = table.is_temp and not db.debug_temp
1246
        # temp tables permanent in debug_temp mode
1247 2760 aaronmk
1248 3085 aaronmk
    # Create table
1249
    while True:
1250
        str_ = 'CREATE'
1251
        if temp: str_ += ' TEMP'
1252
        str_ += ' TABLE '+table.to_str(db)+' (\n'
1253
        str_ += '\n, '.join(c.to_str(db) for c in cols)
1254 3126 aaronmk
        str_ += '\n);'
1255 3085 aaronmk
1256
        try:
1257 3127 aaronmk
            run_query(db, str_, recover=True, cacheable=True, log_level=2,
1258 3085 aaronmk
                log_ignore_excs=(DuplicateException,))
1259
            break
1260
        except DuplicateException:
1261
            table.name = next_version(table.name)
1262
            # try again with next version of name
1263
1264 2760 aaronmk
    # Add indexes
1265 2773 aaronmk
    if has_pkey: has_pkey = already_indexed
1266
    def add_indexes_(): add_indexes(db, table, has_pkey)
1267
    if isinstance(col_indexes, list): col_indexes[0] = add_indexes_ # defer
1268
    elif col_indexes: add_indexes_() # add now
1269 2675 aaronmk
1270 3084 aaronmk
def copy_table_struct(db, src, dest):
1271
    '''Creates a structure-only copy of a table. (Does not copy data.)'''
1272 3085 aaronmk
    create_table(db, dest, has_pkey=False, col_indexes=False, like=src)
1273 3084 aaronmk
1274 3079 aaronmk
### Data
1275 2684 aaronmk
1276 2970 aaronmk
def truncate(db, table, schema='public', **kw_args):
1277
    '''For params, see run_query()'''
1278 2777 aaronmk
    table = sql_gen.as_Table(table, schema)
1279 2970 aaronmk
    return run_query(db, 'TRUNCATE '+table.to_str(db)+' CASCADE', **kw_args)
1280 2732 aaronmk
1281 2965 aaronmk
def empty_temp(db, tables):
1282
    tables = lists.mk_seq(tables)
1283 2971 aaronmk
    for table in tables: truncate(db, table, log_level=3)
1284 2965 aaronmk
1285 1968 aaronmk
def empty_db(db, schema='public', **kw_args):
1286
    '''For kw_args, see tables()'''
1287
    for table in tables(db, schema, **kw_args): truncate(db, table, schema)
1288 3094 aaronmk
1289
def distinct_table(db, table, distinct_on):
1290
    '''Creates a copy of a temp table which is distinct on the given columns.
1291 3099 aaronmk
    The old and new tables will both get an index on these columns, to
1292
    facilitate merge joins.
1293 3097 aaronmk
    @param distinct_on If empty, creates a table with one row. This is useful if
1294
        your distinct_on columns are all literal values.
1295 3099 aaronmk
    @return The new table.
1296 3094 aaronmk
    '''
1297 3099 aaronmk
    new_table = sql_gen.suffixed_table(table, '_distinct')
1298 3094 aaronmk
1299 3099 aaronmk
    copy_table_struct(db, table, new_table)
1300 3097 aaronmk
1301
    limit = None
1302
    if distinct_on == []: limit = 1 # one sample row
1303 3099 aaronmk
    else:
1304
        add_index(db, distinct_on, new_table, unique=True)
1305
        add_index(db, distinct_on, table) # for join optimization
1306 3097 aaronmk
1307 3099 aaronmk
    insert_select(db, new_table, None, mk_select(db, table, start=0,
1308 3097 aaronmk
        limit=limit), ignore=True)
1309 3099 aaronmk
    analyze(db, new_table)
1310 3094 aaronmk
1311 3099 aaronmk
    return new_table