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 5447 aaronmk
        if cond != None: msg += ' with condition '+strings.as_tt(cond)
61 3345 aaronmk
        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 4936 aaronmk
        cols = [sql_gen.Col('data_type'), sql_gen.Col('udt_name'),
470
            'column_default', sql_gen.Cast('boolean',
471
            sql_gen.Col('is_nullable'))]
472 2643 aaronmk
473 3750 aaronmk
        conds = [('table_name', col.table.name),
474
            ('column_name', strings.ustr(col.name))]
475 2643 aaronmk
        schema = col.table.schema
476
        if schema != None: conds.append(('table_schema', schema))
477
478 3638 aaronmk
        cur = select(self, table, cols, conds, order_by='table_schema', limit=1,
479
            cacheable=cacheable, log_level=4) # TODO: order by search_path order
480 4936 aaronmk
        try: type_, extra_type, default, nullable = row(cur)
481 4114 aaronmk
        except StopIteration: raise sql_gen.NoUnderlyingTableException(col)
482 2819 aaronmk
        default = sql_gen.as_Code(default, self)
483 4936 aaronmk
        if type_ == 'USER-DEFINED': type_ = extra_type
484 4939 aaronmk
        elif type_ == 'ARRAY':
485
            type_ = sql_gen.ArrayType(strings.remove_prefix('_', extra_type,
486
                require=True))
487 2819 aaronmk
488
        return sql_gen.TypedCol(col.name, type_, default, nullable)
489 2917 aaronmk
490
    def TempFunction(self, name):
491
        if self.debug_temp: schema = None
492
        else: schema = 'pg_temp'
493
        return sql_gen.Function(name, schema)
494 1849 aaronmk
495 1869 aaronmk
connect = DbConn
496
497 832 aaronmk
##### Recoverable querying
498 15 aaronmk
499 2139 aaronmk
def with_savepoint(db, func): return db.with_savepoint(func)
500 11 aaronmk
501 2791 aaronmk
def run_query(db, query, recover=None, cacheable=False, log_level=2,
502
    log_ignore_excs=None, **kw_args):
503 2794 aaronmk
    '''For params, see DbConn.run_query()'''
504 830 aaronmk
    if recover == None: recover = False
505 2464 aaronmk
    if log_ignore_excs == None: log_ignore_excs = ()
506
    log_ignore_excs = tuple(log_ignore_excs)
507 3236 aaronmk
    debug_msg_ref = [None]
508 830 aaronmk
509 3267 aaronmk
    query = with_explain_comment(db, query)
510 3258 aaronmk
511 2148 aaronmk
    try:
512 2464 aaronmk
        try:
513 2794 aaronmk
            def run(): return db.run_query(query, cacheable, log_level,
514 2793 aaronmk
                debug_msg_ref, **kw_args)
515 2796 aaronmk
            if recover and not db.is_cached(query):
516 2464 aaronmk
                return with_savepoint(db, run)
517
            else: return run() # don't need savepoint if cached
518
        except Exception, e:
519 3095 aaronmk
            msg = strings.ustr(e.args[0])
520 4103 aaronmk
            msg = re.sub(r'^(?:PL/Python: )?ValueError: ', r'', msg)
521 2464 aaronmk
522 3095 aaronmk
            match = re.match(r'^duplicate key value violates unique constraint '
523 3338 aaronmk
                r'"(.+?)"', msg)
524 2464 aaronmk
            if match:
525 3338 aaronmk
                constraint, = match.groups()
526 3025 aaronmk
                cols = []
527 5446 aaronmk
                cond = None
528 3025 aaronmk
                if recover: # need auto-rollback to run index_cols()
529 5446 aaronmk
                    try:
530
                        cols = index_cols(db, constraint)
531
                        cond = index_cond(db, constraint)
532 3025 aaronmk
                    except NotImplementedError: pass
533 5446 aaronmk
                raise DuplicateKeyException(constraint, cond, cols, e)
534 2464 aaronmk
535 3095 aaronmk
            match = re.match(r'^null value in column "(.+?)" violates not-null'
536 2464 aaronmk
                r' constraint', msg)
537 3345 aaronmk
            if match:
538
                col, = match.groups()
539
                raise NullValueException('NOT NULL', None, [col], e)
540 2464 aaronmk
541 3346 aaronmk
            match = re.match(r'^new row for relation "(.+?)" violates check '
542
                r'constraint "(.+?)"', msg)
543
            if match:
544
                table, constraint = match.groups()
545 3347 aaronmk
                constraint = sql_gen.Col(constraint, table)
546 3349 aaronmk
                cond = None
547
                if recover: # need auto-rollback to run constraint_cond()
548
                    try: cond = constraint_cond(db, constraint)
549
                    except NotImplementedError: pass
550
                raise CheckException(constraint.to_str(db), cond, [], e)
551 3346 aaronmk
552 3095 aaronmk
            match = re.match(r'^(?:invalid input (?:syntax|value)\b.*?'
553 3635 aaronmk
                r'|.+? out of range): "(.+?)"', msg)
554 2464 aaronmk
            if match:
555 3109 aaronmk
                value, = match.groups()
556
                raise InvalidValueException(strings.to_unicode(value), e)
557 2464 aaronmk
558 3095 aaronmk
            match = re.match(r'^column "(.+?)" is of type (.+?) but expression '
559 2523 aaronmk
                r'is of type', msg)
560
            if match:
561
                col, type_ = match.groups()
562
                raise MissingCastException(type_, col, e)
563
564 4141 aaronmk
            match = re.match(r'^could not determine polymorphic type because '
565
                r'input has type "unknown"', msg)
566
            if match: raise MissingCastException('text', None, e)
567
568 4485 aaronmk
            match = re.match(r'^.+? types .+? and .+? cannot be matched', msg)
569
            if match: raise MissingCastException('text', None, e)
570
571 4509 aaronmk
            typed_name_re = r'^(\S+) "(.+?)"(?: of relation ".+?")?'
572 3419 aaronmk
573
            match = re.match(typed_name_re+r'.*? already exists', msg)
574 2945 aaronmk
            if match:
575
                type_, name = match.groups()
576
                raise DuplicateException(type_, name, e)
577 2464 aaronmk
578 4145 aaronmk
            match = re.match(r'more than one (\S+) named ""(.+?)""', msg)
579
            if match:
580
                type_, name = match.groups()
581
                raise DuplicateException(type_, name, e)
582
583 3419 aaronmk
            match = re.match(typed_name_re+r' does not exist', msg)
584
            if match:
585
                type_, name = match.groups()
586
                raise DoesNotExistException(type_, name, e)
587
588 2464 aaronmk
            raise # no specific exception raised
589
    except log_ignore_excs:
590
        log_level += 2
591
        raise
592
    finally:
593 3236 aaronmk
        if debug_msg_ref[0] != None: db.log_debug(debug_msg_ref[0], log_level)
594 830 aaronmk
595 832 aaronmk
##### Basic queries
596
597 3256 aaronmk
def is_explainable(query):
598
    # See <http://www.postgresql.org/docs/8.3/static/sql-explain.html#AEN57749>
599 3257 aaronmk
    return re.match(r'^(?:SELECT|INSERT|UPDATE|DELETE|VALUES|EXECUTE|DECLARE)\b'
600
        , query)
601 3256 aaronmk
602 3263 aaronmk
def explain(db, query, **kw_args):
603
    '''
604
    For params, see run_query().
605
    '''
606 3267 aaronmk
    kw_args.setdefault('log_level', 4)
607 3263 aaronmk
608 3750 aaronmk
    return strings.ustr(strings.join_lines(values(run_query(db,
609
        'EXPLAIN '+query, recover=True, cacheable=True, **kw_args))))
610 3256 aaronmk
        # not a higher log_level because it's useful to see what query is being
611
        # run before it's executed, which EXPLAIN effectively provides
612
613 3265 aaronmk
def has_comment(query): return query.endswith('*/')
614
615
def with_explain_comment(db, query, **kw_args):
616 3269 aaronmk
    if db.autoexplain and not has_comment(query) and is_explainable(query):
617 3265 aaronmk
        query += '\n'+sql_gen.esc_comment(' EXPLAIN:\n'
618
            +explain(db, query, **kw_args))
619
    return query
620
621 2153 aaronmk
def next_version(name):
622 2163 aaronmk
    version = 1 # first existing name was version 0
623 2586 aaronmk
    match = re.match(r'^(.*)#(\d+)$', name)
624 2153 aaronmk
    if match:
625 2586 aaronmk
        name, version = match.groups()
626
        version = int(version)+1
627 2932 aaronmk
    return sql_gen.concat(name, '#'+str(version))
628 2153 aaronmk
629 2899 aaronmk
def lock_table(db, table, mode):
630
    table = sql_gen.as_Table(table)
631
    run_query(db, 'LOCK TABLE '+table.to_str(db)+' IN '+mode+' MODE')
632
633 3303 aaronmk
def run_query_into(db, query, into=None, add_pkey_=False, **kw_args):
634 2085 aaronmk
    '''Outputs a query to a temp table.
635
    For params, see run_query().
636
    '''
637 2789 aaronmk
    if into == None: return run_query(db, query, **kw_args)
638 2790 aaronmk
639
    assert isinstance(into, sql_gen.Table)
640
641 2992 aaronmk
    into.is_temp = True
642 3008 aaronmk
    # "temporary tables cannot specify a schema name", so remove schema
643
    into.schema = None
644 2992 aaronmk
645 2790 aaronmk
    kw_args['recover'] = True
646 2945 aaronmk
    kw_args.setdefault('log_ignore_excs', (DuplicateException,))
647 2790 aaronmk
648 2916 aaronmk
    temp = not db.debug_temp # tables are permanent in debug_temp mode
649 2790 aaronmk
650
    # Create table
651
    while True:
652
        create_query = 'CREATE'
653
        if temp: create_query += ' TEMP'
654
        create_query += ' TABLE '+into.to_str(db)+' AS\n'+query
655 2385 aaronmk
656 2790 aaronmk
        try:
657
            cur = run_query(db, create_query, **kw_args)
658
                # CREATE TABLE AS sets rowcount to # rows in query
659
            break
660 2945 aaronmk
        except DuplicateException, e:
661 2790 aaronmk
            into.name = next_version(into.name)
662
            # try again with next version of name
663
664 3303 aaronmk
    if add_pkey_: add_pkey(db, into)
665 3075 aaronmk
666
    # According to the PostgreSQL doc, "The autovacuum daemon cannot access and
667
    # therefore cannot vacuum or analyze temporary tables. [...] if a temporary
668
    # table is going to be used in complex queries, it is wise to run ANALYZE on
669
    # the temporary table after it is populated."
670
    # (http://www.postgresql.org/docs/9.1/static/sql-createtable.html)
671
    # If into is not a temp table, ANALYZE is useful but not required.
672 3073 aaronmk
    analyze(db, into)
673 2790 aaronmk
674
    return cur
675 2085 aaronmk
676 2120 aaronmk
order_by_pkey = object() # tells mk_select() to order by the pkey
677
678 2199 aaronmk
distinct_on_all = object() # tells mk_select() to SELECT DISTINCT ON all columns
679
680 3420 aaronmk
def mk_select(db, tables=None, fields=None, conds=None, distinct_on=[],
681 3494 aaronmk
    limit=None, start=None, order_by=order_by_pkey, default_table=None,
682
    explain=True):
683 1981 aaronmk
    '''
684 2121 aaronmk
    @param tables The single table to select from, or a list of tables to join
685 2280 aaronmk
        together, with tables after the first being sql_gen.Join objects
686 1981 aaronmk
    @param fields Use None to select all fields in the table
687 2377 aaronmk
    @param conds WHERE conditions: [(compare_left_side, compare_right_side),...]
688 2379 aaronmk
        * container can be any iterable type
689 2399 aaronmk
        * compare_left_side: sql_gen.Code|str (for col name)
690
        * compare_right_side: sql_gen.ValueCond|literal value
691 2199 aaronmk
    @param distinct_on The columns to SELECT DISTINCT ON, or distinct_on_all to
692
        use all columns
693 2786 aaronmk
    @return query
694 1981 aaronmk
    '''
695 2315 aaronmk
    # Parse tables param
696 2964 aaronmk
    tables = lists.mk_seq(tables)
697 2141 aaronmk
    tables = list(tables) # don't modify input! (list() copies input)
698 2315 aaronmk
    table0 = sql_gen.as_Table(tables.pop(0)) # first table is separate
699 2121 aaronmk
700 2315 aaronmk
    # Parse other params
701 2376 aaronmk
    if conds == None: conds = []
702 2650 aaronmk
    elif dicts.is_dict(conds): conds = conds.items()
703 2379 aaronmk
    conds = list(conds) # don't modify input! (list() copies input)
704 3129 aaronmk
    assert limit == None or isinstance(limit, (int, long))
705
    assert start == None or isinstance(start, (int, long))
706 2315 aaronmk
    if order_by is order_by_pkey:
707 3421 aaronmk
        if table0 == None or distinct_on != []: order_by = None
708 5388 aaronmk
        else: order_by = pkey_name(db, table0, recover=True)
709 865 aaronmk
710 2315 aaronmk
    query = 'SELECT'
711 2056 aaronmk
712 2315 aaronmk
    def parse_col(col): return sql_gen.as_Col(col, default_table).to_str(db)
713 2056 aaronmk
714 2200 aaronmk
    # DISTINCT ON columns
715 2233 aaronmk
    if distinct_on != []:
716 2467 aaronmk
        query += '\nDISTINCT'
717 2254 aaronmk
        if distinct_on is not distinct_on_all:
718 2200 aaronmk
            query += ' ON ('+(', '.join(map(parse_col, distinct_on)))+')'
719
720
    # Columns
721 3185 aaronmk
    if query.find('\n') >= 0: whitespace = '\n'
722
    else: whitespace = ' '
723
    if fields == None: query += whitespace+'*'
724 2765 aaronmk
    else:
725
        assert fields != []
726 3185 aaronmk
        if len(fields) > 1: whitespace = '\n'
727
        query += whitespace+('\n, '.join(map(parse_col, fields)))
728 2200 aaronmk
729
    # Main table
730 3185 aaronmk
    if query.find('\n') >= 0 or len(tables) > 0: whitespace = '\n'
731
    else: whitespace = ' '
732 3420 aaronmk
    if table0 != None: query += whitespace+'FROM '+table0.to_str(db)
733 865 aaronmk
734 2122 aaronmk
    # Add joins
735 2271 aaronmk
    left_table = table0
736 2263 aaronmk
    for join_ in tables:
737
        table = join_.table
738 2238 aaronmk
739 2343 aaronmk
        # Parse special values
740
        if join_.type_ is sql_gen.filter_out: # filter no match
741 2376 aaronmk
            conds.append((sql_gen.Col(table_not_null_col(db, table), table),
742 2853 aaronmk
                sql_gen.CompareCond(None, '~=')))
743 2343 aaronmk
744 2467 aaronmk
        query += '\n'+join_.to_str(db, left_table)
745 2122 aaronmk
746
        left_table = table
747
748 865 aaronmk
    missing = True
749 2376 aaronmk
    if conds != []:
750 2576 aaronmk
        if len(conds) == 1: whitespace = ' '
751
        else: whitespace = '\n'
752 2578 aaronmk
        query += '\n'+sql_gen.combine_conds([sql_gen.ColValueCond(l, r)
753
            .to_str(db) for l, r in conds], 'WHERE')
754 2227 aaronmk
    if order_by != None:
755 2467 aaronmk
        query += '\nORDER BY '+sql_gen.as_Col(order_by, table0).to_str(db)
756 3297 aaronmk
    if limit != None: query += '\nLIMIT '+str(limit)
757 865 aaronmk
    if start != None:
758 2467 aaronmk
        if start != 0: query += '\nOFFSET '+str(start)
759 865 aaronmk
760 3494 aaronmk
    if explain: query = with_explain_comment(db, query)
761 3266 aaronmk
762 2786 aaronmk
    return query
763 11 aaronmk
764 2054 aaronmk
def select(db, *args, **kw_args):
765
    '''For params, see mk_select() and run_query()'''
766
    recover = kw_args.pop('recover', None)
767
    cacheable = kw_args.pop('cacheable', True)
768 2442 aaronmk
    log_level = kw_args.pop('log_level', 2)
769 2054 aaronmk
770 2791 aaronmk
    return run_query(db, mk_select(db, *args, **kw_args), recover, cacheable,
771
        log_level=log_level)
772 2054 aaronmk
773 2788 aaronmk
def mk_insert_select(db, table, cols=None, select_query=None, returning=None,
774 3181 aaronmk
    embeddable=False, ignore=False, src=None):
775 1960 aaronmk
    '''
776
    @param returning str|None An inserted column (such as pkey) to return
777 2070 aaronmk
    @param embeddable Whether the query should be embeddable as a nested SELECT.
778 2073 aaronmk
        Warning: If you set this and cacheable=True when the query is run, the
779
        query will be fully cached, not just if it raises an exception.
780 3009 aaronmk
    @param ignore Whether to ignore duplicate keys.
781 3181 aaronmk
    @param src Will be included in the name of any created function, to help
782
        identify the data source in pg_stat_activity.
783 1960 aaronmk
    '''
784 2754 aaronmk
    table = sql_gen.remove_table_rename(sql_gen.as_Table(table))
785 2318 aaronmk
    if cols == []: cols = None # no cols (all defaults) = unknown col names
786 3010 aaronmk
    if cols != None: cols = [sql_gen.to_name_only_col(c, table) for c in cols]
787 2063 aaronmk
    if select_query == None: select_query = 'DEFAULT VALUES'
788 2327 aaronmk
    if returning != None: returning = sql_gen.as_Col(returning, table)
789 2063 aaronmk
790 2497 aaronmk
    first_line = 'INSERT INTO '+table.to_str(db)
791 2063 aaronmk
792 3009 aaronmk
    def mk_insert(select_query):
793
        query = first_line
794 3014 aaronmk
        if cols != None:
795
            query += '\n('+(', '.join((c.to_str(db) for c in cols)))+')'
796 3009 aaronmk
        query += '\n'+select_query
797
798
        if returning != None:
799
            returning_name_col = sql_gen.to_name_only_col(returning)
800
            query += '\nRETURNING '+returning_name_col.to_str(db)
801
802
        return query
803 2063 aaronmk
804 3489 aaronmk
    return_type = sql_gen.CustomCode('unknown')
805
    if returning != None: return_type = sql_gen.ColType(returning)
806 3017 aaronmk
807 3009 aaronmk
    if ignore:
808 3017 aaronmk
        # Always return something to set the correct rowcount
809
        if returning == None: returning = sql_gen.NamedCol('NULL', None)
810
811 3009 aaronmk
        embeddable = True # must use function
812 3010 aaronmk
813 3450 aaronmk
        if cols == None: row = [sql_gen.Col(sql_gen.all_cols, 'row')]
814
        else: row = [sql_gen.Col(c.name, 'row') for c in cols]
815 3092 aaronmk
816 3484 aaronmk
        query = sql_gen.RowExcIgnore(sql_gen.RowType(table), select_query,
817 3497 aaronmk
            sql_gen.ReturnQuery(mk_insert(sql_gen.Values(row).to_str(db))),
818
            cols)
819 3009 aaronmk
    else: query = mk_insert(select_query)
820
821 2070 aaronmk
    if embeddable:
822
        # Create function
823 2513 aaronmk
        function_name = sql_gen.clean_name(first_line)
824 3181 aaronmk
        if src != None: function_name = src+': '+function_name
825 2189 aaronmk
        while True:
826
            try:
827 3451 aaronmk
                func = db.TempFunction(function_name)
828 3489 aaronmk
                def_ = sql_gen.FunctionDef(func, sql_gen.SetOf(return_type),
829
                    query)
830 2194 aaronmk
831 3443 aaronmk
                run_query(db, def_.to_str(db), recover=True, cacheable=True,
832 2945 aaronmk
                    log_ignore_excs=(DuplicateException,))
833 2189 aaronmk
                break # this version was successful
834 2945 aaronmk
            except DuplicateException, e:
835 2189 aaronmk
                function_name = next_version(function_name)
836
                # try again with next version of name
837 2070 aaronmk
838 2337 aaronmk
        # Return query that uses function
839 3009 aaronmk
        cols = None
840
        if returning != None: cols = [returning]
841 3451 aaronmk
        func_table = sql_gen.NamedTable('f', sql_gen.FunctionCall(func), cols)
842
            # AS clause requires function alias
843 3298 aaronmk
        return mk_select(db, func_table, order_by=None)
844 2070 aaronmk
845 2787 aaronmk
    return query
846 2066 aaronmk
847 3074 aaronmk
def insert_select(db, table, *args, **kw_args):
848 2085 aaronmk
    '''For params, see mk_insert_select() and run_query_into()
849 2386 aaronmk
    @param into sql_gen.Table with suggested name of temp table to put RETURNING
850
        values in
851 2072 aaronmk
    '''
852 3141 aaronmk
    returning = kw_args.get('returning', None)
853
    ignore = kw_args.get('ignore', False)
854
855 2386 aaronmk
    into = kw_args.pop('into', None)
856
    if into != None: kw_args['embeddable'] = True
857 2066 aaronmk
    recover = kw_args.pop('recover', None)
858 3141 aaronmk
    if ignore: recover = True
859 2066 aaronmk
    cacheable = kw_args.pop('cacheable', True)
860 2673 aaronmk
    log_level = kw_args.pop('log_level', 2)
861 2066 aaronmk
862 3141 aaronmk
    rowcount_only = ignore and returning == None # keep NULL rows on server
863
    if rowcount_only: into = sql_gen.Table('rowcount')
864
865 3074 aaronmk
    cur = run_query_into(db, mk_insert_select(db, table, *args, **kw_args),
866
        into, recover=recover, cacheable=cacheable, log_level=log_level)
867 3141 aaronmk
    if rowcount_only: empty_temp(db, into)
868 3074 aaronmk
    autoanalyze(db, table)
869
    return cur
870 2063 aaronmk
871 2738 aaronmk
default = sql_gen.default # tells insert() to use the default value for a column
872 2066 aaronmk
873 2063 aaronmk
def insert(db, table, row, *args, **kw_args):
874 2085 aaronmk
    '''For params, see insert_select()'''
875 5050 aaronmk
    ignore = kw_args.pop('ignore', False)
876 5094 aaronmk
    if ignore: kw_args.setdefault('recover', True)
877 5050 aaronmk
878 1960 aaronmk
    if lists.is_seq(row): cols = None
879
    else:
880
        cols = row.keys()
881
        row = row.values()
882 2738 aaronmk
    row = list(row) # ensure that "== []" works
883 1960 aaronmk
884 2738 aaronmk
    if row == []: query = None
885
    else: query = sql_gen.Values(row).to_str(db)
886 1961 aaronmk
887 5050 aaronmk
    try: return insert_select(db, table, cols, query, *args, **kw_args)
888 5057 aaronmk
    except (DuplicateKeyException, NullValueException):
889 5050 aaronmk
        if not ignore: raise
890 5163 aaronmk
        return None
891 11 aaronmk
892 3152 aaronmk
def mk_update(db, table, changes=None, cond=None, in_place=False,
893 3153 aaronmk
    cacheable_=True):
894 2402 aaronmk
    '''
895
    @param changes [(col, new_value),...]
896
        * container can be any iterable type
897
        * col: sql_gen.Code|str (for col name)
898
        * new_value: sql_gen.Code|literal value
899
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
900 3056 aaronmk
    @param in_place If set, locks the table and updates rows in place.
901
        This avoids creating dead rows in PostgreSQL.
902
        * cond must be None
903 3153 aaronmk
    @param cacheable_ Whether column structure information used to generate the
904 3152 aaronmk
        query can be cached
905 2402 aaronmk
    @return str query
906
    '''
907 3057 aaronmk
    table = sql_gen.as_Table(table)
908
    changes = [(sql_gen.to_name_only_col(c, table), sql_gen.as_Value(v))
909
        for c, v in changes]
910
911 3056 aaronmk
    if in_place:
912
        assert cond == None
913 3058 aaronmk
914 5398 aaronmk
        def col_type(col):
915
            return sql_gen.canon_type(db.col_info(
916
                sql_gen.with_default_table(c, table), cacheable_).type)
917
        changes = [(c, v, col_type(c)) for c, v in changes]
918 3065 aaronmk
        query = 'ALTER TABLE '+table.to_str(db)+'\n'
919 5396 aaronmk
        query += ',\n'.join(('ALTER COLUMN '+c.to_str(db)+' TYPE '+t+'\nUSING '
920
            +v.to_str(db) for c, v, t in changes))
921 3058 aaronmk
    else:
922
        query = 'UPDATE '+table.to_str(db)+'\nSET\n'
923
        query += ',\n'.join((c.to_str(db)+' = '+v.to_str(db)
924
            for c, v in changes))
925
        if cond != None: query += '\nWHERE\n'+cond.to_str(db)
926 3056 aaronmk
927 3266 aaronmk
    query = with_explain_comment(db, query)
928
929 2402 aaronmk
    return query
930
931 3074 aaronmk
def update(db, table, *args, **kw_args):
932 2402 aaronmk
    '''For params, see mk_update() and run_query()'''
933
    recover = kw_args.pop('recover', None)
934 3043 aaronmk
    cacheable = kw_args.pop('cacheable', False)
935 3030 aaronmk
    log_level = kw_args.pop('log_level', 2)
936 2402 aaronmk
937 3074 aaronmk
    cur = run_query(db, mk_update(db, table, *args, **kw_args), recover,
938
        cacheable, log_level=log_level)
939
    autoanalyze(db, table)
940
    return cur
941 2402 aaronmk
942 3286 aaronmk
def mk_delete(db, table, cond=None):
943
    '''
944
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
945
    @return str query
946
    '''
947
    query = 'DELETE FROM '+table.to_str(db)
948
    if cond != None: query += '\nWHERE '+cond.to_str(db)
949
950
    query = with_explain_comment(db, query)
951
952
    return query
953
954
def delete(db, table, *args, **kw_args):
955
    '''For params, see mk_delete() and run_query()'''
956
    recover = kw_args.pop('recover', None)
957 3295 aaronmk
    cacheable = kw_args.pop('cacheable', True)
958 3286 aaronmk
    log_level = kw_args.pop('log_level', 2)
959
960
    cur = run_query(db, mk_delete(db, table, *args, **kw_args), recover,
961
        cacheable, log_level=log_level)
962
    autoanalyze(db, table)
963
    return cur
964
965 135 aaronmk
def last_insert_id(db):
966 1849 aaronmk
    module = util.root_module(db.db)
967 135 aaronmk
    if module == 'psycopg2': return value(run_query(db, 'SELECT lastval()'))
968
    elif module == 'MySQLdb': return db.insert_id()
969
    else: return None
970 13 aaronmk
971 3490 aaronmk
def define_func(db, def_):
972
    func = def_.function
973
    while True:
974
        try:
975
            run_query(db, def_.to_str(db), recover=True, cacheable=True,
976
                log_ignore_excs=(DuplicateException,))
977
            break # successful
978
        except DuplicateException:
979 3495 aaronmk
            func.name = next_version(func.name)
980 3490 aaronmk
            # try again with next version of name
981
982 2394 aaronmk
def mk_flatten_mapping(db, into, cols, preserve=[], as_items=False):
983 2383 aaronmk
    '''Creates a mapping from original column names (which may have collisions)
984 2415 aaronmk
    to names that will be distinct among the columns' tables.
985 2383 aaronmk
    This is meant to be used for several tables that are being joined together.
986 2415 aaronmk
    @param cols The columns to combine. Duplicates will be removed.
987
    @param into The table for the new columns.
988 2394 aaronmk
    @param preserve [sql_gen.Col...] Columns not to rename. Note that these
989
        columns will be included in the mapping even if they are not in cols.
990
        The tables of the provided Col objects will be changed to into, so make
991
        copies of them if you want to keep the original tables.
992
    @param as_items Whether to return a list of dict items instead of a dict
993 2383 aaronmk
    @return dict(orig_col=new_col, ...)
994
        * orig_col: sql_gen.Col(orig_col_name, orig_table)
995 2392 aaronmk
        * new_col: sql_gen.Col(orig_col_name, into)
996
        * All mappings use the into table so its name can easily be
997 2383 aaronmk
          changed for all columns at once
998
    '''
999 2415 aaronmk
    cols = lists.uniqify(cols)
1000
1001 2394 aaronmk
    items = []
1002 2389 aaronmk
    for col in preserve:
1003 2390 aaronmk
        orig_col = copy.copy(col)
1004 2392 aaronmk
        col.table = into
1005 2394 aaronmk
        items.append((orig_col, col))
1006
    preserve = set(preserve)
1007
    for col in cols:
1008 2716 aaronmk
        if col not in preserve:
1009 3750 aaronmk
            items.append((col, sql_gen.Col(strings.ustr(col), into, col.srcs)))
1010 2394 aaronmk
1011
    if not as_items: items = dict(items)
1012
    return items
1013 2383 aaronmk
1014 2393 aaronmk
def flatten(db, into, joins, cols, limit=None, start=None, **kw_args):
1015 2391 aaronmk
    '''For params, see mk_flatten_mapping()
1016
    @return See return value of mk_flatten_mapping()
1017
    '''
1018 2394 aaronmk
    items = mk_flatten_mapping(db, into, cols, as_items=True, **kw_args)
1019
    cols = [sql_gen.NamedCol(new.name, old) for old, new in items]
1020 3296 aaronmk
    run_query_into(db, mk_select(db, joins, cols, order_by=None, limit=limit,
1021 3305 aaronmk
        start=start), into=into, add_pkey_=True)
1022 3708 aaronmk
        # don't cache because the temp table will usually be truncated after use
1023 2394 aaronmk
    return dict(items)
1024 2391 aaronmk
1025 3079 aaronmk
##### Database structure introspection
1026 2414 aaronmk
1027 3079 aaronmk
#### Tables
1028
1029 4555 aaronmk
def tables(db, schema_like='public', table_like='%', exact=False,
1030
    cacheable=True):
1031 3079 aaronmk
    if exact: compare = '='
1032
    else: compare = 'LIKE'
1033
1034
    module = util.root_module(db.db)
1035
    if module == 'psycopg2':
1036
        conds = [('schemaname', sql_gen.CompareCond(schema_like, compare)),
1037
            ('tablename', sql_gen.CompareCond(table_like, compare))]
1038
        return values(select(db, 'pg_tables', ['tablename'], conds,
1039 4555 aaronmk
            order_by='tablename', cacheable=cacheable, log_level=4))
1040 3079 aaronmk
    elif module == 'MySQLdb':
1041
        return values(run_query(db, 'SHOW TABLES LIKE '+db.esc_value(table_like)
1042
            , cacheable=True, log_level=4))
1043
    else: raise NotImplementedError("Can't list tables for "+module+' database')
1044
1045 4556 aaronmk
def table_exists(db, table, cacheable=True):
1046 3079 aaronmk
    table = sql_gen.as_Table(table)
1047 4556 aaronmk
    return list(tables(db, table.schema, table.name, True, cacheable)) != []
1048 3079 aaronmk
1049 2426 aaronmk
def table_row_count(db, table, recover=None):
1050 2786 aaronmk
    return value(run_query(db, mk_select(db, table, [sql_gen.row_count],
1051 3298 aaronmk
        order_by=None), recover=recover, log_level=3))
1052 2426 aaronmk
1053 5337 aaronmk
def table_col_names(db, table, recover=None):
1054 2414 aaronmk
    return list(col_names(select(db, table, limit=0, order_by=None,
1055 2443 aaronmk
        recover=recover, log_level=4)))
1056 2414 aaronmk
1057 5383 aaronmk
def table_cols(db, table, *args, **kw_args):
1058
    return [sql_gen.as_Col(strings.ustr(c), table)
1059
        for c in table_col_names(db, table, *args, **kw_args)]
1060
1061 5389 aaronmk
def table_pkey_col(db, table, recover=None):
1062 5061 aaronmk
    table = sql_gen.as_Table(table)
1063
1064
    join_cols = ['table_schema', 'table_name', 'constraint_schema',
1065
        'constraint_name']
1066
    tables = [sql_gen.Table('key_column_usage', 'information_schema'),
1067
        sql_gen.Join(sql_gen.Table('table_constraints', 'information_schema'),
1068
            dict(((c, sql_gen.join_same_not_null) for c in join_cols)))]
1069
    cols = [sql_gen.Col('column_name')]
1070
1071
    conds = [('constraint_type', 'PRIMARY KEY'), ('table_name', table.name)]
1072
    schema = table.schema
1073
    if schema != None: conds.append(('table_schema', schema))
1074
    order_by = 'position_in_unique_constraint'
1075
1076 5393 aaronmk
    try: return sql_gen.Col(value(select(db, tables, cols, conds,
1077
        order_by=order_by, limit=1, log_level=4)), table)
1078
    except StopIteration: raise DoesNotExistException('primary key', '')
1079 5389 aaronmk
1080
def pkey_name(db, table, recover=None):
1081
    '''If no pkey, returns the first column in the table.'''
1082 5392 aaronmk
    return pkey_col(db, table, recover).name
1083 832 aaronmk
1084 5390 aaronmk
def pkey_col(db, table, recover=None):
1085 5391 aaronmk
    '''If no pkey, returns the first column in the table.'''
1086 5392 aaronmk
    try: return table_pkey_col(db, table, recover)
1087 5393 aaronmk
    except DoesNotExistException: return table_cols(db, table, recover)[0]
1088 5128 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 5337 aaronmk
    if not_null_col in table_col_names(db, table, recover): return not_null_col
1094 5388 aaronmk
    else: return pkey_name(db, table, recover)
1095 2340 aaronmk
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 5443 aaronmk
    else: raise NotImplementedError("Can't get constraint condition for "
1110
        +module+' database')
1111 3348 aaronmk
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 5367 aaronmk
        return map(sql_gen.parse_expr_col, values(run_query(db, '''\
1121 3322 aaronmk
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 5445 aaronmk
def index_cond(db, index):
1130
    index = sql_gen.as_Table(index)
1131
    module = util.root_module(db.db)
1132
    if module == 'psycopg2':
1133
        qual_index = sql_gen.Literal(index.to_str(db))
1134
        return value(run_query(db, '''\
1135
SELECT pg_get_expr(indpred, indrelid, true)
1136
FROM pg_index
1137
WHERE indexrelid = '''+qual_index.to_str(db)+'''::regclass
1138
'''
1139
            , cacheable=True, log_level=4))
1140
    else: raise NotImplementedError()
1141
1142 3079 aaronmk
#### Functions
1143
1144
def function_exists(db, function):
1145 3423 aaronmk
    qual_function = sql_gen.Literal(function.to_str(db))
1146
    try:
1147 3425 aaronmk
        select(db, fields=[sql_gen.Cast('regproc', qual_function)],
1148
            recover=True, cacheable=True, log_level=4)
1149 3423 aaronmk
    except DoesNotExistException: return False
1150 4146 aaronmk
    except DuplicateException: return True # overloaded function
1151 3423 aaronmk
    else: return True
1152 3079 aaronmk
1153
##### Structural changes
1154
1155
#### Columns
1156
1157 5020 aaronmk
def add_col(db, table, col, comment=None, if_not_exists=False, **kw_args):
1158 3079 aaronmk
    '''
1159
    @param col TypedCol Name may be versioned, so be sure to propagate any
1160
        renaming back to any source column for the TypedCol.
1161
    @param comment None|str SQL comment used to distinguish columns of the same
1162
        name from each other when they contain different data, to allow the
1163
        ADD COLUMN query to be cached. If not set, query will not be cached.
1164
    '''
1165
    assert isinstance(col, sql_gen.TypedCol)
1166
1167
    while True:
1168
        str_ = 'ALTER TABLE '+table.to_str(db)+' ADD COLUMN '+col.to_str(db)
1169
        if comment != None: str_ += ' '+sql_gen.esc_comment(comment)
1170
1171
        try:
1172
            run_query(db, str_, recover=True, cacheable=True, **kw_args)
1173
            break
1174
        except DuplicateException:
1175 5020 aaronmk
            if if_not_exists: raise
1176 3079 aaronmk
            col.name = next_version(col.name)
1177
            # try again with next version of name
1178
1179
def add_not_null(db, col):
1180
    table = col.table
1181
    col = sql_gen.to_name_only_col(col)
1182
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ALTER COLUMN '
1183
        +col.to_str(db)+' SET NOT NULL', cacheable=True, log_level=3)
1184
1185 4443 aaronmk
def drop_not_null(db, col):
1186
    table = col.table
1187
    col = sql_gen.to_name_only_col(col)
1188
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ALTER COLUMN '
1189
        +col.to_str(db)+' DROP NOT NULL', cacheable=True, log_level=3)
1190
1191 2096 aaronmk
row_num_col = '_row_num'
1192
1193 4997 aaronmk
row_num_col_def = sql_gen.TypedCol('', 'serial', nullable=False,
1194 3079 aaronmk
    constraints='PRIMARY KEY')
1195
1196 4997 aaronmk
def add_row_num(db, table, name=row_num_col):
1197
    '''Adds a row number column to a table. Its definition is in
1198
    row_num_col_def. It will be the primary key.'''
1199
    col_def = copy.copy(row_num_col_def)
1200
    col_def.name = name
1201 5021 aaronmk
    add_col(db, table, col_def, comment='', if_not_exists=True, log_level=3)
1202 3079 aaronmk
1203
#### Indexes
1204
1205
def add_pkey(db, table, cols=None, recover=None):
1206
    '''Adds a primary key.
1207
    @param cols [sql_gen.Col,...] The columns in the primary key.
1208
        Defaults to the first column in the table.
1209
    @pre The table must not already have a primary key.
1210
    '''
1211
    table = sql_gen.as_Table(table)
1212 5388 aaronmk
    if cols == None: cols = [pkey_name(db, table, recover)]
1213 3079 aaronmk
    col_strs = [sql_gen.to_name_only_col(v).to_str(db) for v in cols]
1214
1215
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ADD PRIMARY KEY ('
1216
        +(', '.join(col_strs))+')', recover=True, cacheable=True, log_level=3,
1217
        log_ignore_excs=(DuplicateException,))
1218
1219 2998 aaronmk
def add_index(db, exprs, table=None, unique=False, ensure_not_null_=True):
1220 2688 aaronmk
    '''Adds an index on column(s) or expression(s) if it doesn't already exist.
1221 3356 aaronmk
    Currently, only function calls and literal values are supported expressions.
1222 2998 aaronmk
    @param ensure_not_null_ If set, translates NULL values to sentinel values.
1223 2847 aaronmk
        This allows indexes to be used for comparisons where NULLs are equal.
1224 2538 aaronmk
    '''
1225 2964 aaronmk
    exprs = lists.mk_seq(exprs)
1226 2538 aaronmk
1227 2688 aaronmk
    # Parse exprs
1228
    old_exprs = exprs[:]
1229
    exprs = []
1230
    cols = []
1231
    for i, expr in enumerate(old_exprs):
1232 2823 aaronmk
        expr = sql_gen.as_Col(expr, table)
1233 2688 aaronmk
1234 2823 aaronmk
        # Handle nullable columns
1235 2998 aaronmk
        if ensure_not_null_:
1236 3164 aaronmk
            try: expr = sql_gen.ensure_not_null(db, expr)
1237 2860 aaronmk
            except KeyError: pass # unknown type, so just create plain index
1238 2823 aaronmk
1239 2688 aaronmk
        # Extract col
1240 3002 aaronmk
        expr = copy.deepcopy(expr) # don't modify input!
1241 3356 aaronmk
        col = expr
1242
        if isinstance(expr, sql_gen.FunctionCall): col = expr.args[0]
1243
        expr = sql_gen.cast_literal(expr)
1244
        if not isinstance(expr, (sql_gen.Expr, sql_gen.Col)):
1245 2688 aaronmk
            expr = sql_gen.Expr(expr)
1246 3356 aaronmk
1247 2688 aaronmk
1248
        # Extract table
1249
        if table == None:
1250
            assert sql_gen.is_table_col(col)
1251
            table = col.table
1252
1253 3356 aaronmk
        if isinstance(col, sql_gen.Col): col.table = None
1254 2688 aaronmk
1255
        exprs.append(expr)
1256
        cols.append(col)
1257 2408 aaronmk
1258 2688 aaronmk
    table = sql_gen.as_Table(table)
1259
1260 3005 aaronmk
    # Add index
1261 3148 aaronmk
    str_ = 'CREATE'
1262
    if unique: str_ += ' UNIQUE'
1263
    str_ += ' INDEX ON '+table.to_str(db)+' ('+(
1264
        ', '.join((v.to_str(db) for v in exprs)))+')'
1265
    run_query(db, str_, recover=True, cacheable=True, log_level=3)
1266 2408 aaronmk
1267 3083 aaronmk
already_indexed = object() # tells add_indexes() the pkey has already been added
1268
1269
def add_indexes(db, table, has_pkey=True):
1270
    '''Adds an index on all columns in a table.
1271
    @param has_pkey bool|already_indexed Whether a pkey instead of a regular
1272
        index should be added on the first column.
1273
        * If already_indexed, the pkey is assumed to have already been added
1274
    '''
1275 5337 aaronmk
    cols = table_col_names(db, table)
1276 3083 aaronmk
    if has_pkey:
1277
        if has_pkey is not already_indexed: add_pkey(db, table)
1278
        cols = cols[1:]
1279
    for col in cols: add_index(db, col, table)
1280
1281 3079 aaronmk
#### Tables
1282 2772 aaronmk
1283 3079 aaronmk
### Maintenance
1284 2772 aaronmk
1285 3079 aaronmk
def analyze(db, table):
1286
    table = sql_gen.as_Table(table)
1287
    run_query(db, 'ANALYZE '+table.to_str(db), log_level=3)
1288 2934 aaronmk
1289 3079 aaronmk
def autoanalyze(db, table):
1290
    if db.autoanalyze: analyze(db, table)
1291 2935 aaronmk
1292 3079 aaronmk
def vacuum(db, table):
1293
    table = sql_gen.as_Table(table)
1294
    db.with_autocommit(lambda: run_query(db, 'VACUUM ANALYZE '+table.to_str(db),
1295
        log_level=3))
1296 2086 aaronmk
1297 3079 aaronmk
### Lifecycle
1298
1299 3247 aaronmk
def drop(db, type_, name):
1300
    name = sql_gen.as_Name(name)
1301
    run_query(db, 'DROP '+type_+' IF EXISTS '+name.to_str(db)+' CASCADE')
1302 2889 aaronmk
1303 3247 aaronmk
def drop_table(db, table): drop(db, 'TABLE', table)
1304
1305 3082 aaronmk
def create_table(db, table, cols=[], has_pkey=True, col_indexes=True,
1306
    like=None):
1307 2675 aaronmk
    '''Creates a table.
1308 2681 aaronmk
    @param cols [sql_gen.TypedCol,...] The column names and types
1309
    @param has_pkey If set, the first column becomes the primary key.
1310 2760 aaronmk
    @param col_indexes bool|[ref]
1311
        * If True, indexes will be added on all non-pkey columns.
1312
        * If a list reference, [0] will be set to a function to do this.
1313
          This can be used to delay index creation until the table is populated.
1314 2675 aaronmk
    '''
1315
    table = sql_gen.as_Table(table)
1316
1317 3082 aaronmk
    if like != None:
1318
        cols = [sql_gen.CustomCode('LIKE '+like.to_str(db)+' INCLUDING ALL')
1319
            ]+cols
1320 2681 aaronmk
    if has_pkey:
1321
        cols[0] = pkey = copy.copy(cols[0]) # don't modify input!
1322 2872 aaronmk
        pkey.constraints = 'PRIMARY KEY'
1323 2681 aaronmk
1324 3085 aaronmk
    temp = table.is_temp and not db.debug_temp
1325
        # temp tables permanent in debug_temp mode
1326 2760 aaronmk
1327 3085 aaronmk
    # Create table
1328 3383 aaronmk
    def create():
1329 3085 aaronmk
        str_ = 'CREATE'
1330
        if temp: str_ += ' TEMP'
1331
        str_ += ' TABLE '+table.to_str(db)+' (\n'
1332
        str_ += '\n, '.join(c.to_str(db) for c in cols)
1333 3126 aaronmk
        str_ += '\n);'
1334 3085 aaronmk
1335 3383 aaronmk
        run_query(db, str_, recover=True, cacheable=True, log_level=2,
1336
            log_ignore_excs=(DuplicateException,))
1337
    if table.is_temp:
1338
        while True:
1339
            try:
1340
                create()
1341
                break
1342
            except DuplicateException:
1343
                table.name = next_version(table.name)
1344
                # try again with next version of name
1345
    else: create()
1346 3085 aaronmk
1347 2760 aaronmk
    # Add indexes
1348 2773 aaronmk
    if has_pkey: has_pkey = already_indexed
1349
    def add_indexes_(): add_indexes(db, table, has_pkey)
1350
    if isinstance(col_indexes, list): col_indexes[0] = add_indexes_ # defer
1351
    elif col_indexes: add_indexes_() # add now
1352 2675 aaronmk
1353 3084 aaronmk
def copy_table_struct(db, src, dest):
1354
    '''Creates a structure-only copy of a table. (Does not copy data.)'''
1355 3085 aaronmk
    create_table(db, dest, has_pkey=False, col_indexes=False, like=src)
1356 3084 aaronmk
1357 3079 aaronmk
### Data
1358 2684 aaronmk
1359 2970 aaronmk
def truncate(db, table, schema='public', **kw_args):
1360
    '''For params, see run_query()'''
1361 2777 aaronmk
    table = sql_gen.as_Table(table, schema)
1362 2970 aaronmk
    return run_query(db, 'TRUNCATE '+table.to_str(db)+' CASCADE', **kw_args)
1363 2732 aaronmk
1364 2965 aaronmk
def empty_temp(db, tables):
1365
    tables = lists.mk_seq(tables)
1366 2971 aaronmk
    for table in tables: truncate(db, table, log_level=3)
1367 2965 aaronmk
1368 1968 aaronmk
def empty_db(db, schema='public', **kw_args):
1369
    '''For kw_args, see tables()'''
1370
    for table in tables(db, schema, **kw_args): truncate(db, table, schema)
1371 3094 aaronmk
1372
def distinct_table(db, table, distinct_on):
1373
    '''Creates a copy of a temp table which is distinct on the given columns.
1374 3099 aaronmk
    The old and new tables will both get an index on these columns, to
1375
    facilitate merge joins.
1376 3097 aaronmk
    @param distinct_on If empty, creates a table with one row. This is useful if
1377
        your distinct_on columns are all literal values.
1378 3099 aaronmk
    @return The new table.
1379 3094 aaronmk
    '''
1380 3099 aaronmk
    new_table = sql_gen.suffixed_table(table, '_distinct')
1381 3411 aaronmk
    distinct_on = filter(sql_gen.is_table_col, distinct_on)
1382 3094 aaronmk
1383 3099 aaronmk
    copy_table_struct(db, table, new_table)
1384 3097 aaronmk
1385
    limit = None
1386
    if distinct_on == []: limit = 1 # one sample row
1387 3099 aaronmk
    else:
1388
        add_index(db, distinct_on, new_table, unique=True)
1389
        add_index(db, distinct_on, table) # for join optimization
1390 3097 aaronmk
1391 3313 aaronmk
    insert_select(db, new_table, None, mk_select(db, table, order_by=None,
1392
        limit=limit), ignore=True)
1393 3099 aaronmk
    analyze(db, new_table)
1394 3094 aaronmk
1395 3099 aaronmk
    return new_table