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 4491 aaronmk
        DbException.__init__(self, 'for name: '
41
            +strings.as_tt(strings.ustr(name)), cause)
42 2143 aaronmk
        self.name = name
43 360 aaronmk
44 3109 aaronmk
class ExceptionWithValue(DbException):
45
    def __init__(self, value, cause=None):
46 4492 aaronmk
        DbException.__init__(self, 'for value: '
47
            +strings.as_tt(strings.urepr(value)), cause)
48 2240 aaronmk
        self.value = value
49
50 2945 aaronmk
class ExceptionWithNameType(DbException):
51
    def __init__(self, type_, name, cause=None):
52 4491 aaronmk
        DbException.__init__(self, 'for type: '+strings.as_tt(strings.ustr(
53
            type_))+'; name: '+strings.as_tt(name), cause)
54 2945 aaronmk
        self.type = type_
55
        self.name = name
56
57 2306 aaronmk
class ConstraintException(DbException):
58 3345 aaronmk
    def __init__(self, name, cond, cols, cause=None):
59
        msg = 'Violated '+strings.as_tt(name)+' constraint'
60
        if cond != None: msg += ' with condition '+cond
61
        if cols != []: msg += ' on columns: '+strings.as_tt(', '.join(cols))
62
        DbException.__init__(self, msg, cause)
63 2306 aaronmk
        self.name = name
64 3345 aaronmk
        self.cond = cond
65 468 aaronmk
        self.cols = cols
66 11 aaronmk
67 2523 aaronmk
class MissingCastException(DbException):
68 4139 aaronmk
    def __init__(self, type_, col=None, cause=None):
69
        msg = 'Missing cast to type '+strings.as_tt(type_)
70
        if col != None: msg += ' on column: '+strings.as_tt(col)
71
        DbException.__init__(self, msg, cause)
72 2523 aaronmk
        self.type = type_
73
        self.col = col
74
75 2143 aaronmk
class NameException(DbException): pass
76
77 2306 aaronmk
class DuplicateKeyException(ConstraintException): pass
78 13 aaronmk
79 2306 aaronmk
class NullValueException(ConstraintException): pass
80 13 aaronmk
81 3346 aaronmk
class CheckException(ConstraintException): pass
82
83 3109 aaronmk
class InvalidValueException(ExceptionWithValue): pass
84 2239 aaronmk
85 2945 aaronmk
class DuplicateException(ExceptionWithNameType): pass
86 2143 aaronmk
87 3419 aaronmk
class DoesNotExistException(ExceptionWithNameType): pass
88
89 89 aaronmk
class EmptyRowException(DbException): pass
90
91 865 aaronmk
##### Warnings
92
93
class DbWarning(UserWarning): pass
94
95 1930 aaronmk
##### Result retrieval
96
97
def col_names(cur): return (col[0] for col in cur.description)
98
99
def rows(cur): return iter(lambda: cur.fetchone(), None)
100
101
def consume_rows(cur):
102
    '''Used to fetch all rows so result will be cached'''
103
    iters.consume_iter(rows(cur))
104
105
def next_row(cur): return rows(cur).next()
106
107
def row(cur):
108
    row_ = next_row(cur)
109
    consume_rows(cur)
110
    return row_
111
112
def next_value(cur): return next_row(cur)[0]
113
114
def value(cur): return row(cur)[0]
115
116
def values(cur): return iters.func_iter(lambda: next_value(cur))
117
118
def value_or_none(cur):
119
    try: return value(cur)
120
    except StopIteration: return None
121
122 2762 aaronmk
##### Escaping
123 2101 aaronmk
124 2573 aaronmk
def esc_name_by_module(module, name):
125
    if module == 'psycopg2' or module == None: quote = '"'
126 2101 aaronmk
    elif module == 'MySQLdb': quote = '`'
127
    else: raise NotImplementedError("Can't escape name for "+module+' database')
128 2500 aaronmk
    return sql_gen.esc_name(name, quote)
129 2101 aaronmk
130
def esc_name_by_engine(engine, name, **kw_args):
131
    return esc_name_by_module(db_engines[engine][0], name, **kw_args)
132
133
def esc_name(db, name, **kw_args):
134
    return esc_name_by_module(util.root_module(db.db), name, **kw_args)
135
136
def qual_name(db, schema, table):
137
    def esc_name_(name): return esc_name(db, name)
138
    table = esc_name_(table)
139
    if schema != None: return esc_name_(schema)+'.'+table
140
    else: return table
141
142 1869 aaronmk
##### Database connections
143 1849 aaronmk
144 2097 aaronmk
db_config_names = ['engine', 'host', 'user', 'password', 'database', 'schemas']
145 1926 aaronmk
146 1869 aaronmk
db_engines = {
147
    'MySQL': ('MySQLdb', {'password': 'passwd', 'database': 'db'}),
148
    'PostgreSQL': ('psycopg2', {}),
149
}
150
151
DatabaseErrors_set = set([DbException])
152
DatabaseErrors = tuple(DatabaseErrors_set)
153
154
def _add_module(module):
155
    DatabaseErrors_set.add(module.DatabaseError)
156
    global DatabaseErrors
157
    DatabaseErrors = tuple(DatabaseErrors_set)
158
159
def db_config_str(db_config):
160
    return db_config['engine']+' database '+db_config['database']
161
162 2448 aaronmk
log_debug_none = lambda msg, level=2: None
163 1901 aaronmk
164 1849 aaronmk
class DbConn:
165 2923 aaronmk
    def __init__(self, db_config, autocommit=True, caching=True,
166 3183 aaronmk
        log_debug=log_debug_none, debug_temp=False, src=None):
167 2915 aaronmk
        '''
168
        @param debug_temp Whether temporary objects should instead be permanent.
169
            This assists in debugging the internal objects used by the program.
170 3183 aaronmk
        @param src In autocommit mode, will be included in a comment in every
171
            query, to help identify the data source in pg_stat_activity.
172 2915 aaronmk
        '''
173 1869 aaronmk
        self.db_config = db_config
174 2190 aaronmk
        self.autocommit = autocommit
175
        self.caching = caching
176 1901 aaronmk
        self.log_debug = log_debug
177 2193 aaronmk
        self.debug = log_debug != log_debug_none
178 2915 aaronmk
        self.debug_temp = debug_temp
179 3183 aaronmk
        self.src = src
180 3074 aaronmk
        self.autoanalyze = False
181 3269 aaronmk
        self.autoexplain = False
182
        self.profile_row_ct = None
183 1869 aaronmk
184 3124 aaronmk
        self._savepoint = 0
185 3120 aaronmk
        self._reset()
186 1869 aaronmk
187
    def __getattr__(self, name):
188
        if name == '__dict__': raise Exception('getting __dict__')
189
        if name == 'db': return self._db()
190
        else: raise AttributeError()
191
192
    def __getstate__(self):
193
        state = copy.copy(self.__dict__) # shallow copy
194 1915 aaronmk
        state['log_debug'] = None # don't pickle the debug callback
195 1869 aaronmk
        state['_DbConn__db'] = None # don't pickle the connection
196
        return state
197
198 3118 aaronmk
    def clear_cache(self): self.query_results = {}
199
200 3120 aaronmk
    def _reset(self):
201 3118 aaronmk
        self.clear_cache()
202 3124 aaronmk
        assert self._savepoint == 0
203 3118 aaronmk
        self._notices_seen = set()
204
        self.__db = None
205
206 2165 aaronmk
    def connected(self): return self.__db != None
207
208 3116 aaronmk
    def close(self):
209 3119 aaronmk
        if not self.connected(): return
210
211 3135 aaronmk
        # Record that the automatic transaction is now closed
212 3136 aaronmk
        self._savepoint -= 1
213 3135 aaronmk
214 3119 aaronmk
        self.db.close()
215 3120 aaronmk
        self._reset()
216 3116 aaronmk
217 3125 aaronmk
    def reconnect(self):
218
        # Do not do this in test mode as it would roll back everything
219
        if self.autocommit: self.close()
220
        # Connection will be reopened automatically on first query
221
222 1869 aaronmk
    def _db(self):
223
        if self.__db == None:
224
            # Process db_config
225
            db_config = self.db_config.copy() # don't modify input!
226 2097 aaronmk
            schemas = db_config.pop('schemas', None)
227 1869 aaronmk
            module_name, mappings = db_engines[db_config.pop('engine')]
228
            module = __import__(module_name)
229
            _add_module(module)
230
            for orig, new in mappings.iteritems():
231
                try: util.rename_key(db_config, orig, new)
232
                except KeyError: pass
233
234
            # Connect
235
            self.__db = module.connect(**db_config)
236
237 3161 aaronmk
            # Record that a transaction is already open
238
            self._savepoint += 1
239
240 1869 aaronmk
            # Configure connection
241 2906 aaronmk
            if hasattr(self.db, 'set_isolation_level'):
242
                import psycopg2.extensions
243
                self.db.set_isolation_level(
244
                    psycopg2.extensions.ISOLATION_LEVEL_READ_COMMITTED)
245 2101 aaronmk
            if schemas != None:
246 2893 aaronmk
                search_path = [self.esc_name(s) for s in schemas.split(',')]
247
                search_path.append(value(run_query(self, 'SHOW search_path',
248
                    log_level=4)))
249
                run_query(self, 'SET search_path TO '+(','.join(search_path)),
250
                    log_level=3)
251 1869 aaronmk
252
        return self.__db
253 1889 aaronmk
254 1891 aaronmk
    class DbCursor(Proxy):
255 1927 aaronmk
        def __init__(self, outer):
256 1891 aaronmk
            Proxy.__init__(self, outer.db.cursor())
257 2191 aaronmk
            self.outer = outer
258 1927 aaronmk
            self.query_results = outer.query_results
259 1894 aaronmk
            self.query_lookup = None
260 1891 aaronmk
            self.result = []
261 1889 aaronmk
262 2802 aaronmk
        def execute(self, query):
263 2764 aaronmk
            self._is_insert = query.startswith('INSERT')
264 2797 aaronmk
            self.query_lookup = query
265 2148 aaronmk
            try:
266 3162 aaronmk
                try: cur = self.inner.execute(query)
267 2802 aaronmk
                finally: self.query = get_cur_query(self.inner, query)
268 1904 aaronmk
            except Exception, e:
269
                self.result = e # cache the exception as the result
270
                self._cache_result()
271
                raise
272 3004 aaronmk
273
            # Always cache certain queries
274 3183 aaronmk
            query = sql_gen.lstrip(query)
275 3004 aaronmk
            if query.startswith('CREATE') or query.startswith('ALTER'):
276 3007 aaronmk
                # structural changes
277 3040 aaronmk
                # Rest of query must be unique in the face of name collisions,
278
                # so don't cache ADD COLUMN unless it has distinguishing comment
279
                if query.find('ADD COLUMN') < 0 or query.endswith('*/'):
280 3007 aaronmk
                    self._cache_result()
281 3004 aaronmk
            elif self.rowcount == 0 and query.startswith('SELECT'): # empty
282 2800 aaronmk
                consume_rows(self) # fetch all rows so result will be cached
283 3004 aaronmk
284 2762 aaronmk
            return cur
285 1894 aaronmk
286 1891 aaronmk
        def fetchone(self):
287
            row = self.inner.fetchone()
288 1899 aaronmk
            if row != None: self.result.append(row)
289
            # otherwise, fetched all rows
290 1904 aaronmk
            else: self._cache_result()
291
            return row
292
293
        def _cache_result(self):
294 2948 aaronmk
            # For inserts that return a result set, don't cache result set since
295
            # inserts are not idempotent. Other non-SELECT queries don't have
296
            # their result set read, so only exceptions will be cached (an
297
            # invalid query will always be invalid).
298 1930 aaronmk
            if self.query_results != None and (not self._is_insert
299 1906 aaronmk
                or isinstance(self.result, Exception)):
300
301 1894 aaronmk
                assert self.query_lookup != None
302 1916 aaronmk
                self.query_results[self.query_lookup] = self.CacheCursor(
303
                    util.dict_subset(dicts.AttrsDictView(self),
304
                    ['query', 'result', 'rowcount', 'description']))
305 1906 aaronmk
306 1916 aaronmk
        class CacheCursor:
307
            def __init__(self, cached_result): self.__dict__ = cached_result
308
309 1927 aaronmk
            def execute(self, *args, **kw_args):
310 1916 aaronmk
                if isinstance(self.result, Exception): raise self.result
311
                # otherwise, result is a rows list
312
                self.iter = iter(self.result)
313
314
            def fetchone(self):
315
                try: return self.iter.next()
316
                except StopIteration: return None
317 1891 aaronmk
318 2212 aaronmk
    def esc_value(self, value):
319 2663 aaronmk
        try: str_ = self.mogrify('%s', [value])
320
        except NotImplementedError, e:
321
            module = util.root_module(self.db)
322
            if module == 'MySQLdb':
323
                import _mysql
324
                str_ = _mysql.escape_string(value)
325
            else: raise e
326 2374 aaronmk
        return strings.to_unicode(str_)
327 2212 aaronmk
328 2347 aaronmk
    def esc_name(self, name): return esc_name(self, name) # calls global func
329
330 2814 aaronmk
    def std_code(self, str_):
331
        '''Standardizes SQL code.
332
        * Ensures that string literals are prefixed by `E`
333
        '''
334
        if str_.startswith("'"): str_ = 'E'+str_
335
        return str_
336
337 2665 aaronmk
    def can_mogrify(self):
338 2663 aaronmk
        module = util.root_module(self.db)
339 2665 aaronmk
        return module == 'psycopg2'
340 2663 aaronmk
341 2665 aaronmk
    def mogrify(self, query, params=None):
342
        if self.can_mogrify(): return self.db.cursor().mogrify(query, params)
343
        else: raise NotImplementedError("Can't mogrify query")
344
345 2671 aaronmk
    def print_notices(self):
346 2725 aaronmk
        if hasattr(self.db, 'notices'):
347
            for msg in self.db.notices:
348
                if msg not in self._notices_seen:
349
                    self._notices_seen.add(msg)
350
                    self.log_debug(msg, level=2)
351 2671 aaronmk
352 2793 aaronmk
    def run_query(self, query, cacheable=False, log_level=2,
353 2464 aaronmk
        debug_msg_ref=None):
354 2445 aaronmk
        '''
355 2464 aaronmk
        @param log_ignore_excs The log_level will be increased by 2 if the query
356
            throws one of these exceptions.
357 2664 aaronmk
        @param debug_msg_ref If specified, the log message will be returned in
358
            this instead of being output. This allows you to filter log messages
359
            depending on the result of the query.
360 2445 aaronmk
        '''
361 2167 aaronmk
        assert query != None
362
363 3183 aaronmk
        if self.autocommit and self.src != None:
364 3206 aaronmk
            query = sql_gen.esc_comment(self.src)+'\t'+query
365 3183 aaronmk
366 2047 aaronmk
        if not self.caching: cacheable = False
367 1903 aaronmk
        used_cache = False
368 2664 aaronmk
369 3242 aaronmk
        if self.debug:
370
            profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
371 1903 aaronmk
        try:
372 1927 aaronmk
            # Get cursor
373
            if cacheable:
374 3238 aaronmk
                try: cur = self.query_results[query]
375 1927 aaronmk
                except KeyError: cur = self.DbCursor(self)
376 3238 aaronmk
                else: used_cache = True
377 1927 aaronmk
            else: cur = self.db.cursor()
378
379
            # Run query
380 3238 aaronmk
            try: cur.execute(query)
381 3162 aaronmk
            except Exception, e:
382
                _add_cursor_info(e, self, query)
383
                raise
384 3238 aaronmk
            else: self.do_autocommit()
385 1903 aaronmk
        finally:
386 3242 aaronmk
            if self.debug:
387 3244 aaronmk
                profiler.stop(self.profile_row_ct)
388 3242 aaronmk
389
                ## Log or return query
390
391 4491 aaronmk
                query = strings.ustr(get_cur_query(cur, query))
392 3281 aaronmk
                # Put the src comment on a separate line in the log file
393
                query = query.replace('\t', '\n', 1)
394 3239 aaronmk
395 3240 aaronmk
                msg = 'DB query: '
396 3239 aaronmk
397 3240 aaronmk
                if used_cache: msg += 'cache hit'
398
                elif cacheable: msg += 'cache miss'
399
                else: msg += 'non-cacheable'
400 3239 aaronmk
401 3241 aaronmk
                msg += ':\n'+profiler.msg()+'\n'+strings.as_code(query, 'SQL')
402 3240 aaronmk
403 3237 aaronmk
                if debug_msg_ref != None: debug_msg_ref[0] = msg
404
                else: self.log_debug(msg, log_level)
405 3245 aaronmk
406
                self.print_notices()
407 1903 aaronmk
408
        return cur
409 1914 aaronmk
410 2797 aaronmk
    def is_cached(self, query): return query in self.query_results
411 2139 aaronmk
412 2907 aaronmk
    def with_autocommit(self, func):
413 2801 aaronmk
        import psycopg2.extensions
414
415
        prev_isolation_level = self.db.isolation_level
416 2907 aaronmk
        self.db.set_isolation_level(
417
            psycopg2.extensions.ISOLATION_LEVEL_AUTOCOMMIT)
418 2683 aaronmk
        try: return func()
419 2801 aaronmk
        finally: self.db.set_isolation_level(prev_isolation_level)
420 2683 aaronmk
421 2139 aaronmk
    def with_savepoint(self, func):
422 3137 aaronmk
        top = self._savepoint == 0
423 2171 aaronmk
        savepoint = 'level_'+str(self._savepoint)
424 3137 aaronmk
425 3272 aaronmk
        if self.debug:
426 3273 aaronmk
            self.log_debug('Begin transaction', level=4)
427 3272 aaronmk
            profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
428
429 3160 aaronmk
        # Must happen before running queries so they don't get autocommitted
430
        self._savepoint += 1
431
432 3137 aaronmk
        if top: query = 'START TRANSACTION ISOLATION LEVEL READ COMMITTED'
433
        else: query = 'SAVEPOINT '+savepoint
434
        self.run_query(query, log_level=4)
435
        try:
436
            return func()
437
            if top: self.run_query('COMMIT', log_level=4)
438 2139 aaronmk
        except:
439 3137 aaronmk
            if top: query = 'ROLLBACK'
440
            else: query = 'ROLLBACK TO SAVEPOINT '+savepoint
441
            self.run_query(query, log_level=4)
442
443 2139 aaronmk
            raise
444 2930 aaronmk
        finally:
445
            # Always release savepoint, because after ROLLBACK TO SAVEPOINT,
446
            # "The savepoint remains valid and can be rolled back to again"
447
            # (http://www.postgresql.org/docs/8.3/static/sql-rollback-to.html).
448 3137 aaronmk
            if not top:
449
                self.run_query('RELEASE SAVEPOINT '+savepoint, log_level=4)
450 2930 aaronmk
451
            self._savepoint -= 1
452
            assert self._savepoint >= 0
453
454 3272 aaronmk
            if self.debug:
455
                profiler.stop(self.profile_row_ct)
456 3273 aaronmk
                self.log_debug('End transaction\n'+profiler.msg(), level=4)
457 3272 aaronmk
458 2930 aaronmk
            self.do_autocommit() # OK to do this after ROLLBACK TO SAVEPOINT
459 2191 aaronmk
460
    def do_autocommit(self):
461
        '''Autocommits if outside savepoint'''
462 3135 aaronmk
        assert self._savepoint >= 1
463
        if self.autocommit and self._savepoint == 1:
464 2924 aaronmk
            self.log_debug('Autocommitting', level=4)
465 2191 aaronmk
            self.db.commit()
466 2643 aaronmk
467 3155 aaronmk
    def col_info(self, col, cacheable=True):
468 2643 aaronmk
        table = sql_gen.Table('columns', 'information_schema')
469 3063 aaronmk
        type_ = sql_gen.Coalesce(sql_gen.Nullif(sql_gen.Col('data_type'),
470
            'USER-DEFINED'), sql_gen.Col('udt_name'))
471 3078 aaronmk
        cols = [type_, 'column_default',
472
            sql_gen.Cast('boolean', sql_gen.Col('is_nullable'))]
473 2643 aaronmk
474 3750 aaronmk
        conds = [('table_name', col.table.name),
475
            ('column_name', strings.ustr(col.name))]
476 2643 aaronmk
        schema = col.table.schema
477
        if schema != None: conds.append(('table_schema', schema))
478
479 3638 aaronmk
        cur = select(self, table, cols, conds, order_by='table_schema', limit=1,
480
            cacheable=cacheable, log_level=4) # TODO: order by search_path order
481
        try: type_, default, nullable = row(cur)
482 4114 aaronmk
        except StopIteration: raise sql_gen.NoUnderlyingTableException(col)
483 2819 aaronmk
        default = sql_gen.as_Code(default, self)
484
485
        return sql_gen.TypedCol(col.name, type_, default, nullable)
486 2917 aaronmk
487
    def TempFunction(self, name):
488
        if self.debug_temp: schema = None
489
        else: schema = 'pg_temp'
490
        return sql_gen.Function(name, schema)
491 1849 aaronmk
492 1869 aaronmk
connect = DbConn
493
494 832 aaronmk
##### Recoverable querying
495 15 aaronmk
496 2139 aaronmk
def with_savepoint(db, func): return db.with_savepoint(func)
497 11 aaronmk
498 2791 aaronmk
def run_query(db, query, recover=None, cacheable=False, log_level=2,
499
    log_ignore_excs=None, **kw_args):
500 2794 aaronmk
    '''For params, see DbConn.run_query()'''
501 830 aaronmk
    if recover == None: recover = False
502 2464 aaronmk
    if log_ignore_excs == None: log_ignore_excs = ()
503
    log_ignore_excs = tuple(log_ignore_excs)
504 3236 aaronmk
    debug_msg_ref = [None]
505 830 aaronmk
506 3267 aaronmk
    query = with_explain_comment(db, query)
507 3258 aaronmk
508 2148 aaronmk
    try:
509 2464 aaronmk
        try:
510 2794 aaronmk
            def run(): return db.run_query(query, cacheable, log_level,
511 2793 aaronmk
                debug_msg_ref, **kw_args)
512 2796 aaronmk
            if recover and not db.is_cached(query):
513 2464 aaronmk
                return with_savepoint(db, run)
514
            else: return run() # don't need savepoint if cached
515
        except Exception, e:
516 3095 aaronmk
            msg = strings.ustr(e.args[0])
517 4103 aaronmk
            msg = re.sub(r'^(?:PL/Python: )?ValueError: ', r'', msg)
518 2464 aaronmk
519 3095 aaronmk
            match = re.match(r'^duplicate key value violates unique constraint '
520 3338 aaronmk
                r'"(.+?)"', msg)
521 2464 aaronmk
            if match:
522 3338 aaronmk
                constraint, = match.groups()
523 3025 aaronmk
                cols = []
524
                if recover: # need auto-rollback to run index_cols()
525 3319 aaronmk
                    try: cols = index_cols(db, constraint)
526 3025 aaronmk
                    except NotImplementedError: pass
527 3345 aaronmk
                raise DuplicateKeyException(constraint, None, cols, e)
528 2464 aaronmk
529 3095 aaronmk
            match = re.match(r'^null value in column "(.+?)" violates not-null'
530 2464 aaronmk
                r' constraint', msg)
531 3345 aaronmk
            if match:
532
                col, = match.groups()
533
                raise NullValueException('NOT NULL', None, [col], e)
534 2464 aaronmk
535 3346 aaronmk
            match = re.match(r'^new row for relation "(.+?)" violates check '
536
                r'constraint "(.+?)"', msg)
537
            if match:
538
                table, constraint = match.groups()
539 3347 aaronmk
                constraint = sql_gen.Col(constraint, table)
540 3349 aaronmk
                cond = None
541
                if recover: # need auto-rollback to run constraint_cond()
542
                    try: cond = constraint_cond(db, constraint)
543
                    except NotImplementedError: pass
544
                raise CheckException(constraint.to_str(db), cond, [], e)
545 3346 aaronmk
546 3095 aaronmk
            match = re.match(r'^(?:invalid input (?:syntax|value)\b.*?'
547 3635 aaronmk
                r'|.+? out of range): "(.+?)"', msg)
548 2464 aaronmk
            if match:
549 3109 aaronmk
                value, = match.groups()
550
                raise InvalidValueException(strings.to_unicode(value), e)
551 2464 aaronmk
552 3095 aaronmk
            match = re.match(r'^column "(.+?)" is of type (.+?) but expression '
553 2523 aaronmk
                r'is of type', msg)
554
            if match:
555
                col, type_ = match.groups()
556
                raise MissingCastException(type_, col, e)
557
558 4141 aaronmk
            match = re.match(r'^could not determine polymorphic type because '
559
                r'input has type "unknown"', msg)
560
            if match: raise MissingCastException('text', None, e)
561
562 4485 aaronmk
            match = re.match(r'^.+? types .+? and .+? cannot be matched', msg)
563
            if match: raise MissingCastException('text', None, e)
564
565 4509 aaronmk
            typed_name_re = r'^(\S+) "(.+?)"(?: of relation ".+?")?'
566 3419 aaronmk
567
            match = re.match(typed_name_re+r'.*? already exists', msg)
568 2945 aaronmk
            if match:
569
                type_, name = match.groups()
570
                raise DuplicateException(type_, name, e)
571 2464 aaronmk
572 4145 aaronmk
            match = re.match(r'more than one (\S+) named ""(.+?)""', msg)
573
            if match:
574
                type_, name = match.groups()
575
                raise DuplicateException(type_, name, e)
576
577 3419 aaronmk
            match = re.match(typed_name_re+r' does not exist', msg)
578
            if match:
579
                type_, name = match.groups()
580
                raise DoesNotExistException(type_, name, e)
581
582 2464 aaronmk
            raise # no specific exception raised
583
    except log_ignore_excs:
584
        log_level += 2
585
        raise
586
    finally:
587 3236 aaronmk
        if debug_msg_ref[0] != None: db.log_debug(debug_msg_ref[0], log_level)
588 830 aaronmk
589 832 aaronmk
##### Basic queries
590
591 3256 aaronmk
def is_explainable(query):
592
    # See <http://www.postgresql.org/docs/8.3/static/sql-explain.html#AEN57749>
593 3257 aaronmk
    return re.match(r'^(?:SELECT|INSERT|UPDATE|DELETE|VALUES|EXECUTE|DECLARE)\b'
594
        , query)
595 3256 aaronmk
596 3263 aaronmk
def explain(db, query, **kw_args):
597
    '''
598
    For params, see run_query().
599
    '''
600 3267 aaronmk
    kw_args.setdefault('log_level', 4)
601 3263 aaronmk
602 3750 aaronmk
    return strings.ustr(strings.join_lines(values(run_query(db,
603
        'EXPLAIN '+query, recover=True, cacheable=True, **kw_args))))
604 3256 aaronmk
        # not a higher log_level because it's useful to see what query is being
605
        # run before it's executed, which EXPLAIN effectively provides
606
607 3265 aaronmk
def has_comment(query): return query.endswith('*/')
608
609
def with_explain_comment(db, query, **kw_args):
610 3269 aaronmk
    if db.autoexplain and not has_comment(query) and is_explainable(query):
611 3265 aaronmk
        query += '\n'+sql_gen.esc_comment(' EXPLAIN:\n'
612
            +explain(db, query, **kw_args))
613
    return query
614
615 2153 aaronmk
def next_version(name):
616 2163 aaronmk
    version = 1 # first existing name was version 0
617 2586 aaronmk
    match = re.match(r'^(.*)#(\d+)$', name)
618 2153 aaronmk
    if match:
619 2586 aaronmk
        name, version = match.groups()
620
        version = int(version)+1
621 2932 aaronmk
    return sql_gen.concat(name, '#'+str(version))
622 2153 aaronmk
623 2899 aaronmk
def lock_table(db, table, mode):
624
    table = sql_gen.as_Table(table)
625
    run_query(db, 'LOCK TABLE '+table.to_str(db)+' IN '+mode+' MODE')
626
627 3303 aaronmk
def run_query_into(db, query, into=None, add_pkey_=False, **kw_args):
628 2085 aaronmk
    '''Outputs a query to a temp table.
629
    For params, see run_query().
630
    '''
631 2789 aaronmk
    if into == None: return run_query(db, query, **kw_args)
632 2790 aaronmk
633
    assert isinstance(into, sql_gen.Table)
634
635 2992 aaronmk
    into.is_temp = True
636 3008 aaronmk
    # "temporary tables cannot specify a schema name", so remove schema
637
    into.schema = None
638 2992 aaronmk
639 2790 aaronmk
    kw_args['recover'] = True
640 2945 aaronmk
    kw_args.setdefault('log_ignore_excs', (DuplicateException,))
641 2790 aaronmk
642 2916 aaronmk
    temp = not db.debug_temp # tables are permanent in debug_temp mode
643 2790 aaronmk
644
    # Create table
645
    while True:
646
        create_query = 'CREATE'
647
        if temp: create_query += ' TEMP'
648
        create_query += ' TABLE '+into.to_str(db)+' AS\n'+query
649 2385 aaronmk
650 2790 aaronmk
        try:
651
            cur = run_query(db, create_query, **kw_args)
652
                # CREATE TABLE AS sets rowcount to # rows in query
653
            break
654 2945 aaronmk
        except DuplicateException, e:
655 2790 aaronmk
            into.name = next_version(into.name)
656
            # try again with next version of name
657
658 3303 aaronmk
    if add_pkey_: add_pkey(db, into)
659 3075 aaronmk
660
    # According to the PostgreSQL doc, "The autovacuum daemon cannot access and
661
    # therefore cannot vacuum or analyze temporary tables. [...] if a temporary
662
    # table is going to be used in complex queries, it is wise to run ANALYZE on
663
    # the temporary table after it is populated."
664
    # (http://www.postgresql.org/docs/9.1/static/sql-createtable.html)
665
    # If into is not a temp table, ANALYZE is useful but not required.
666 3073 aaronmk
    analyze(db, into)
667 2790 aaronmk
668
    return cur
669 2085 aaronmk
670 2120 aaronmk
order_by_pkey = object() # tells mk_select() to order by the pkey
671
672 2199 aaronmk
distinct_on_all = object() # tells mk_select() to SELECT DISTINCT ON all columns
673
674 3420 aaronmk
def mk_select(db, tables=None, fields=None, conds=None, distinct_on=[],
675 3494 aaronmk
    limit=None, start=None, order_by=order_by_pkey, default_table=None,
676
    explain=True):
677 1981 aaronmk
    '''
678 2121 aaronmk
    @param tables The single table to select from, or a list of tables to join
679 2280 aaronmk
        together, with tables after the first being sql_gen.Join objects
680 1981 aaronmk
    @param fields Use None to select all fields in the table
681 2377 aaronmk
    @param conds WHERE conditions: [(compare_left_side, compare_right_side),...]
682 2379 aaronmk
        * container can be any iterable type
683 2399 aaronmk
        * compare_left_side: sql_gen.Code|str (for col name)
684
        * compare_right_side: sql_gen.ValueCond|literal value
685 2199 aaronmk
    @param distinct_on The columns to SELECT DISTINCT ON, or distinct_on_all to
686
        use all columns
687 2786 aaronmk
    @return query
688 1981 aaronmk
    '''
689 2315 aaronmk
    # Parse tables param
690 2964 aaronmk
    tables = lists.mk_seq(tables)
691 2141 aaronmk
    tables = list(tables) # don't modify input! (list() copies input)
692 2315 aaronmk
    table0 = sql_gen.as_Table(tables.pop(0)) # first table is separate
693 2121 aaronmk
694 2315 aaronmk
    # Parse other params
695 2376 aaronmk
    if conds == None: conds = []
696 2650 aaronmk
    elif dicts.is_dict(conds): conds = conds.items()
697 2379 aaronmk
    conds = list(conds) # don't modify input! (list() copies input)
698 3129 aaronmk
    assert limit == None or isinstance(limit, (int, long))
699
    assert start == None or isinstance(start, (int, long))
700 2315 aaronmk
    if order_by is order_by_pkey:
701 3421 aaronmk
        if table0 == None or distinct_on != []: order_by = None
702 2315 aaronmk
        else: order_by = pkey(db, table0, recover=True)
703 865 aaronmk
704 2315 aaronmk
    query = 'SELECT'
705 2056 aaronmk
706 2315 aaronmk
    def parse_col(col): return sql_gen.as_Col(col, default_table).to_str(db)
707 2056 aaronmk
708 2200 aaronmk
    # DISTINCT ON columns
709 2233 aaronmk
    if distinct_on != []:
710 2467 aaronmk
        query += '\nDISTINCT'
711 2254 aaronmk
        if distinct_on is not distinct_on_all:
712 2200 aaronmk
            query += ' ON ('+(', '.join(map(parse_col, distinct_on)))+')'
713
714
    # Columns
715 3185 aaronmk
    if query.find('\n') >= 0: whitespace = '\n'
716
    else: whitespace = ' '
717
    if fields == None: query += whitespace+'*'
718 2765 aaronmk
    else:
719
        assert fields != []
720 3185 aaronmk
        if len(fields) > 1: whitespace = '\n'
721
        query += whitespace+('\n, '.join(map(parse_col, fields)))
722 2200 aaronmk
723
    # Main table
724 3185 aaronmk
    if query.find('\n') >= 0 or len(tables) > 0: whitespace = '\n'
725
    else: whitespace = ' '
726 3420 aaronmk
    if table0 != None: query += whitespace+'FROM '+table0.to_str(db)
727 865 aaronmk
728 2122 aaronmk
    # Add joins
729 2271 aaronmk
    left_table = table0
730 2263 aaronmk
    for join_ in tables:
731
        table = join_.table
732 2238 aaronmk
733 2343 aaronmk
        # Parse special values
734
        if join_.type_ is sql_gen.filter_out: # filter no match
735 2376 aaronmk
            conds.append((sql_gen.Col(table_not_null_col(db, table), table),
736 2853 aaronmk
                sql_gen.CompareCond(None, '~=')))
737 2343 aaronmk
738 2467 aaronmk
        query += '\n'+join_.to_str(db, left_table)
739 2122 aaronmk
740
        left_table = table
741
742 865 aaronmk
    missing = True
743 2376 aaronmk
    if conds != []:
744 2576 aaronmk
        if len(conds) == 1: whitespace = ' '
745
        else: whitespace = '\n'
746 2578 aaronmk
        query += '\n'+sql_gen.combine_conds([sql_gen.ColValueCond(l, r)
747
            .to_str(db) for l, r in conds], 'WHERE')
748 2227 aaronmk
    if order_by != None:
749 2467 aaronmk
        query += '\nORDER BY '+sql_gen.as_Col(order_by, table0).to_str(db)
750 3297 aaronmk
    if limit != None: query += '\nLIMIT '+str(limit)
751 865 aaronmk
    if start != None:
752 2467 aaronmk
        if start != 0: query += '\nOFFSET '+str(start)
753 865 aaronmk
754 3494 aaronmk
    if explain: query = with_explain_comment(db, query)
755 3266 aaronmk
756 2786 aaronmk
    return query
757 11 aaronmk
758 2054 aaronmk
def select(db, *args, **kw_args):
759
    '''For params, see mk_select() and run_query()'''
760
    recover = kw_args.pop('recover', None)
761
    cacheable = kw_args.pop('cacheable', True)
762 2442 aaronmk
    log_level = kw_args.pop('log_level', 2)
763 2054 aaronmk
764 2791 aaronmk
    return run_query(db, mk_select(db, *args, **kw_args), recover, cacheable,
765
        log_level=log_level)
766 2054 aaronmk
767 2788 aaronmk
def mk_insert_select(db, table, cols=None, select_query=None, returning=None,
768 3181 aaronmk
    embeddable=False, ignore=False, src=None):
769 1960 aaronmk
    '''
770
    @param returning str|None An inserted column (such as pkey) to return
771 2070 aaronmk
    @param embeddable Whether the query should be embeddable as a nested SELECT.
772 2073 aaronmk
        Warning: If you set this and cacheable=True when the query is run, the
773
        query will be fully cached, not just if it raises an exception.
774 3009 aaronmk
    @param ignore Whether to ignore duplicate keys.
775 3181 aaronmk
    @param src Will be included in the name of any created function, to help
776
        identify the data source in pg_stat_activity.
777 1960 aaronmk
    '''
778 2754 aaronmk
    table = sql_gen.remove_table_rename(sql_gen.as_Table(table))
779 2318 aaronmk
    if cols == []: cols = None # no cols (all defaults) = unknown col names
780 3010 aaronmk
    if cols != None: cols = [sql_gen.to_name_only_col(c, table) for c in cols]
781 2063 aaronmk
    if select_query == None: select_query = 'DEFAULT VALUES'
782 2327 aaronmk
    if returning != None: returning = sql_gen.as_Col(returning, table)
783 2063 aaronmk
784 2497 aaronmk
    first_line = 'INSERT INTO '+table.to_str(db)
785 2063 aaronmk
786 3009 aaronmk
    def mk_insert(select_query):
787
        query = first_line
788 3014 aaronmk
        if cols != None:
789
            query += '\n('+(', '.join((c.to_str(db) for c in cols)))+')'
790 3009 aaronmk
        query += '\n'+select_query
791
792
        if returning != None:
793
            returning_name_col = sql_gen.to_name_only_col(returning)
794
            query += '\nRETURNING '+returning_name_col.to_str(db)
795
796
        return query
797 2063 aaronmk
798 3489 aaronmk
    return_type = sql_gen.CustomCode('unknown')
799
    if returning != None: return_type = sql_gen.ColType(returning)
800 3017 aaronmk
801 3009 aaronmk
    if ignore:
802 3017 aaronmk
        # Always return something to set the correct rowcount
803
        if returning == None: returning = sql_gen.NamedCol('NULL', None)
804
805 3009 aaronmk
        embeddable = True # must use function
806 3010 aaronmk
807 3450 aaronmk
        if cols == None: row = [sql_gen.Col(sql_gen.all_cols, 'row')]
808
        else: row = [sql_gen.Col(c.name, 'row') for c in cols]
809 3092 aaronmk
810 3484 aaronmk
        query = sql_gen.RowExcIgnore(sql_gen.RowType(table), select_query,
811 3497 aaronmk
            sql_gen.ReturnQuery(mk_insert(sql_gen.Values(row).to_str(db))),
812
            cols)
813 3009 aaronmk
    else: query = mk_insert(select_query)
814
815 2070 aaronmk
    if embeddable:
816
        # Create function
817 2513 aaronmk
        function_name = sql_gen.clean_name(first_line)
818 3181 aaronmk
        if src != None: function_name = src+': '+function_name
819 2189 aaronmk
        while True:
820
            try:
821 3451 aaronmk
                func = db.TempFunction(function_name)
822 3489 aaronmk
                def_ = sql_gen.FunctionDef(func, sql_gen.SetOf(return_type),
823
                    query)
824 2194 aaronmk
825 3443 aaronmk
                run_query(db, def_.to_str(db), recover=True, cacheable=True,
826 2945 aaronmk
                    log_ignore_excs=(DuplicateException,))
827 2189 aaronmk
                break # this version was successful
828 2945 aaronmk
            except DuplicateException, e:
829 2189 aaronmk
                function_name = next_version(function_name)
830
                # try again with next version of name
831 2070 aaronmk
832 2337 aaronmk
        # Return query that uses function
833 3009 aaronmk
        cols = None
834
        if returning != None: cols = [returning]
835 3451 aaronmk
        func_table = sql_gen.NamedTable('f', sql_gen.FunctionCall(func), cols)
836
            # AS clause requires function alias
837 3298 aaronmk
        return mk_select(db, func_table, order_by=None)
838 2070 aaronmk
839 2787 aaronmk
    return query
840 2066 aaronmk
841 3074 aaronmk
def insert_select(db, table, *args, **kw_args):
842 2085 aaronmk
    '''For params, see mk_insert_select() and run_query_into()
843 2386 aaronmk
    @param into sql_gen.Table with suggested name of temp table to put RETURNING
844
        values in
845 2072 aaronmk
    '''
846 3141 aaronmk
    returning = kw_args.get('returning', None)
847
    ignore = kw_args.get('ignore', False)
848
849 2386 aaronmk
    into = kw_args.pop('into', None)
850
    if into != None: kw_args['embeddable'] = True
851 2066 aaronmk
    recover = kw_args.pop('recover', None)
852 3141 aaronmk
    if ignore: recover = True
853 2066 aaronmk
    cacheable = kw_args.pop('cacheable', True)
854 2673 aaronmk
    log_level = kw_args.pop('log_level', 2)
855 2066 aaronmk
856 3141 aaronmk
    rowcount_only = ignore and returning == None # keep NULL rows on server
857
    if rowcount_only: into = sql_gen.Table('rowcount')
858
859 3074 aaronmk
    cur = run_query_into(db, mk_insert_select(db, table, *args, **kw_args),
860
        into, recover=recover, cacheable=cacheable, log_level=log_level)
861 3141 aaronmk
    if rowcount_only: empty_temp(db, into)
862 3074 aaronmk
    autoanalyze(db, table)
863
    return cur
864 2063 aaronmk
865 2738 aaronmk
default = sql_gen.default # tells insert() to use the default value for a column
866 2066 aaronmk
867 2063 aaronmk
def insert(db, table, row, *args, **kw_args):
868 2085 aaronmk
    '''For params, see insert_select()'''
869 1960 aaronmk
    if lists.is_seq(row): cols = None
870
    else:
871
        cols = row.keys()
872
        row = row.values()
873 2738 aaronmk
    row = list(row) # ensure that "== []" works
874 1960 aaronmk
875 2738 aaronmk
    if row == []: query = None
876
    else: query = sql_gen.Values(row).to_str(db)
877 1961 aaronmk
878 2788 aaronmk
    return insert_select(db, table, cols, query, *args, **kw_args)
879 11 aaronmk
880 3152 aaronmk
def mk_update(db, table, changes=None, cond=None, in_place=False,
881 3153 aaronmk
    cacheable_=True):
882 2402 aaronmk
    '''
883
    @param changes [(col, new_value),...]
884
        * container can be any iterable type
885
        * col: sql_gen.Code|str (for col name)
886
        * new_value: sql_gen.Code|literal value
887
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
888 3056 aaronmk
    @param in_place If set, locks the table and updates rows in place.
889
        This avoids creating dead rows in PostgreSQL.
890
        * cond must be None
891 3153 aaronmk
    @param cacheable_ Whether column structure information used to generate the
892 3152 aaronmk
        query can be cached
893 2402 aaronmk
    @return str query
894
    '''
895 3057 aaronmk
    table = sql_gen.as_Table(table)
896
    changes = [(sql_gen.to_name_only_col(c, table), sql_gen.as_Value(v))
897
        for c, v in changes]
898
899 3056 aaronmk
    if in_place:
900
        assert cond == None
901 3058 aaronmk
902 3065 aaronmk
        query = 'ALTER TABLE '+table.to_str(db)+'\n'
903
        query += ',\n'.join(('ALTER COLUMN '+c.to_str(db)+' TYPE '
904 3153 aaronmk
            +db.col_info(sql_gen.with_default_table(c, table), cacheable_).type
905 3065 aaronmk
            +'\nUSING '+v.to_str(db) for c, v in changes))
906 3058 aaronmk
    else:
907
        query = 'UPDATE '+table.to_str(db)+'\nSET\n'
908
        query += ',\n'.join((c.to_str(db)+' = '+v.to_str(db)
909
            for c, v in changes))
910
        if cond != None: query += '\nWHERE\n'+cond.to_str(db)
911 3056 aaronmk
912 3266 aaronmk
    query = with_explain_comment(db, query)
913
914 2402 aaronmk
    return query
915
916 3074 aaronmk
def update(db, table, *args, **kw_args):
917 2402 aaronmk
    '''For params, see mk_update() and run_query()'''
918
    recover = kw_args.pop('recover', None)
919 3043 aaronmk
    cacheable = kw_args.pop('cacheable', False)
920 3030 aaronmk
    log_level = kw_args.pop('log_level', 2)
921 2402 aaronmk
922 3074 aaronmk
    cur = run_query(db, mk_update(db, table, *args, **kw_args), recover,
923
        cacheable, log_level=log_level)
924
    autoanalyze(db, table)
925
    return cur
926 2402 aaronmk
927 3286 aaronmk
def mk_delete(db, table, cond=None):
928
    '''
929
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
930
    @return str query
931
    '''
932
    query = 'DELETE FROM '+table.to_str(db)
933
    if cond != None: query += '\nWHERE '+cond.to_str(db)
934
935
    query = with_explain_comment(db, query)
936
937
    return query
938
939
def delete(db, table, *args, **kw_args):
940
    '''For params, see mk_delete() and run_query()'''
941
    recover = kw_args.pop('recover', None)
942 3295 aaronmk
    cacheable = kw_args.pop('cacheable', True)
943 3286 aaronmk
    log_level = kw_args.pop('log_level', 2)
944
945
    cur = run_query(db, mk_delete(db, table, *args, **kw_args), recover,
946
        cacheable, log_level=log_level)
947
    autoanalyze(db, table)
948
    return cur
949
950 135 aaronmk
def last_insert_id(db):
951 1849 aaronmk
    module = util.root_module(db.db)
952 135 aaronmk
    if module == 'psycopg2': return value(run_query(db, 'SELECT lastval()'))
953
    elif module == 'MySQLdb': return db.insert_id()
954
    else: return None
955 13 aaronmk
956 3490 aaronmk
def define_func(db, def_):
957
    func = def_.function
958
    while True:
959
        try:
960
            run_query(db, def_.to_str(db), recover=True, cacheable=True,
961
                log_ignore_excs=(DuplicateException,))
962
            break # successful
963
        except DuplicateException:
964 3495 aaronmk
            func.name = next_version(func.name)
965 3490 aaronmk
            # try again with next version of name
966
967 2394 aaronmk
def mk_flatten_mapping(db, into, cols, preserve=[], as_items=False):
968 2383 aaronmk
    '''Creates a mapping from original column names (which may have collisions)
969 2415 aaronmk
    to names that will be distinct among the columns' tables.
970 2383 aaronmk
    This is meant to be used for several tables that are being joined together.
971 2415 aaronmk
    @param cols The columns to combine. Duplicates will be removed.
972
    @param into The table for the new columns.
973 2394 aaronmk
    @param preserve [sql_gen.Col...] Columns not to rename. Note that these
974
        columns will be included in the mapping even if they are not in cols.
975
        The tables of the provided Col objects will be changed to into, so make
976
        copies of them if you want to keep the original tables.
977
    @param as_items Whether to return a list of dict items instead of a dict
978 2383 aaronmk
    @return dict(orig_col=new_col, ...)
979
        * orig_col: sql_gen.Col(orig_col_name, orig_table)
980 2392 aaronmk
        * new_col: sql_gen.Col(orig_col_name, into)
981
        * All mappings use the into table so its name can easily be
982 2383 aaronmk
          changed for all columns at once
983
    '''
984 2415 aaronmk
    cols = lists.uniqify(cols)
985
986 2394 aaronmk
    items = []
987 2389 aaronmk
    for col in preserve:
988 2390 aaronmk
        orig_col = copy.copy(col)
989 2392 aaronmk
        col.table = into
990 2394 aaronmk
        items.append((orig_col, col))
991
    preserve = set(preserve)
992
    for col in cols:
993 2716 aaronmk
        if col not in preserve:
994 3750 aaronmk
            items.append((col, sql_gen.Col(strings.ustr(col), into, col.srcs)))
995 2394 aaronmk
996
    if not as_items: items = dict(items)
997
    return items
998 2383 aaronmk
999 2393 aaronmk
def flatten(db, into, joins, cols, limit=None, start=None, **kw_args):
1000 2391 aaronmk
    '''For params, see mk_flatten_mapping()
1001
    @return See return value of mk_flatten_mapping()
1002
    '''
1003 2394 aaronmk
    items = mk_flatten_mapping(db, into, cols, as_items=True, **kw_args)
1004
    cols = [sql_gen.NamedCol(new.name, old) for old, new in items]
1005 3296 aaronmk
    run_query_into(db, mk_select(db, joins, cols, order_by=None, limit=limit,
1006 3305 aaronmk
        start=start), into=into, add_pkey_=True)
1007 3708 aaronmk
        # don't cache because the temp table will usually be truncated after use
1008 2394 aaronmk
    return dict(items)
1009 2391 aaronmk
1010 3079 aaronmk
##### Database structure introspection
1011 2414 aaronmk
1012 3321 aaronmk
#### Expressions
1013
1014 3353 aaronmk
bool_re = r'(?:true|false)'
1015
1016
def simplify_expr(expr):
1017
    expr = expr.replace('(NULL IS NULL)', 'true')
1018
    expr = expr.replace('(NULL IS NOT NULL)', 'false')
1019
    expr = re.sub(r' OR '+bool_re, r'', expr)
1020
    expr = re.sub(bool_re+r' OR ', r'', expr)
1021
    while True:
1022
        expr, n = re.subn(r'\((\([^()]*\))\)', r'\1', expr)
1023
        if n == 0: break
1024
    return expr
1025
1026 3321 aaronmk
name_re = r'(?:\w+|(?:"[^"]*")+)'
1027
1028
def parse_expr_col(str_):
1029
    match = re.match(r'^\('+name_re+r'\(('+name_re+r').*\)\)$', str_)
1030
    if match: str_ = match.group(1)
1031
    return sql_gen.unesc_name(str_)
1032
1033 3351 aaronmk
def map_expr(db, expr, mapping, in_cols_found=None):
1034
    '''Replaces output columns with input columns in an expression.
1035
    @param in_cols_found If set, will be filled in with the expr's (input) cols
1036
    '''
1037
    for out, in_ in mapping.iteritems():
1038
        orig_expr = expr
1039
        out = sql_gen.to_name_only_col(out)
1040
        in_str = sql_gen.to_name_only_col(sql_gen.remove_col_rename(in_)
1041
            ).to_str(db)
1042
1043
        # Replace out both with and without quotes
1044
        expr = expr.replace(out.to_str(db), in_str)
1045 4490 aaronmk
        expr = re.sub(r'(?<!\.)\b'+out.name+r'\b(?!\.)', in_str, expr)
1046 3351 aaronmk
1047
        if in_cols_found != None and expr != orig_expr: # replaced something
1048
            in_cols_found.append(in_)
1049 3353 aaronmk
1050
    return simplify_expr(expr)
1051 3351 aaronmk
1052 3079 aaronmk
#### Tables
1053
1054
def tables(db, schema_like='public', table_like='%', exact=False):
1055
    if exact: compare = '='
1056
    else: compare = 'LIKE'
1057
1058
    module = util.root_module(db.db)
1059
    if module == 'psycopg2':
1060
        conds = [('schemaname', sql_gen.CompareCond(schema_like, compare)),
1061
            ('tablename', sql_gen.CompareCond(table_like, compare))]
1062
        return values(select(db, 'pg_tables', ['tablename'], conds,
1063
            order_by='tablename', log_level=4))
1064
    elif module == 'MySQLdb':
1065
        return values(run_query(db, 'SHOW TABLES LIKE '+db.esc_value(table_like)
1066
            , cacheable=True, log_level=4))
1067
    else: raise NotImplementedError("Can't list tables for "+module+' database')
1068
1069
def table_exists(db, table):
1070
    table = sql_gen.as_Table(table)
1071
    return list(tables(db, table.schema, table.name, exact=True)) != []
1072
1073 2426 aaronmk
def table_row_count(db, table, recover=None):
1074 2786 aaronmk
    return value(run_query(db, mk_select(db, table, [sql_gen.row_count],
1075 3298 aaronmk
        order_by=None), recover=recover, log_level=3))
1076 2426 aaronmk
1077 2414 aaronmk
def table_cols(db, table, recover=None):
1078
    return list(col_names(select(db, table, limit=0, order_by=None,
1079 2443 aaronmk
        recover=recover, log_level=4)))
1080 2414 aaronmk
1081 4261 aaronmk
pkey_col = 'row_num'
1082
1083 2291 aaronmk
def pkey(db, table, recover=None):
1084 4261 aaronmk
    '''Uses pkey_col, or if not found, the first column in the table.'''
1085
    cols = table_cols(db, table, recover)
1086
    if pkey_col in cols: return pkey_col
1087
    else: return cols[0]
1088 832 aaronmk
1089 2559 aaronmk
not_null_col = 'not_null_col'
1090 2340 aaronmk
1091
def table_not_null_col(db, table, recover=None):
1092
    '''Name assumed to be the value of not_null_col. If not found, uses pkey.'''
1093
    if not_null_col in table_cols(db, table, recover): return not_null_col
1094
    else: return pkey(db, table, recover)
1095
1096 3348 aaronmk
def constraint_cond(db, constraint):
1097
    module = util.root_module(db.db)
1098
    if module == 'psycopg2':
1099
        table_str = sql_gen.Literal(constraint.table.to_str(db))
1100
        name_str = sql_gen.Literal(constraint.name)
1101
        return value(run_query(db, '''\
1102
SELECT consrc
1103
FROM pg_constraint
1104
WHERE
1105
conrelid = '''+table_str.to_str(db)+'''::regclass
1106
AND conname = '''+name_str.to_str(db)+'''
1107
'''
1108
            , cacheable=True, log_level=4))
1109
    else: raise NotImplementedError("Can't list index columns for "+module+
1110
        ' database')
1111
1112 3319 aaronmk
def index_cols(db, index):
1113 853 aaronmk
    '''Can also use this for UNIQUE constraints, because a UNIQUE index is
1114
    automatically created. When you don't know whether something is a UNIQUE
1115
    constraint or a UNIQUE index, use this function.'''
1116 3322 aaronmk
    index = sql_gen.as_Table(index)
1117 1909 aaronmk
    module = util.root_module(db.db)
1118
    if module == 'psycopg2':
1119 3322 aaronmk
        qual_index = sql_gen.Literal(index.to_str(db))
1120
        return map(parse_expr_col, values(run_query(db, '''\
1121
SELECT pg_get_indexdef(indexrelid, generate_series(1, indnatts), true)
1122
FROM pg_index
1123
WHERE indexrelid = '''+qual_index.to_str(db)+'''::regclass
1124 2782 aaronmk
'''
1125
            , cacheable=True, log_level=4)))
1126 1909 aaronmk
    else: raise NotImplementedError("Can't list index columns for "+module+
1127
        ' database')
1128 853 aaronmk
1129 3079 aaronmk
#### Functions
1130
1131
def function_exists(db, function):
1132 3423 aaronmk
    qual_function = sql_gen.Literal(function.to_str(db))
1133
    try:
1134 3425 aaronmk
        select(db, fields=[sql_gen.Cast('regproc', qual_function)],
1135
            recover=True, cacheable=True, log_level=4)
1136 3423 aaronmk
    except DoesNotExistException: return False
1137 4146 aaronmk
    except DuplicateException: return True # overloaded function
1138 3423 aaronmk
    else: return True
1139 3079 aaronmk
1140
##### Structural changes
1141
1142
#### Columns
1143
1144
def add_col(db, table, col, comment=None, **kw_args):
1145
    '''
1146
    @param col TypedCol Name may be versioned, so be sure to propagate any
1147
        renaming back to any source column for the TypedCol.
1148
    @param comment None|str SQL comment used to distinguish columns of the same
1149
        name from each other when they contain different data, to allow the
1150
        ADD COLUMN query to be cached. If not set, query will not be cached.
1151
    '''
1152
    assert isinstance(col, sql_gen.TypedCol)
1153
1154
    while True:
1155
        str_ = 'ALTER TABLE '+table.to_str(db)+' ADD COLUMN '+col.to_str(db)
1156
        if comment != None: str_ += ' '+sql_gen.esc_comment(comment)
1157
1158
        try:
1159
            run_query(db, str_, recover=True, cacheable=True, **kw_args)
1160
            break
1161
        except DuplicateException:
1162
            col.name = next_version(col.name)
1163
            # try again with next version of name
1164
1165
def add_not_null(db, col):
1166
    table = col.table
1167
    col = sql_gen.to_name_only_col(col)
1168
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ALTER COLUMN '
1169
        +col.to_str(db)+' SET NOT NULL', cacheable=True, log_level=3)
1170
1171 4443 aaronmk
def drop_not_null(db, col):
1172
    table = col.table
1173
    col = sql_gen.to_name_only_col(col)
1174
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ALTER COLUMN '
1175
        +col.to_str(db)+' DROP NOT NULL', cacheable=True, log_level=3)
1176
1177 2096 aaronmk
row_num_col = '_row_num'
1178
1179 3079 aaronmk
row_num_typed_col = sql_gen.TypedCol(row_num_col, 'serial', nullable=False,
1180
    constraints='PRIMARY KEY')
1181
1182
def add_row_num(db, table):
1183
    '''Adds a row number column to a table. Its name is in row_num_col. It will
1184
    be the primary key.'''
1185
    add_col(db, table, row_num_typed_col, log_level=3)
1186
1187
#### Indexes
1188
1189
def add_pkey(db, table, cols=None, recover=None):
1190
    '''Adds a primary key.
1191
    @param cols [sql_gen.Col,...] The columns in the primary key.
1192
        Defaults to the first column in the table.
1193
    @pre The table must not already have a primary key.
1194
    '''
1195
    table = sql_gen.as_Table(table)
1196
    if cols == None: cols = [pkey(db, table, recover)]
1197
    col_strs = [sql_gen.to_name_only_col(v).to_str(db) for v in cols]
1198
1199
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ADD PRIMARY KEY ('
1200
        +(', '.join(col_strs))+')', recover=True, cacheable=True, log_level=3,
1201
        log_ignore_excs=(DuplicateException,))
1202
1203 2998 aaronmk
def add_index(db, exprs, table=None, unique=False, ensure_not_null_=True):
1204 2688 aaronmk
    '''Adds an index on column(s) or expression(s) if it doesn't already exist.
1205 3356 aaronmk
    Currently, only function calls and literal values are supported expressions.
1206 2998 aaronmk
    @param ensure_not_null_ If set, translates NULL values to sentinel values.
1207 2847 aaronmk
        This allows indexes to be used for comparisons where NULLs are equal.
1208 2538 aaronmk
    '''
1209 2964 aaronmk
    exprs = lists.mk_seq(exprs)
1210 2538 aaronmk
1211 2688 aaronmk
    # Parse exprs
1212
    old_exprs = exprs[:]
1213
    exprs = []
1214
    cols = []
1215
    for i, expr in enumerate(old_exprs):
1216 2823 aaronmk
        expr = sql_gen.as_Col(expr, table)
1217 2688 aaronmk
1218 2823 aaronmk
        # Handle nullable columns
1219 2998 aaronmk
        if ensure_not_null_:
1220 3164 aaronmk
            try: expr = sql_gen.ensure_not_null(db, expr)
1221 2860 aaronmk
            except KeyError: pass # unknown type, so just create plain index
1222 2823 aaronmk
1223 2688 aaronmk
        # Extract col
1224 3002 aaronmk
        expr = copy.deepcopy(expr) # don't modify input!
1225 3356 aaronmk
        col = expr
1226
        if isinstance(expr, sql_gen.FunctionCall): col = expr.args[0]
1227
        expr = sql_gen.cast_literal(expr)
1228
        if not isinstance(expr, (sql_gen.Expr, sql_gen.Col)):
1229 2688 aaronmk
            expr = sql_gen.Expr(expr)
1230 3356 aaronmk
1231 2688 aaronmk
1232
        # Extract table
1233
        if table == None:
1234
            assert sql_gen.is_table_col(col)
1235
            table = col.table
1236
1237 3356 aaronmk
        if isinstance(col, sql_gen.Col): col.table = None
1238 2688 aaronmk
1239
        exprs.append(expr)
1240
        cols.append(col)
1241 2408 aaronmk
1242 2688 aaronmk
    table = sql_gen.as_Table(table)
1243
1244 3005 aaronmk
    # Add index
1245 3148 aaronmk
    str_ = 'CREATE'
1246
    if unique: str_ += ' UNIQUE'
1247
    str_ += ' INDEX ON '+table.to_str(db)+' ('+(
1248
        ', '.join((v.to_str(db) for v in exprs)))+')'
1249
    run_query(db, str_, recover=True, cacheable=True, log_level=3)
1250 2408 aaronmk
1251 3083 aaronmk
already_indexed = object() # tells add_indexes() the pkey has already been added
1252
1253
def add_indexes(db, table, has_pkey=True):
1254
    '''Adds an index on all columns in a table.
1255
    @param has_pkey bool|already_indexed Whether a pkey instead of a regular
1256
        index should be added on the first column.
1257
        * If already_indexed, the pkey is assumed to have already been added
1258
    '''
1259
    cols = table_cols(db, table)
1260
    if has_pkey:
1261
        if has_pkey is not already_indexed: add_pkey(db, table)
1262
        cols = cols[1:]
1263
    for col in cols: add_index(db, col, table)
1264
1265 3079 aaronmk
#### Tables
1266 2772 aaronmk
1267 3079 aaronmk
### Maintenance
1268 2772 aaronmk
1269 3079 aaronmk
def analyze(db, table):
1270
    table = sql_gen.as_Table(table)
1271
    run_query(db, 'ANALYZE '+table.to_str(db), log_level=3)
1272 2934 aaronmk
1273 3079 aaronmk
def autoanalyze(db, table):
1274
    if db.autoanalyze: analyze(db, table)
1275 2935 aaronmk
1276 3079 aaronmk
def vacuum(db, table):
1277
    table = sql_gen.as_Table(table)
1278
    db.with_autocommit(lambda: run_query(db, 'VACUUM ANALYZE '+table.to_str(db),
1279
        log_level=3))
1280 2086 aaronmk
1281 3079 aaronmk
### Lifecycle
1282
1283 3247 aaronmk
def drop(db, type_, name):
1284
    name = sql_gen.as_Name(name)
1285
    run_query(db, 'DROP '+type_+' IF EXISTS '+name.to_str(db)+' CASCADE')
1286 2889 aaronmk
1287 3247 aaronmk
def drop_table(db, table): drop(db, 'TABLE', table)
1288
1289 3082 aaronmk
def create_table(db, table, cols=[], has_pkey=True, col_indexes=True,
1290
    like=None):
1291 2675 aaronmk
    '''Creates a table.
1292 2681 aaronmk
    @param cols [sql_gen.TypedCol,...] The column names and types
1293
    @param has_pkey If set, the first column becomes the primary key.
1294 2760 aaronmk
    @param col_indexes bool|[ref]
1295
        * If True, indexes will be added on all non-pkey columns.
1296
        * If a list reference, [0] will be set to a function to do this.
1297
          This can be used to delay index creation until the table is populated.
1298 2675 aaronmk
    '''
1299
    table = sql_gen.as_Table(table)
1300
1301 3082 aaronmk
    if like != None:
1302
        cols = [sql_gen.CustomCode('LIKE '+like.to_str(db)+' INCLUDING ALL')
1303
            ]+cols
1304 2681 aaronmk
    if has_pkey:
1305
        cols[0] = pkey = copy.copy(cols[0]) # don't modify input!
1306 2872 aaronmk
        pkey.constraints = 'PRIMARY KEY'
1307 2681 aaronmk
1308 3085 aaronmk
    temp = table.is_temp and not db.debug_temp
1309
        # temp tables permanent in debug_temp mode
1310 2760 aaronmk
1311 3085 aaronmk
    # Create table
1312 3383 aaronmk
    def create():
1313 3085 aaronmk
        str_ = 'CREATE'
1314
        if temp: str_ += ' TEMP'
1315
        str_ += ' TABLE '+table.to_str(db)+' (\n'
1316
        str_ += '\n, '.join(c.to_str(db) for c in cols)
1317 3126 aaronmk
        str_ += '\n);'
1318 3085 aaronmk
1319 3383 aaronmk
        run_query(db, str_, recover=True, cacheable=True, log_level=2,
1320
            log_ignore_excs=(DuplicateException,))
1321
    if table.is_temp:
1322
        while True:
1323
            try:
1324
                create()
1325
                break
1326
            except DuplicateException:
1327
                table.name = next_version(table.name)
1328
                # try again with next version of name
1329
    else: create()
1330 3085 aaronmk
1331 2760 aaronmk
    # Add indexes
1332 2773 aaronmk
    if has_pkey: has_pkey = already_indexed
1333
    def add_indexes_(): add_indexes(db, table, has_pkey)
1334
    if isinstance(col_indexes, list): col_indexes[0] = add_indexes_ # defer
1335
    elif col_indexes: add_indexes_() # add now
1336 2675 aaronmk
1337 3084 aaronmk
def copy_table_struct(db, src, dest):
1338
    '''Creates a structure-only copy of a table. (Does not copy data.)'''
1339 3085 aaronmk
    create_table(db, dest, has_pkey=False, col_indexes=False, like=src)
1340 3084 aaronmk
1341 3079 aaronmk
### Data
1342 2684 aaronmk
1343 2970 aaronmk
def truncate(db, table, schema='public', **kw_args):
1344
    '''For params, see run_query()'''
1345 2777 aaronmk
    table = sql_gen.as_Table(table, schema)
1346 2970 aaronmk
    return run_query(db, 'TRUNCATE '+table.to_str(db)+' CASCADE', **kw_args)
1347 2732 aaronmk
1348 2965 aaronmk
def empty_temp(db, tables):
1349
    tables = lists.mk_seq(tables)
1350 2971 aaronmk
    for table in tables: truncate(db, table, log_level=3)
1351 2965 aaronmk
1352 1968 aaronmk
def empty_db(db, schema='public', **kw_args):
1353
    '''For kw_args, see tables()'''
1354
    for table in tables(db, schema, **kw_args): truncate(db, table, schema)
1355 3094 aaronmk
1356
def distinct_table(db, table, distinct_on):
1357
    '''Creates a copy of a temp table which is distinct on the given columns.
1358 3099 aaronmk
    The old and new tables will both get an index on these columns, to
1359
    facilitate merge joins.
1360 3097 aaronmk
    @param distinct_on If empty, creates a table with one row. This is useful if
1361
        your distinct_on columns are all literal values.
1362 3099 aaronmk
    @return The new table.
1363 3094 aaronmk
    '''
1364 3099 aaronmk
    new_table = sql_gen.suffixed_table(table, '_distinct')
1365 3411 aaronmk
    distinct_on = filter(sql_gen.is_table_col, distinct_on)
1366 3094 aaronmk
1367 3099 aaronmk
    copy_table_struct(db, table, new_table)
1368 3097 aaronmk
1369
    limit = None
1370
    if distinct_on == []: limit = 1 # one sample row
1371 3099 aaronmk
    else:
1372
        add_index(db, distinct_on, new_table, unique=True)
1373
        add_index(db, distinct_on, table) # for join optimization
1374 3097 aaronmk
1375 3313 aaronmk
    insert_select(db, new_table, None, mk_select(db, table, order_by=None,
1376
        limit=limit), ignore=True)
1377 3099 aaronmk
    analyze(db, new_table)
1378 3094 aaronmk
1379 3099 aaronmk
    return new_table