Project

General

Profile

1 11 aaronmk
# Database access
2
3 1869 aaronmk
import copy
4 11 aaronmk
import re
5 3238 aaronmk
import time
6 865 aaronmk
import warnings
7 11 aaronmk
8 300 aaronmk
import exc
9 1909 aaronmk
import dicts
10 1893 aaronmk
import iters
11 1960 aaronmk
import lists
12 3241 aaronmk
import profiling
13 1889 aaronmk
from Proxy import Proxy
14 1872 aaronmk
import rand
15 2217 aaronmk
import sql_gen
16 862 aaronmk
import strings
17 131 aaronmk
import util
18 11 aaronmk
19 832 aaronmk
##### Exceptions
20
21 2804 aaronmk
def get_cur_query(cur, input_query=None):
22 2168 aaronmk
    raw_query = None
23
    if hasattr(cur, 'query'): raw_query = cur.query
24
    elif hasattr(cur, '_last_executed'): raw_query = cur._last_executed
25 2170 aaronmk
26
    if raw_query != None: return raw_query
27 2804 aaronmk
    else: return '[input] '+strings.ustr(input_query)
28 14 aaronmk
29 2170 aaronmk
def _add_cursor_info(e, *args, **kw_args):
30
    '''For params, see get_cur_query()'''
31 2771 aaronmk
    exc.add_msg(e, 'query: '+strings.ustr(get_cur_query(*args, **kw_args)))
32 135 aaronmk
33 300 aaronmk
class DbException(exc.ExceptionWithCause):
34 14 aaronmk
    def __init__(self, msg, cause=None, cur=None):
35 2145 aaronmk
        exc.ExceptionWithCause.__init__(self, msg, cause, cause_newline=True)
36 14 aaronmk
        if cur != None: _add_cursor_info(self, cur)
37
38 2143 aaronmk
class ExceptionWithName(DbException):
39
    def __init__(self, name, cause=None):
40 2484 aaronmk
        DbException.__init__(self, 'for name: '+strings.as_tt(str(name)), cause)
41 2143 aaronmk
        self.name = name
42 360 aaronmk
43 3109 aaronmk
class ExceptionWithValue(DbException):
44
    def __init__(self, value, cause=None):
45
        DbException.__init__(self, 'for value: '+strings.as_tt(repr(value)),
46
            cause)
47 2240 aaronmk
        self.value = value
48
49 2945 aaronmk
class ExceptionWithNameType(DbException):
50
    def __init__(self, type_, name, cause=None):
51
        DbException.__init__(self, 'for type: '+strings.as_tt(str(type_))
52
            +'; name: '+strings.as_tt(name), cause)
53
        self.type = type_
54
        self.name = name
55
56 2306 aaronmk
class ConstraintException(DbException):
57
    def __init__(self, name, cols, cause=None):
58 2484 aaronmk
        DbException.__init__(self, 'Violated '+strings.as_tt(name)
59
            +' constraint on columns: '+strings.as_tt(', '.join(cols)), cause)
60 2306 aaronmk
        self.name = name
61 468 aaronmk
        self.cols = cols
62 11 aaronmk
63 2523 aaronmk
class MissingCastException(DbException):
64
    def __init__(self, type_, col, cause=None):
65
        DbException.__init__(self, 'Missing cast to type '+strings.as_tt(type_)
66
            +' on column: '+strings.as_tt(col), cause)
67
        self.type = type_
68
        self.col = col
69
70 2143 aaronmk
class NameException(DbException): pass
71
72 2306 aaronmk
class DuplicateKeyException(ConstraintException): pass
73 13 aaronmk
74 2306 aaronmk
class NullValueException(ConstraintException): pass
75 13 aaronmk
76 3109 aaronmk
class InvalidValueException(ExceptionWithValue): pass
77 2239 aaronmk
78 2945 aaronmk
class DuplicateException(ExceptionWithNameType): pass
79 2143 aaronmk
80 89 aaronmk
class EmptyRowException(DbException): pass
81
82 865 aaronmk
##### Warnings
83
84
class DbWarning(UserWarning): pass
85
86 1930 aaronmk
##### Result retrieval
87
88
def col_names(cur): return (col[0] for col in cur.description)
89
90
def rows(cur): return iter(lambda: cur.fetchone(), None)
91
92
def consume_rows(cur):
93
    '''Used to fetch all rows so result will be cached'''
94
    iters.consume_iter(rows(cur))
95
96
def next_row(cur): return rows(cur).next()
97
98
def row(cur):
99
    row_ = next_row(cur)
100
    consume_rows(cur)
101
    return row_
102
103
def next_value(cur): return next_row(cur)[0]
104
105
def value(cur): return row(cur)[0]
106
107
def values(cur): return iters.func_iter(lambda: next_value(cur))
108
109
def value_or_none(cur):
110
    try: return value(cur)
111
    except StopIteration: return None
112
113 2762 aaronmk
##### Escaping
114 2101 aaronmk
115 2573 aaronmk
def esc_name_by_module(module, name):
116
    if module == 'psycopg2' or module == None: quote = '"'
117 2101 aaronmk
    elif module == 'MySQLdb': quote = '`'
118
    else: raise NotImplementedError("Can't escape name for "+module+' database')
119 2500 aaronmk
    return sql_gen.esc_name(name, quote)
120 2101 aaronmk
121
def esc_name_by_engine(engine, name, **kw_args):
122
    return esc_name_by_module(db_engines[engine][0], name, **kw_args)
123
124
def esc_name(db, name, **kw_args):
125
    return esc_name_by_module(util.root_module(db.db), name, **kw_args)
126
127
def qual_name(db, schema, table):
128
    def esc_name_(name): return esc_name(db, name)
129
    table = esc_name_(table)
130
    if schema != None: return esc_name_(schema)+'.'+table
131
    else: return table
132
133 1869 aaronmk
##### Database connections
134 1849 aaronmk
135 2097 aaronmk
db_config_names = ['engine', 'host', 'user', 'password', 'database', 'schemas']
136 1926 aaronmk
137 1869 aaronmk
db_engines = {
138
    'MySQL': ('MySQLdb', {'password': 'passwd', 'database': 'db'}),
139
    'PostgreSQL': ('psycopg2', {}),
140
}
141
142
DatabaseErrors_set = set([DbException])
143
DatabaseErrors = tuple(DatabaseErrors_set)
144
145
def _add_module(module):
146
    DatabaseErrors_set.add(module.DatabaseError)
147
    global DatabaseErrors
148
    DatabaseErrors = tuple(DatabaseErrors_set)
149
150
def db_config_str(db_config):
151
    return db_config['engine']+' database '+db_config['database']
152
153 2448 aaronmk
log_debug_none = lambda msg, level=2: None
154 1901 aaronmk
155 1849 aaronmk
class DbConn:
156 2923 aaronmk
    def __init__(self, db_config, autocommit=True, caching=True,
157 3183 aaronmk
        log_debug=log_debug_none, debug_temp=False, src=None):
158 2915 aaronmk
        '''
159
        @param debug_temp Whether temporary objects should instead be permanent.
160
            This assists in debugging the internal objects used by the program.
161 3183 aaronmk
        @param src In autocommit mode, will be included in a comment in every
162
            query, to help identify the data source in pg_stat_activity.
163 2915 aaronmk
        '''
164 1869 aaronmk
        self.db_config = db_config
165 2190 aaronmk
        self.autocommit = autocommit
166
        self.caching = caching
167 1901 aaronmk
        self.log_debug = log_debug
168 2193 aaronmk
        self.debug = log_debug != log_debug_none
169 2915 aaronmk
        self.debug_temp = debug_temp
170 3244 aaronmk
        self.profile_row_ct = None
171 3183 aaronmk
        self.src = src
172 3074 aaronmk
        self.autoanalyze = False
173 1869 aaronmk
174 3124 aaronmk
        self._savepoint = 0
175 3120 aaronmk
        self._reset()
176 1869 aaronmk
177
    def __getattr__(self, name):
178
        if name == '__dict__': raise Exception('getting __dict__')
179
        if name == 'db': return self._db()
180
        else: raise AttributeError()
181
182
    def __getstate__(self):
183
        state = copy.copy(self.__dict__) # shallow copy
184 1915 aaronmk
        state['log_debug'] = None # don't pickle the debug callback
185 1869 aaronmk
        state['_DbConn__db'] = None # don't pickle the connection
186
        return state
187
188 3118 aaronmk
    def clear_cache(self): self.query_results = {}
189
190 3120 aaronmk
    def _reset(self):
191 3118 aaronmk
        self.clear_cache()
192 3124 aaronmk
        assert self._savepoint == 0
193 3118 aaronmk
        self._notices_seen = set()
194
        self.__db = None
195
196 2165 aaronmk
    def connected(self): return self.__db != None
197
198 3116 aaronmk
    def close(self):
199 3119 aaronmk
        if not self.connected(): return
200
201 3135 aaronmk
        # Record that the automatic transaction is now closed
202 3136 aaronmk
        self._savepoint -= 1
203 3135 aaronmk
204 3119 aaronmk
        self.db.close()
205 3120 aaronmk
        self._reset()
206 3116 aaronmk
207 3125 aaronmk
    def reconnect(self):
208
        # Do not do this in test mode as it would roll back everything
209
        if self.autocommit: self.close()
210
        # Connection will be reopened automatically on first query
211
212 1869 aaronmk
    def _db(self):
213
        if self.__db == None:
214
            # Process db_config
215
            db_config = self.db_config.copy() # don't modify input!
216 2097 aaronmk
            schemas = db_config.pop('schemas', None)
217 1869 aaronmk
            module_name, mappings = db_engines[db_config.pop('engine')]
218
            module = __import__(module_name)
219
            _add_module(module)
220
            for orig, new in mappings.iteritems():
221
                try: util.rename_key(db_config, orig, new)
222
                except KeyError: pass
223
224
            # Connect
225
            self.__db = module.connect(**db_config)
226
227 3161 aaronmk
            # Record that a transaction is already open
228
            self._savepoint += 1
229
230 1869 aaronmk
            # Configure connection
231 2906 aaronmk
            if hasattr(self.db, 'set_isolation_level'):
232
                import psycopg2.extensions
233
                self.db.set_isolation_level(
234
                    psycopg2.extensions.ISOLATION_LEVEL_READ_COMMITTED)
235 2101 aaronmk
            if schemas != None:
236 2893 aaronmk
                search_path = [self.esc_name(s) for s in schemas.split(',')]
237
                search_path.append(value(run_query(self, 'SHOW search_path',
238
                    log_level=4)))
239
                run_query(self, 'SET search_path TO '+(','.join(search_path)),
240
                    log_level=3)
241 1869 aaronmk
242
        return self.__db
243 1889 aaronmk
244 1891 aaronmk
    class DbCursor(Proxy):
245 1927 aaronmk
        def __init__(self, outer):
246 1891 aaronmk
            Proxy.__init__(self, outer.db.cursor())
247 2191 aaronmk
            self.outer = outer
248 1927 aaronmk
            self.query_results = outer.query_results
249 1894 aaronmk
            self.query_lookup = None
250 1891 aaronmk
            self.result = []
251 1889 aaronmk
252 2802 aaronmk
        def execute(self, query):
253 2764 aaronmk
            self._is_insert = query.startswith('INSERT')
254 2797 aaronmk
            self.query_lookup = query
255 2148 aaronmk
            try:
256 3162 aaronmk
                try: cur = self.inner.execute(query)
257 2802 aaronmk
                finally: self.query = get_cur_query(self.inner, query)
258 1904 aaronmk
            except Exception, e:
259
                self.result = e # cache the exception as the result
260
                self._cache_result()
261
                raise
262 3004 aaronmk
263
            # Always cache certain queries
264 3183 aaronmk
            query = sql_gen.lstrip(query)
265 3004 aaronmk
            if query.startswith('CREATE') or query.startswith('ALTER'):
266 3007 aaronmk
                # structural changes
267 3040 aaronmk
                # Rest of query must be unique in the face of name collisions,
268
                # so don't cache ADD COLUMN unless it has distinguishing comment
269
                if query.find('ADD COLUMN') < 0 or query.endswith('*/'):
270 3007 aaronmk
                    self._cache_result()
271 3004 aaronmk
            elif self.rowcount == 0 and query.startswith('SELECT'): # empty
272 2800 aaronmk
                consume_rows(self) # fetch all rows so result will be cached
273 3004 aaronmk
274 2762 aaronmk
            return cur
275 1894 aaronmk
276 1891 aaronmk
        def fetchone(self):
277
            row = self.inner.fetchone()
278 1899 aaronmk
            if row != None: self.result.append(row)
279
            # otherwise, fetched all rows
280 1904 aaronmk
            else: self._cache_result()
281
            return row
282
283
        def _cache_result(self):
284 2948 aaronmk
            # For inserts that return a result set, don't cache result set since
285
            # inserts are not idempotent. Other non-SELECT queries don't have
286
            # their result set read, so only exceptions will be cached (an
287
            # invalid query will always be invalid).
288 1930 aaronmk
            if self.query_results != None and (not self._is_insert
289 1906 aaronmk
                or isinstance(self.result, Exception)):
290
291 1894 aaronmk
                assert self.query_lookup != None
292 1916 aaronmk
                self.query_results[self.query_lookup] = self.CacheCursor(
293
                    util.dict_subset(dicts.AttrsDictView(self),
294
                    ['query', 'result', 'rowcount', 'description']))
295 1906 aaronmk
296 1916 aaronmk
        class CacheCursor:
297
            def __init__(self, cached_result): self.__dict__ = cached_result
298
299 1927 aaronmk
            def execute(self, *args, **kw_args):
300 1916 aaronmk
                if isinstance(self.result, Exception): raise self.result
301
                # otherwise, result is a rows list
302
                self.iter = iter(self.result)
303
304
            def fetchone(self):
305
                try: return self.iter.next()
306
                except StopIteration: return None
307 1891 aaronmk
308 2212 aaronmk
    def esc_value(self, value):
309 2663 aaronmk
        try: str_ = self.mogrify('%s', [value])
310
        except NotImplementedError, e:
311
            module = util.root_module(self.db)
312
            if module == 'MySQLdb':
313
                import _mysql
314
                str_ = _mysql.escape_string(value)
315
            else: raise e
316 2374 aaronmk
        return strings.to_unicode(str_)
317 2212 aaronmk
318 2347 aaronmk
    def esc_name(self, name): return esc_name(self, name) # calls global func
319
320 2814 aaronmk
    def std_code(self, str_):
321
        '''Standardizes SQL code.
322
        * Ensures that string literals are prefixed by `E`
323
        '''
324
        if str_.startswith("'"): str_ = 'E'+str_
325
        return str_
326
327 2665 aaronmk
    def can_mogrify(self):
328 2663 aaronmk
        module = util.root_module(self.db)
329 2665 aaronmk
        return module == 'psycopg2'
330 2663 aaronmk
331 2665 aaronmk
    def mogrify(self, query, params=None):
332
        if self.can_mogrify(): return self.db.cursor().mogrify(query, params)
333
        else: raise NotImplementedError("Can't mogrify query")
334
335 2671 aaronmk
    def print_notices(self):
336 2725 aaronmk
        if hasattr(self.db, 'notices'):
337
            for msg in self.db.notices:
338
                if msg not in self._notices_seen:
339
                    self._notices_seen.add(msg)
340
                    self.log_debug(msg, level=2)
341 2671 aaronmk
342 2793 aaronmk
    def run_query(self, query, cacheable=False, log_level=2,
343 2464 aaronmk
        debug_msg_ref=None):
344 2445 aaronmk
        '''
345 2464 aaronmk
        @param log_ignore_excs The log_level will be increased by 2 if the query
346
            throws one of these exceptions.
347 2664 aaronmk
        @param debug_msg_ref If specified, the log message will be returned in
348
            this instead of being output. This allows you to filter log messages
349
            depending on the result of the query.
350 2445 aaronmk
        '''
351 2167 aaronmk
        assert query != None
352
353 3183 aaronmk
        if self.autocommit and self.src != None:
354 3206 aaronmk
            query = sql_gen.esc_comment(self.src)+'\t'+query
355 3183 aaronmk
356 2047 aaronmk
        if not self.caching: cacheable = False
357 1903 aaronmk
        used_cache = False
358 2664 aaronmk
359 3242 aaronmk
        if self.debug:
360
            profiler = profiling.ItersProfiler(start_now=True, iter_text='row')
361 1903 aaronmk
        try:
362 1927 aaronmk
            # Get cursor
363
            if cacheable:
364 3238 aaronmk
                try: cur = self.query_results[query]
365 1927 aaronmk
                except KeyError: cur = self.DbCursor(self)
366 3238 aaronmk
                else: used_cache = True
367 1927 aaronmk
            else: cur = self.db.cursor()
368
369
            # Run query
370 3238 aaronmk
            try: cur.execute(query)
371 3162 aaronmk
            except Exception, e:
372
                _add_cursor_info(e, self, query)
373
                raise
374 3238 aaronmk
            else: self.do_autocommit()
375 1903 aaronmk
        finally:
376 3242 aaronmk
            if self.debug:
377 3244 aaronmk
                profiler.stop(self.profile_row_ct)
378 3242 aaronmk
379
                ## Log or return query
380
381 3239 aaronmk
                query = str(get_cur_query(cur, query))
382
                # So that the src comment is hidden when cating the file, and
383
                # put on a separate line when viewed in a text editor.
384
                query = query.replace('\t', '\r', 1)
385
386 3240 aaronmk
                msg = 'DB query: '
387 3239 aaronmk
388 3240 aaronmk
                if used_cache: msg += 'cache hit'
389
                elif cacheable: msg += 'cache miss'
390
                else: msg += 'non-cacheable'
391 3239 aaronmk
392 3241 aaronmk
                msg += ':\n'+profiler.msg()+'\n'+strings.as_code(query, 'SQL')
393 3240 aaronmk
394 3237 aaronmk
                if debug_msg_ref != None: debug_msg_ref[0] = msg
395
                else: self.log_debug(msg, log_level)
396 3242 aaronmk
            self.print_notices()
397 1903 aaronmk
398
        return cur
399 1914 aaronmk
400 2797 aaronmk
    def is_cached(self, query): return query in self.query_results
401 2139 aaronmk
402 2907 aaronmk
    def with_autocommit(self, func):
403 2801 aaronmk
        import psycopg2.extensions
404
405
        prev_isolation_level = self.db.isolation_level
406 2907 aaronmk
        self.db.set_isolation_level(
407
            psycopg2.extensions.ISOLATION_LEVEL_AUTOCOMMIT)
408 2683 aaronmk
        try: return func()
409 2801 aaronmk
        finally: self.db.set_isolation_level(prev_isolation_level)
410 2683 aaronmk
411 2139 aaronmk
    def with_savepoint(self, func):
412 3137 aaronmk
        top = self._savepoint == 0
413 2171 aaronmk
        savepoint = 'level_'+str(self._savepoint)
414 3137 aaronmk
415 3160 aaronmk
        # Must happen before running queries so they don't get autocommitted
416
        self._savepoint += 1
417
418 3137 aaronmk
        if top: query = 'START TRANSACTION ISOLATION LEVEL READ COMMITTED'
419
        else: query = 'SAVEPOINT '+savepoint
420
        self.run_query(query, log_level=4)
421
        try:
422
            return func()
423
            if top: self.run_query('COMMIT', log_level=4)
424 2139 aaronmk
        except:
425 3137 aaronmk
            if top: query = 'ROLLBACK'
426
            else: query = 'ROLLBACK TO SAVEPOINT '+savepoint
427
            self.run_query(query, log_level=4)
428
429 2139 aaronmk
            raise
430 2930 aaronmk
        finally:
431
            # Always release savepoint, because after ROLLBACK TO SAVEPOINT,
432
            # "The savepoint remains valid and can be rolled back to again"
433
            # (http://www.postgresql.org/docs/8.3/static/sql-rollback-to.html).
434 3137 aaronmk
            if not top:
435
                self.run_query('RELEASE SAVEPOINT '+savepoint, log_level=4)
436 2930 aaronmk
437
            self._savepoint -= 1
438
            assert self._savepoint >= 0
439
440
            self.do_autocommit() # OK to do this after ROLLBACK TO SAVEPOINT
441 2191 aaronmk
442
    def do_autocommit(self):
443
        '''Autocommits if outside savepoint'''
444 3135 aaronmk
        assert self._savepoint >= 1
445
        if self.autocommit and self._savepoint == 1:
446 2924 aaronmk
            self.log_debug('Autocommitting', level=4)
447 2191 aaronmk
            self.db.commit()
448 2643 aaronmk
449 3155 aaronmk
    def col_info(self, col, cacheable=True):
450 2643 aaronmk
        table = sql_gen.Table('columns', 'information_schema')
451 3063 aaronmk
        type_ = sql_gen.Coalesce(sql_gen.Nullif(sql_gen.Col('data_type'),
452
            'USER-DEFINED'), sql_gen.Col('udt_name'))
453 3078 aaronmk
        cols = [type_, 'column_default',
454
            sql_gen.Cast('boolean', sql_gen.Col('is_nullable'))]
455 2643 aaronmk
456
        conds = [('table_name', col.table.name), ('column_name', col.name)]
457
        schema = col.table.schema
458
        if schema != None: conds.append(('table_schema', schema))
459
460 2819 aaronmk
        type_, default, nullable = row(select(self, table, cols, conds,
461 3150 aaronmk
            order_by='table_schema', limit=1, cacheable=cacheable, log_level=4))
462 2643 aaronmk
            # TODO: order_by search_path schema order
463 2819 aaronmk
        default = sql_gen.as_Code(default, self)
464
465
        return sql_gen.TypedCol(col.name, type_, default, nullable)
466 2917 aaronmk
467
    def TempFunction(self, name):
468
        if self.debug_temp: schema = None
469
        else: schema = 'pg_temp'
470
        return sql_gen.Function(name, schema)
471 1849 aaronmk
472 1869 aaronmk
connect = DbConn
473
474 832 aaronmk
##### Recoverable querying
475 15 aaronmk
476 2139 aaronmk
def with_savepoint(db, func): return db.with_savepoint(func)
477 11 aaronmk
478 2791 aaronmk
def run_query(db, query, recover=None, cacheable=False, log_level=2,
479
    log_ignore_excs=None, **kw_args):
480 2794 aaronmk
    '''For params, see DbConn.run_query()'''
481 830 aaronmk
    if recover == None: recover = False
482 2464 aaronmk
    if log_ignore_excs == None: log_ignore_excs = ()
483
    log_ignore_excs = tuple(log_ignore_excs)
484 3236 aaronmk
    debug_msg_ref = [None]
485 830 aaronmk
486 2148 aaronmk
    try:
487 2464 aaronmk
        try:
488 2794 aaronmk
            def run(): return db.run_query(query, cacheable, log_level,
489 2793 aaronmk
                debug_msg_ref, **kw_args)
490 2796 aaronmk
            if recover and not db.is_cached(query):
491 2464 aaronmk
                return with_savepoint(db, run)
492
            else: return run() # don't need savepoint if cached
493
        except Exception, e:
494 3095 aaronmk
            msg = strings.ustr(e.args[0])
495 2464 aaronmk
496 3095 aaronmk
            match = re.match(r'^duplicate key value violates unique constraint '
497 3096 aaronmk
                r'"((_?[^\W_]+(?=[._]))?.+?)"', msg)
498 2464 aaronmk
            if match:
499
                constraint, table = match.groups()
500 3025 aaronmk
                cols = []
501
                if recover: # need auto-rollback to run index_cols()
502
                    try: cols = index_cols(db, table, constraint)
503
                    except NotImplementedError: pass
504
                raise DuplicateKeyException(constraint, cols, e)
505 2464 aaronmk
506 3095 aaronmk
            match = re.match(r'^null value in column "(.+?)" violates not-null'
507 2464 aaronmk
                r' constraint', msg)
508
            if match: raise NullValueException('NOT NULL', [match.group(1)], e)
509
510 3095 aaronmk
            match = re.match(r'^(?:invalid input (?:syntax|value)\b.*?'
511 3109 aaronmk
                r'|.+? field value out of range): "(.+?)"', msg)
512 2464 aaronmk
            if match:
513 3109 aaronmk
                value, = match.groups()
514
                raise InvalidValueException(strings.to_unicode(value), e)
515 2464 aaronmk
516 3095 aaronmk
            match = re.match(r'^column "(.+?)" is of type (.+?) but expression '
517 2523 aaronmk
                r'is of type', msg)
518
            if match:
519
                col, type_ = match.groups()
520
                raise MissingCastException(type_, col, e)
521
522 3095 aaronmk
            match = re.match(r'^(\S+) "(.+?)".*? already exists', msg)
523 2945 aaronmk
            if match:
524
                type_, name = match.groups()
525
                raise DuplicateException(type_, name, e)
526 2464 aaronmk
527
            raise # no specific exception raised
528
    except log_ignore_excs:
529
        log_level += 2
530
        raise
531
    finally:
532 3236 aaronmk
        if debug_msg_ref[0] != None: db.log_debug(debug_msg_ref[0], log_level)
533 830 aaronmk
534 832 aaronmk
##### Basic queries
535
536 2153 aaronmk
def next_version(name):
537 2163 aaronmk
    version = 1 # first existing name was version 0
538 2586 aaronmk
    match = re.match(r'^(.*)#(\d+)$', name)
539 2153 aaronmk
    if match:
540 2586 aaronmk
        name, version = match.groups()
541
        version = int(version)+1
542 2932 aaronmk
    return sql_gen.concat(name, '#'+str(version))
543 2153 aaronmk
544 2899 aaronmk
def lock_table(db, table, mode):
545
    table = sql_gen.as_Table(table)
546
    run_query(db, 'LOCK TABLE '+table.to_str(db)+' IN '+mode+' MODE')
547
548 2789 aaronmk
def run_query_into(db, query, into=None, add_indexes_=False, **kw_args):
549 2085 aaronmk
    '''Outputs a query to a temp table.
550
    For params, see run_query().
551
    '''
552 2789 aaronmk
    if into == None: return run_query(db, query, **kw_args)
553 2790 aaronmk
554
    assert isinstance(into, sql_gen.Table)
555
556 2992 aaronmk
    into.is_temp = True
557 3008 aaronmk
    # "temporary tables cannot specify a schema name", so remove schema
558
    into.schema = None
559 2992 aaronmk
560 2790 aaronmk
    kw_args['recover'] = True
561 2945 aaronmk
    kw_args.setdefault('log_ignore_excs', (DuplicateException,))
562 2790 aaronmk
563 2916 aaronmk
    temp = not db.debug_temp # tables are permanent in debug_temp mode
564 2790 aaronmk
565
    # Create table
566
    while True:
567
        create_query = 'CREATE'
568
        if temp: create_query += ' TEMP'
569
        create_query += ' TABLE '+into.to_str(db)+' AS\n'+query
570 2385 aaronmk
571 2790 aaronmk
        try:
572
            cur = run_query(db, create_query, **kw_args)
573
                # CREATE TABLE AS sets rowcount to # rows in query
574
            break
575 2945 aaronmk
        except DuplicateException, e:
576 2790 aaronmk
            into.name = next_version(into.name)
577
            # try again with next version of name
578
579
    if add_indexes_: add_indexes(db, into)
580 3075 aaronmk
581
    # According to the PostgreSQL doc, "The autovacuum daemon cannot access and
582
    # therefore cannot vacuum or analyze temporary tables. [...] if a temporary
583
    # table is going to be used in complex queries, it is wise to run ANALYZE on
584
    # the temporary table after it is populated."
585
    # (http://www.postgresql.org/docs/9.1/static/sql-createtable.html)
586
    # If into is not a temp table, ANALYZE is useful but not required.
587 3073 aaronmk
    analyze(db, into)
588 2790 aaronmk
589
    return cur
590 2085 aaronmk
591 2120 aaronmk
order_by_pkey = object() # tells mk_select() to order by the pkey
592
593 2199 aaronmk
distinct_on_all = object() # tells mk_select() to SELECT DISTINCT ON all columns
594
595 2233 aaronmk
def mk_select(db, tables, fields=None, conds=None, distinct_on=[], limit=None,
596 2293 aaronmk
    start=None, order_by=order_by_pkey, default_table=None):
597 1981 aaronmk
    '''
598 2121 aaronmk
    @param tables The single table to select from, or a list of tables to join
599 2280 aaronmk
        together, with tables after the first being sql_gen.Join objects
600 1981 aaronmk
    @param fields Use None to select all fields in the table
601 2377 aaronmk
    @param conds WHERE conditions: [(compare_left_side, compare_right_side),...]
602 2379 aaronmk
        * container can be any iterable type
603 2399 aaronmk
        * compare_left_side: sql_gen.Code|str (for col name)
604
        * compare_right_side: sql_gen.ValueCond|literal value
605 2199 aaronmk
    @param distinct_on The columns to SELECT DISTINCT ON, or distinct_on_all to
606
        use all columns
607 2786 aaronmk
    @return query
608 1981 aaronmk
    '''
609 2315 aaronmk
    # Parse tables param
610 2964 aaronmk
    tables = lists.mk_seq(tables)
611 2141 aaronmk
    tables = list(tables) # don't modify input! (list() copies input)
612 2315 aaronmk
    table0 = sql_gen.as_Table(tables.pop(0)) # first table is separate
613 2121 aaronmk
614 2315 aaronmk
    # Parse other params
615 2376 aaronmk
    if conds == None: conds = []
616 2650 aaronmk
    elif dicts.is_dict(conds): conds = conds.items()
617 2379 aaronmk
    conds = list(conds) # don't modify input! (list() copies input)
618 3129 aaronmk
    assert limit == None or isinstance(limit, (int, long))
619
    assert start == None or isinstance(start, (int, long))
620 2315 aaronmk
    if order_by is order_by_pkey:
621
        if distinct_on != []: order_by = None
622
        else: order_by = pkey(db, table0, recover=True)
623 865 aaronmk
624 2315 aaronmk
    query = 'SELECT'
625 2056 aaronmk
626 2315 aaronmk
    def parse_col(col): return sql_gen.as_Col(col, default_table).to_str(db)
627 2056 aaronmk
628 2200 aaronmk
    # DISTINCT ON columns
629 2233 aaronmk
    if distinct_on != []:
630 2467 aaronmk
        query += '\nDISTINCT'
631 2254 aaronmk
        if distinct_on is not distinct_on_all:
632 2200 aaronmk
            query += ' ON ('+(', '.join(map(parse_col, distinct_on)))+')'
633
634
    # Columns
635 3185 aaronmk
    if query.find('\n') >= 0: whitespace = '\n'
636
    else: whitespace = ' '
637
    if fields == None: query += whitespace+'*'
638 2765 aaronmk
    else:
639
        assert fields != []
640 3185 aaronmk
        if len(fields) > 1: whitespace = '\n'
641
        query += whitespace+('\n, '.join(map(parse_col, fields)))
642 2200 aaronmk
643
    # Main table
644 3185 aaronmk
    if query.find('\n') >= 0 or len(tables) > 0: whitespace = '\n'
645
    else: whitespace = ' '
646
    query += whitespace+'FROM '+table0.to_str(db)
647 865 aaronmk
648 2122 aaronmk
    # Add joins
649 2271 aaronmk
    left_table = table0
650 2263 aaronmk
    for join_ in tables:
651
        table = join_.table
652 2238 aaronmk
653 2343 aaronmk
        # Parse special values
654
        if join_.type_ is sql_gen.filter_out: # filter no match
655 2376 aaronmk
            conds.append((sql_gen.Col(table_not_null_col(db, table), table),
656 2853 aaronmk
                sql_gen.CompareCond(None, '~=')))
657 2343 aaronmk
658 2467 aaronmk
        query += '\n'+join_.to_str(db, left_table)
659 2122 aaronmk
660
        left_table = table
661
662 865 aaronmk
    missing = True
663 2376 aaronmk
    if conds != []:
664 2576 aaronmk
        if len(conds) == 1: whitespace = ' '
665
        else: whitespace = '\n'
666 2578 aaronmk
        query += '\n'+sql_gen.combine_conds([sql_gen.ColValueCond(l, r)
667
            .to_str(db) for l, r in conds], 'WHERE')
668 865 aaronmk
        missing = False
669 2227 aaronmk
    if order_by != None:
670 2467 aaronmk
        query += '\nORDER BY '+sql_gen.as_Col(order_by, table0).to_str(db)
671
    if limit != None: query += '\nLIMIT '+str(limit); missing = False
672 865 aaronmk
    if start != None:
673 2467 aaronmk
        if start != 0: query += '\nOFFSET '+str(start)
674 865 aaronmk
        missing = False
675
    if missing: warnings.warn(DbWarning(
676
        'SELECT statement missing a WHERE, LIMIT, or OFFSET clause: '+query))
677
678 2786 aaronmk
    return query
679 11 aaronmk
680 2054 aaronmk
def select(db, *args, **kw_args):
681
    '''For params, see mk_select() and run_query()'''
682
    recover = kw_args.pop('recover', None)
683
    cacheable = kw_args.pop('cacheable', True)
684 2442 aaronmk
    log_level = kw_args.pop('log_level', 2)
685 2054 aaronmk
686 2791 aaronmk
    return run_query(db, mk_select(db, *args, **kw_args), recover, cacheable,
687
        log_level=log_level)
688 2054 aaronmk
689 2788 aaronmk
def mk_insert_select(db, table, cols=None, select_query=None, returning=None,
690 3181 aaronmk
    embeddable=False, ignore=False, src=None):
691 1960 aaronmk
    '''
692
    @param returning str|None An inserted column (such as pkey) to return
693 2070 aaronmk
    @param embeddable Whether the query should be embeddable as a nested SELECT.
694 2073 aaronmk
        Warning: If you set this and cacheable=True when the query is run, the
695
        query will be fully cached, not just if it raises an exception.
696 3009 aaronmk
    @param ignore Whether to ignore duplicate keys.
697 3181 aaronmk
    @param src Will be included in the name of any created function, to help
698
        identify the data source in pg_stat_activity.
699 1960 aaronmk
    '''
700 2754 aaronmk
    table = sql_gen.remove_table_rename(sql_gen.as_Table(table))
701 2318 aaronmk
    if cols == []: cols = None # no cols (all defaults) = unknown col names
702 3010 aaronmk
    if cols != None: cols = [sql_gen.to_name_only_col(c, table) for c in cols]
703 2063 aaronmk
    if select_query == None: select_query = 'DEFAULT VALUES'
704 2327 aaronmk
    if returning != None: returning = sql_gen.as_Col(returning, table)
705 2063 aaronmk
706 2497 aaronmk
    first_line = 'INSERT INTO '+table.to_str(db)
707 2063 aaronmk
708 3009 aaronmk
    def mk_insert(select_query):
709
        query = first_line
710 3014 aaronmk
        if cols != None:
711
            query += '\n('+(', '.join((c.to_str(db) for c in cols)))+')'
712 3009 aaronmk
        query += '\n'+select_query
713
714
        if returning != None:
715
            returning_name_col = sql_gen.to_name_only_col(returning)
716
            query += '\nRETURNING '+returning_name_col.to_str(db)
717
718
        return query
719 2063 aaronmk
720 3017 aaronmk
    return_type = 'unknown'
721
    if returning != None: return_type = returning.to_str(db)+'%TYPE'
722
723 3009 aaronmk
    lang = 'sql'
724
    if ignore:
725 3017 aaronmk
        # Always return something to set the correct rowcount
726
        if returning == None: returning = sql_gen.NamedCol('NULL', None)
727
728 3009 aaronmk
        embeddable = True # must use function
729
        lang = 'plpgsql'
730 3010 aaronmk
731 3092 aaronmk
        if cols == None:
732
            row = [sql_gen.Col(sql_gen.all_cols, 'row')]
733
            row_vars = [sql_gen.Table('row')]
734
        else:
735
            row_vars = row = [sql_gen.Col(c.name, 'row') for c in cols]
736
737 3009 aaronmk
        query = '''\
738 3010 aaronmk
DECLARE
739 3014 aaronmk
    row '''+table.to_str(db)+'''%ROWTYPE;
740 3009 aaronmk
BEGIN
741 3019 aaronmk
    /* Need an EXCEPTION block for each individual row because "When an error is
742
    caught by an EXCEPTION clause, [...] all changes to persistent database
743
    state within the block are rolled back."
744
    This is unfortunate because "A block containing an EXCEPTION clause is
745
    significantly more expensive to enter and exit than a block without one."
746 3015 aaronmk
    (http://www.postgresql.org/docs/8.3/static/plpgsql-control-structures.html\
747
#PLPGSQL-ERROR-TRAPPING)
748
    */
749 3092 aaronmk
    FOR '''+(', '.join((v.to_str(db) for v in row_vars)))+''' IN
750 3034 aaronmk
'''+select_query+'''
751
    LOOP
752 3015 aaronmk
        BEGIN
753 3019 aaronmk
            RETURN QUERY
754 3014 aaronmk
'''+mk_insert(sql_gen.Values(row).to_str(db))+'''
755 3010 aaronmk
;
756 3015 aaronmk
        EXCEPTION
757 3019 aaronmk
            WHEN unique_violation THEN NULL; -- continue to next row
758 3015 aaronmk
        END;
759 3010 aaronmk
    END LOOP;
760
END;\
761 3009 aaronmk
'''
762
    else: query = mk_insert(select_query)
763
764 2070 aaronmk
    if embeddable:
765
        # Create function
766 2513 aaronmk
        function_name = sql_gen.clean_name(first_line)
767 3181 aaronmk
        if src != None: function_name = src+': '+function_name
768 2189 aaronmk
        while True:
769
            try:
770 2918 aaronmk
                function = db.TempFunction(function_name)
771 2194 aaronmk
772 2189 aaronmk
                function_query = '''\
773 2698 aaronmk
CREATE FUNCTION '''+function.to_str(db)+'''()
774 3017 aaronmk
RETURNS SETOF '''+return_type+'''
775 3009 aaronmk
LANGUAGE '''+lang+'''
776 2467 aaronmk
AS $$
777 3009 aaronmk
'''+query+'''
778 2467 aaronmk
$$;
779 2070 aaronmk
'''
780 2446 aaronmk
                run_query(db, function_query, recover=True, cacheable=True,
781 2945 aaronmk
                    log_ignore_excs=(DuplicateException,))
782 2189 aaronmk
                break # this version was successful
783 2945 aaronmk
            except DuplicateException, e:
784 2189 aaronmk
                function_name = next_version(function_name)
785
                # try again with next version of name
786 2070 aaronmk
787 2337 aaronmk
        # Return query that uses function
788 3009 aaronmk
        cols = None
789
        if returning != None: cols = [returning]
790 2698 aaronmk
        func_table = sql_gen.NamedTable('f', sql_gen.FunctionCall(function),
791 3009 aaronmk
            cols) # AS clause requires function alias
792 2787 aaronmk
        return mk_select(db, func_table, start=0, order_by=None)
793 2070 aaronmk
794 2787 aaronmk
    return query
795 2066 aaronmk
796 3074 aaronmk
def insert_select(db, table, *args, **kw_args):
797 2085 aaronmk
    '''For params, see mk_insert_select() and run_query_into()
798 2386 aaronmk
    @param into sql_gen.Table with suggested name of temp table to put RETURNING
799
        values in
800 2072 aaronmk
    '''
801 3141 aaronmk
    returning = kw_args.get('returning', None)
802
    ignore = kw_args.get('ignore', False)
803
804 2386 aaronmk
    into = kw_args.pop('into', None)
805
    if into != None: kw_args['embeddable'] = True
806 2066 aaronmk
    recover = kw_args.pop('recover', None)
807 3141 aaronmk
    if ignore: recover = True
808 2066 aaronmk
    cacheable = kw_args.pop('cacheable', True)
809 2673 aaronmk
    log_level = kw_args.pop('log_level', 2)
810 2066 aaronmk
811 3141 aaronmk
    rowcount_only = ignore and returning == None # keep NULL rows on server
812
    if rowcount_only: into = sql_gen.Table('rowcount')
813
814 3074 aaronmk
    cur = run_query_into(db, mk_insert_select(db, table, *args, **kw_args),
815
        into, recover=recover, cacheable=cacheable, log_level=log_level)
816 3141 aaronmk
    if rowcount_only: empty_temp(db, into)
817 3074 aaronmk
    autoanalyze(db, table)
818
    return cur
819 2063 aaronmk
820 2738 aaronmk
default = sql_gen.default # tells insert() to use the default value for a column
821 2066 aaronmk
822 2063 aaronmk
def insert(db, table, row, *args, **kw_args):
823 2085 aaronmk
    '''For params, see insert_select()'''
824 1960 aaronmk
    if lists.is_seq(row): cols = None
825
    else:
826
        cols = row.keys()
827
        row = row.values()
828 2738 aaronmk
    row = list(row) # ensure that "== []" works
829 1960 aaronmk
830 2738 aaronmk
    if row == []: query = None
831
    else: query = sql_gen.Values(row).to_str(db)
832 1961 aaronmk
833 2788 aaronmk
    return insert_select(db, table, cols, query, *args, **kw_args)
834 11 aaronmk
835 3152 aaronmk
def mk_update(db, table, changes=None, cond=None, in_place=False,
836 3153 aaronmk
    cacheable_=True):
837 2402 aaronmk
    '''
838
    @param changes [(col, new_value),...]
839
        * container can be any iterable type
840
        * col: sql_gen.Code|str (for col name)
841
        * new_value: sql_gen.Code|literal value
842
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
843 3056 aaronmk
    @param in_place If set, locks the table and updates rows in place.
844
        This avoids creating dead rows in PostgreSQL.
845
        * cond must be None
846 3153 aaronmk
    @param cacheable_ Whether column structure information used to generate the
847 3152 aaronmk
        query can be cached
848 2402 aaronmk
    @return str query
849
    '''
850 3057 aaronmk
    table = sql_gen.as_Table(table)
851
    changes = [(sql_gen.to_name_only_col(c, table), sql_gen.as_Value(v))
852
        for c, v in changes]
853
854 3056 aaronmk
    if in_place:
855
        assert cond == None
856 3058 aaronmk
857 3065 aaronmk
        query = 'ALTER TABLE '+table.to_str(db)+'\n'
858
        query += ',\n'.join(('ALTER COLUMN '+c.to_str(db)+' TYPE '
859 3153 aaronmk
            +db.col_info(sql_gen.with_default_table(c, table), cacheable_).type
860 3065 aaronmk
            +'\nUSING '+v.to_str(db) for c, v in changes))
861 3058 aaronmk
    else:
862
        query = 'UPDATE '+table.to_str(db)+'\nSET\n'
863
        query += ',\n'.join((c.to_str(db)+' = '+v.to_str(db)
864
            for c, v in changes))
865
        if cond != None: query += '\nWHERE\n'+cond.to_str(db)
866 3056 aaronmk
867 2402 aaronmk
    return query
868
869 3074 aaronmk
def update(db, table, *args, **kw_args):
870 2402 aaronmk
    '''For params, see mk_update() and run_query()'''
871
    recover = kw_args.pop('recover', None)
872 3043 aaronmk
    cacheable = kw_args.pop('cacheable', False)
873 3030 aaronmk
    log_level = kw_args.pop('log_level', 2)
874 2402 aaronmk
875 3074 aaronmk
    cur = run_query(db, mk_update(db, table, *args, **kw_args), recover,
876
        cacheable, log_level=log_level)
877
    autoanalyze(db, table)
878
    return cur
879 2402 aaronmk
880 135 aaronmk
def last_insert_id(db):
881 1849 aaronmk
    module = util.root_module(db.db)
882 135 aaronmk
    if module == 'psycopg2': return value(run_query(db, 'SELECT lastval()'))
883
    elif module == 'MySQLdb': return db.insert_id()
884
    else: return None
885 13 aaronmk
886 2394 aaronmk
def mk_flatten_mapping(db, into, cols, preserve=[], as_items=False):
887 2383 aaronmk
    '''Creates a mapping from original column names (which may have collisions)
888 2415 aaronmk
    to names that will be distinct among the columns' tables.
889 2383 aaronmk
    This is meant to be used for several tables that are being joined together.
890 2415 aaronmk
    @param cols The columns to combine. Duplicates will be removed.
891
    @param into The table for the new columns.
892 2394 aaronmk
    @param preserve [sql_gen.Col...] Columns not to rename. Note that these
893
        columns will be included in the mapping even if they are not in cols.
894
        The tables of the provided Col objects will be changed to into, so make
895
        copies of them if you want to keep the original tables.
896
    @param as_items Whether to return a list of dict items instead of a dict
897 2383 aaronmk
    @return dict(orig_col=new_col, ...)
898
        * orig_col: sql_gen.Col(orig_col_name, orig_table)
899 2392 aaronmk
        * new_col: sql_gen.Col(orig_col_name, into)
900
        * All mappings use the into table so its name can easily be
901 2383 aaronmk
          changed for all columns at once
902
    '''
903 2415 aaronmk
    cols = lists.uniqify(cols)
904
905 2394 aaronmk
    items = []
906 2389 aaronmk
    for col in preserve:
907 2390 aaronmk
        orig_col = copy.copy(col)
908 2392 aaronmk
        col.table = into
909 2394 aaronmk
        items.append((orig_col, col))
910
    preserve = set(preserve)
911
    for col in cols:
912 2716 aaronmk
        if col not in preserve:
913
            items.append((col, sql_gen.Col(str(col), into, col.srcs)))
914 2394 aaronmk
915
    if not as_items: items = dict(items)
916
    return items
917 2383 aaronmk
918 2393 aaronmk
def flatten(db, into, joins, cols, limit=None, start=None, **kw_args):
919 2391 aaronmk
    '''For params, see mk_flatten_mapping()
920
    @return See return value of mk_flatten_mapping()
921
    '''
922 2394 aaronmk
    items = mk_flatten_mapping(db, into, cols, as_items=True, **kw_args)
923
    cols = [sql_gen.NamedCol(new.name, old) for old, new in items]
924 2786 aaronmk
    run_query_into(db, mk_select(db, joins, cols, limit=limit, start=start),
925 3166 aaronmk
        into=into)
926 2394 aaronmk
    return dict(items)
927 2391 aaronmk
928 3079 aaronmk
##### Database structure introspection
929 2414 aaronmk
930 3079 aaronmk
#### Tables
931
932
def tables(db, schema_like='public', table_like='%', exact=False):
933
    if exact: compare = '='
934
    else: compare = 'LIKE'
935
936
    module = util.root_module(db.db)
937
    if module == 'psycopg2':
938
        conds = [('schemaname', sql_gen.CompareCond(schema_like, compare)),
939
            ('tablename', sql_gen.CompareCond(table_like, compare))]
940
        return values(select(db, 'pg_tables', ['tablename'], conds,
941
            order_by='tablename', log_level=4))
942
    elif module == 'MySQLdb':
943
        return values(run_query(db, 'SHOW TABLES LIKE '+db.esc_value(table_like)
944
            , cacheable=True, log_level=4))
945
    else: raise NotImplementedError("Can't list tables for "+module+' database')
946
947
def table_exists(db, table):
948
    table = sql_gen.as_Table(table)
949
    return list(tables(db, table.schema, table.name, exact=True)) != []
950
951 2426 aaronmk
def table_row_count(db, table, recover=None):
952 2786 aaronmk
    return value(run_query(db, mk_select(db, table, [sql_gen.row_count],
953 2443 aaronmk
        order_by=None, start=0), recover=recover, log_level=3))
954 2426 aaronmk
955 2414 aaronmk
def table_cols(db, table, recover=None):
956
    return list(col_names(select(db, table, limit=0, order_by=None,
957 2443 aaronmk
        recover=recover, log_level=4)))
958 2414 aaronmk
959 2291 aaronmk
def pkey(db, table, recover=None):
960 832 aaronmk
    '''Assumed to be first column in table'''
961 2339 aaronmk
    return table_cols(db, table, recover)[0]
962 832 aaronmk
963 2559 aaronmk
not_null_col = 'not_null_col'
964 2340 aaronmk
965
def table_not_null_col(db, table, recover=None):
966
    '''Name assumed to be the value of not_null_col. If not found, uses pkey.'''
967
    if not_null_col in table_cols(db, table, recover): return not_null_col
968
    else: return pkey(db, table, recover)
969
970 853 aaronmk
def index_cols(db, table, index):
971
    '''Can also use this for UNIQUE constraints, because a UNIQUE index is
972
    automatically created. When you don't know whether something is a UNIQUE
973
    constraint or a UNIQUE index, use this function.'''
974 1909 aaronmk
    module = util.root_module(db.db)
975
    if module == 'psycopg2':
976
        return list(values(run_query(db, '''\
977 853 aaronmk
SELECT attname
978 866 aaronmk
FROM
979
(
980
        SELECT attnum, attname
981
        FROM pg_index
982
        JOIN pg_class index ON index.oid = indexrelid
983
        JOIN pg_class table_ ON table_.oid = indrelid
984
        JOIN pg_attribute ON attrelid = indrelid AND attnum = ANY (indkey)
985
        WHERE
986 2782 aaronmk
            table_.relname = '''+db.esc_value(table)+'''
987
            AND index.relname = '''+db.esc_value(index)+'''
988 866 aaronmk
    UNION
989
        SELECT attnum, attname
990
        FROM
991
        (
992
            SELECT
993
                indrelid
994
                , (regexp_matches(indexprs, E':varattno (\\\\d+)', 'g'))[1]::int
995
                    AS indkey
996
            FROM pg_index
997
            JOIN pg_class index ON index.oid = indexrelid
998
            JOIN pg_class table_ ON table_.oid = indrelid
999
            WHERE
1000 2782 aaronmk
                table_.relname = '''+db.esc_value(table)+'''
1001
                AND index.relname = '''+db.esc_value(index)+'''
1002 866 aaronmk
        ) s
1003
        JOIN pg_attribute ON attrelid = indrelid AND attnum = indkey
1004
) s
1005 853 aaronmk
ORDER BY attnum
1006 2782 aaronmk
'''
1007
            , cacheable=True, log_level=4)))
1008 1909 aaronmk
    else: raise NotImplementedError("Can't list index columns for "+module+
1009
        ' database')
1010 853 aaronmk
1011 464 aaronmk
def constraint_cols(db, table, constraint):
1012 1849 aaronmk
    module = util.root_module(db.db)
1013 464 aaronmk
    if module == 'psycopg2':
1014
        return list(values(run_query(db, '''\
1015
SELECT attname
1016
FROM pg_constraint
1017
JOIN pg_class ON pg_class.oid = conrelid
1018
JOIN pg_attribute ON attrelid = conrelid AND attnum = ANY (conkey)
1019
WHERE
1020 2783 aaronmk
    relname = '''+db.esc_value(table)+'''
1021
    AND conname = '''+db.esc_value(constraint)+'''
1022 464 aaronmk
ORDER BY attnum
1023 2783 aaronmk
'''
1024
            )))
1025 464 aaronmk
    else: raise NotImplementedError("Can't list constraint columns for "+module+
1026
        ' database')
1027
1028 3079 aaronmk
#### Functions
1029
1030
def function_exists(db, function):
1031
    function = sql_gen.as_Function(function)
1032
1033
    info_table = sql_gen.Table('routines', 'information_schema')
1034
    conds = [('routine_name', function.name)]
1035
    schema = function.schema
1036
    if schema != None: conds.append(('routine_schema', schema))
1037
    # Exclude trigger functions, since they cannot be called directly
1038
    conds.append(('data_type', sql_gen.CompareCond('trigger', '!=')))
1039
1040
    return list(values(select(db, info_table, ['routine_name'], conds,
1041
        order_by='routine_schema', limit=1, log_level=4))) != []
1042
        # TODO: order_by search_path schema order
1043
1044
##### Structural changes
1045
1046
#### Columns
1047
1048
def add_col(db, table, col, comment=None, **kw_args):
1049
    '''
1050
    @param col TypedCol Name may be versioned, so be sure to propagate any
1051
        renaming back to any source column for the TypedCol.
1052
    @param comment None|str SQL comment used to distinguish columns of the same
1053
        name from each other when they contain different data, to allow the
1054
        ADD COLUMN query to be cached. If not set, query will not be cached.
1055
    '''
1056
    assert isinstance(col, sql_gen.TypedCol)
1057
1058
    while True:
1059
        str_ = 'ALTER TABLE '+table.to_str(db)+' ADD COLUMN '+col.to_str(db)
1060
        if comment != None: str_ += ' '+sql_gen.esc_comment(comment)
1061
1062
        try:
1063
            run_query(db, str_, recover=True, cacheable=True, **kw_args)
1064
            break
1065
        except DuplicateException:
1066
            col.name = next_version(col.name)
1067
            # try again with next version of name
1068
1069
def add_not_null(db, col):
1070
    table = col.table
1071
    col = sql_gen.to_name_only_col(col)
1072
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ALTER COLUMN '
1073
        +col.to_str(db)+' SET NOT NULL', cacheable=True, log_level=3)
1074
1075 2096 aaronmk
row_num_col = '_row_num'
1076
1077 3079 aaronmk
row_num_typed_col = sql_gen.TypedCol(row_num_col, 'serial', nullable=False,
1078
    constraints='PRIMARY KEY')
1079
1080
def add_row_num(db, table):
1081
    '''Adds a row number column to a table. Its name is in row_num_col. It will
1082
    be the primary key.'''
1083
    add_col(db, table, row_num_typed_col, log_level=3)
1084
1085
#### Indexes
1086
1087
def add_pkey(db, table, cols=None, recover=None):
1088
    '''Adds a primary key.
1089
    @param cols [sql_gen.Col,...] The columns in the primary key.
1090
        Defaults to the first column in the table.
1091
    @pre The table must not already have a primary key.
1092
    '''
1093
    table = sql_gen.as_Table(table)
1094
    if cols == None: cols = [pkey(db, table, recover)]
1095
    col_strs = [sql_gen.to_name_only_col(v).to_str(db) for v in cols]
1096
1097
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ADD PRIMARY KEY ('
1098
        +(', '.join(col_strs))+')', recover=True, cacheable=True, log_level=3,
1099
        log_ignore_excs=(DuplicateException,))
1100
1101 2998 aaronmk
def add_index(db, exprs, table=None, unique=False, ensure_not_null_=True):
1102 2688 aaronmk
    '''Adds an index on column(s) or expression(s) if it doesn't already exist.
1103 2538 aaronmk
    Currently, only function calls are supported as expressions.
1104 2998 aaronmk
    @param ensure_not_null_ If set, translates NULL values to sentinel values.
1105 2847 aaronmk
        This allows indexes to be used for comparisons where NULLs are equal.
1106 2538 aaronmk
    '''
1107 2964 aaronmk
    exprs = lists.mk_seq(exprs)
1108 2538 aaronmk
1109 2688 aaronmk
    # Parse exprs
1110
    old_exprs = exprs[:]
1111
    exprs = []
1112
    cols = []
1113
    for i, expr in enumerate(old_exprs):
1114 2823 aaronmk
        expr = sql_gen.as_Col(expr, table)
1115 2688 aaronmk
1116 2823 aaronmk
        # Handle nullable columns
1117 2998 aaronmk
        if ensure_not_null_:
1118 3164 aaronmk
            try: expr = sql_gen.ensure_not_null(db, expr)
1119 2860 aaronmk
            except KeyError: pass # unknown type, so just create plain index
1120 2823 aaronmk
1121 2688 aaronmk
        # Extract col
1122 3002 aaronmk
        expr = copy.deepcopy(expr) # don't modify input!
1123 2688 aaronmk
        if isinstance(expr, sql_gen.FunctionCall):
1124
            col = expr.args[0]
1125
            expr = sql_gen.Expr(expr)
1126
        else: col = expr
1127 2823 aaronmk
        assert isinstance(col, sql_gen.Col)
1128 2688 aaronmk
1129
        # Extract table
1130
        if table == None:
1131
            assert sql_gen.is_table_col(col)
1132
            table = col.table
1133
1134
        col.table = None
1135
1136
        exprs.append(expr)
1137
        cols.append(col)
1138 2408 aaronmk
1139 2688 aaronmk
    table = sql_gen.as_Table(table)
1140
1141 3005 aaronmk
    # Add index
1142 3148 aaronmk
    str_ = 'CREATE'
1143
    if unique: str_ += ' UNIQUE'
1144
    str_ += ' INDEX ON '+table.to_str(db)+' ('+(
1145
        ', '.join((v.to_str(db) for v in exprs)))+')'
1146
    run_query(db, str_, recover=True, cacheable=True, log_level=3)
1147 2408 aaronmk
1148 3083 aaronmk
already_indexed = object() # tells add_indexes() the pkey has already been added
1149
1150
def add_indexes(db, table, has_pkey=True):
1151
    '''Adds an index on all columns in a table.
1152
    @param has_pkey bool|already_indexed Whether a pkey instead of a regular
1153
        index should be added on the first column.
1154
        * If already_indexed, the pkey is assumed to have already been added
1155
    '''
1156
    cols = table_cols(db, table)
1157
    if has_pkey:
1158
        if has_pkey is not already_indexed: add_pkey(db, table)
1159
        cols = cols[1:]
1160
    for col in cols: add_index(db, col, table)
1161
1162 3079 aaronmk
#### Tables
1163 2772 aaronmk
1164 3079 aaronmk
### Maintenance
1165 2772 aaronmk
1166 3079 aaronmk
def analyze(db, table):
1167
    table = sql_gen.as_Table(table)
1168
    run_query(db, 'ANALYZE '+table.to_str(db), log_level=3)
1169 2934 aaronmk
1170 3079 aaronmk
def autoanalyze(db, table):
1171
    if db.autoanalyze: analyze(db, table)
1172 2935 aaronmk
1173 3079 aaronmk
def vacuum(db, table):
1174
    table = sql_gen.as_Table(table)
1175
    db.with_autocommit(lambda: run_query(db, 'VACUUM ANALYZE '+table.to_str(db),
1176
        log_level=3))
1177 2086 aaronmk
1178 3079 aaronmk
### Lifecycle
1179
1180 2889 aaronmk
def drop_table(db, table):
1181
    table = sql_gen.as_Table(table)
1182
    return run_query(db, 'DROP TABLE IF EXISTS '+table.to_str(db)+' CASCADE')
1183
1184 3082 aaronmk
def create_table(db, table, cols=[], has_pkey=True, col_indexes=True,
1185
    like=None):
1186 2675 aaronmk
    '''Creates a table.
1187 2681 aaronmk
    @param cols [sql_gen.TypedCol,...] The column names and types
1188
    @param has_pkey If set, the first column becomes the primary key.
1189 2760 aaronmk
    @param col_indexes bool|[ref]
1190
        * If True, indexes will be added on all non-pkey columns.
1191
        * If a list reference, [0] will be set to a function to do this.
1192
          This can be used to delay index creation until the table is populated.
1193 2675 aaronmk
    '''
1194
    table = sql_gen.as_Table(table)
1195
1196 3082 aaronmk
    if like != None:
1197
        cols = [sql_gen.CustomCode('LIKE '+like.to_str(db)+' INCLUDING ALL')
1198
            ]+cols
1199 2681 aaronmk
    if has_pkey:
1200
        cols[0] = pkey = copy.copy(cols[0]) # don't modify input!
1201 2872 aaronmk
        pkey.constraints = 'PRIMARY KEY'
1202 2681 aaronmk
1203 3085 aaronmk
    temp = table.is_temp and not db.debug_temp
1204
        # temp tables permanent in debug_temp mode
1205 2760 aaronmk
1206 3085 aaronmk
    # Create table
1207
    while True:
1208
        str_ = 'CREATE'
1209
        if temp: str_ += ' TEMP'
1210
        str_ += ' TABLE '+table.to_str(db)+' (\n'
1211
        str_ += '\n, '.join(c.to_str(db) for c in cols)
1212 3126 aaronmk
        str_ += '\n);'
1213 3085 aaronmk
1214
        try:
1215 3127 aaronmk
            run_query(db, str_, recover=True, cacheable=True, log_level=2,
1216 3085 aaronmk
                log_ignore_excs=(DuplicateException,))
1217
            break
1218
        except DuplicateException:
1219
            table.name = next_version(table.name)
1220
            # try again with next version of name
1221
1222 2760 aaronmk
    # Add indexes
1223 2773 aaronmk
    if has_pkey: has_pkey = already_indexed
1224
    def add_indexes_(): add_indexes(db, table, has_pkey)
1225
    if isinstance(col_indexes, list): col_indexes[0] = add_indexes_ # defer
1226
    elif col_indexes: add_indexes_() # add now
1227 2675 aaronmk
1228 3084 aaronmk
def copy_table_struct(db, src, dest):
1229
    '''Creates a structure-only copy of a table. (Does not copy data.)'''
1230 3085 aaronmk
    create_table(db, dest, has_pkey=False, col_indexes=False, like=src)
1231 3084 aaronmk
1232 3079 aaronmk
### Data
1233 2684 aaronmk
1234 2970 aaronmk
def truncate(db, table, schema='public', **kw_args):
1235
    '''For params, see run_query()'''
1236 2777 aaronmk
    table = sql_gen.as_Table(table, schema)
1237 2970 aaronmk
    return run_query(db, 'TRUNCATE '+table.to_str(db)+' CASCADE', **kw_args)
1238 2732 aaronmk
1239 2965 aaronmk
def empty_temp(db, tables):
1240
    tables = lists.mk_seq(tables)
1241 2971 aaronmk
    for table in tables: truncate(db, table, log_level=3)
1242 2965 aaronmk
1243 1968 aaronmk
def empty_db(db, schema='public', **kw_args):
1244
    '''For kw_args, see tables()'''
1245
    for table in tables(db, schema, **kw_args): truncate(db, table, schema)
1246 3094 aaronmk
1247
def distinct_table(db, table, distinct_on):
1248
    '''Creates a copy of a temp table which is distinct on the given columns.
1249 3099 aaronmk
    The old and new tables will both get an index on these columns, to
1250
    facilitate merge joins.
1251 3097 aaronmk
    @param distinct_on If empty, creates a table with one row. This is useful if
1252
        your distinct_on columns are all literal values.
1253 3099 aaronmk
    @return The new table.
1254 3094 aaronmk
    '''
1255 3099 aaronmk
    new_table = sql_gen.suffixed_table(table, '_distinct')
1256 3094 aaronmk
1257 3099 aaronmk
    copy_table_struct(db, table, new_table)
1258 3097 aaronmk
1259
    limit = None
1260
    if distinct_on == []: limit = 1 # one sample row
1261 3099 aaronmk
    else:
1262
        add_index(db, distinct_on, new_table, unique=True)
1263
        add_index(db, distinct_on, table) # for join optimization
1264 3097 aaronmk
1265 3099 aaronmk
    insert_select(db, new_table, None, mk_select(db, table, start=0,
1266 3097 aaronmk
        limit=limit), ignore=True)
1267 3099 aaronmk
    analyze(db, new_table)
1268 3094 aaronmk
1269 3099 aaronmk
    return new_table