Project

General

Profile

1 11 aaronmk
# Database access
2
3 1869 aaronmk
import copy
4 2127 aaronmk
import operator
5 11 aaronmk
import re
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 1889 aaronmk
from Proxy import Proxy
13 1872 aaronmk
import rand
14 2217 aaronmk
import sql_gen
15 862 aaronmk
import strings
16 131 aaronmk
import util
17 11 aaronmk
18 832 aaronmk
##### Exceptions
19
20 2804 aaronmk
def get_cur_query(cur, input_query=None):
21 2168 aaronmk
    raw_query = None
22
    if hasattr(cur, 'query'): raw_query = cur.query
23
    elif hasattr(cur, '_last_executed'): raw_query = cur._last_executed
24 2170 aaronmk
25
    if raw_query != None: return raw_query
26 2804 aaronmk
    else: return '[input] '+strings.ustr(input_query)
27 14 aaronmk
28 2170 aaronmk
def _add_cursor_info(e, *args, **kw_args):
29
    '''For params, see get_cur_query()'''
30 2771 aaronmk
    exc.add_msg(e, 'query: '+strings.ustr(get_cur_query(*args, **kw_args)))
31 135 aaronmk
32 300 aaronmk
class DbException(exc.ExceptionWithCause):
33 14 aaronmk
    def __init__(self, msg, cause=None, cur=None):
34 2145 aaronmk
        exc.ExceptionWithCause.__init__(self, msg, cause, cause_newline=True)
35 14 aaronmk
        if cur != None: _add_cursor_info(self, cur)
36
37 2143 aaronmk
class ExceptionWithName(DbException):
38
    def __init__(self, name, cause=None):
39 2484 aaronmk
        DbException.__init__(self, 'for name: '+strings.as_tt(str(name)), cause)
40 2143 aaronmk
        self.name = name
41 360 aaronmk
42 2240 aaronmk
class ExceptionWithNameValue(DbException):
43
    def __init__(self, name, value, cause=None):
44 2484 aaronmk
        DbException.__init__(self, 'for name: '+strings.as_tt(str(name))
45
            +'; value: '+strings.as_tt(repr(value)), cause)
46 2240 aaronmk
        self.name = name
47
        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 2240 aaronmk
class FunctionValueException(ExceptionWithNameValue): 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 2915 aaronmk
        log_debug=log_debug_none, debug_temp=False):
158
        '''
159
        @param debug_temp Whether temporary objects should instead be permanent.
160
            This assists in debugging the internal objects used by the program.
161
        '''
162 1869 aaronmk
        self.db_config = db_config
163 2190 aaronmk
        self.autocommit = autocommit
164
        self.caching = caching
165 1901 aaronmk
        self.log_debug = log_debug
166 2193 aaronmk
        self.debug = log_debug != log_debug_none
167 2915 aaronmk
        self.debug_temp = debug_temp
168 1869 aaronmk
169
        self.__db = None
170 1889 aaronmk
        self.query_results = {}
171 2139 aaronmk
        self._savepoint = 0
172 2671 aaronmk
        self._notices_seen = set()
173 1869 aaronmk
174
    def __getattr__(self, name):
175
        if name == '__dict__': raise Exception('getting __dict__')
176
        if name == 'db': return self._db()
177
        else: raise AttributeError()
178
179
    def __getstate__(self):
180
        state = copy.copy(self.__dict__) # shallow copy
181 1915 aaronmk
        state['log_debug'] = None # don't pickle the debug callback
182 1869 aaronmk
        state['_DbConn__db'] = None # don't pickle the connection
183
        return state
184
185 2165 aaronmk
    def connected(self): return self.__db != None
186
187 1869 aaronmk
    def _db(self):
188
        if self.__db == None:
189
            # Process db_config
190
            db_config = self.db_config.copy() # don't modify input!
191 2097 aaronmk
            schemas = db_config.pop('schemas', None)
192 1869 aaronmk
            module_name, mappings = db_engines[db_config.pop('engine')]
193
            module = __import__(module_name)
194
            _add_module(module)
195
            for orig, new in mappings.iteritems():
196
                try: util.rename_key(db_config, orig, new)
197
                except KeyError: pass
198
199
            # Connect
200
            self.__db = module.connect(**db_config)
201
202
            # Configure connection
203 2906 aaronmk
            if hasattr(self.db, 'set_isolation_level'):
204
                import psycopg2.extensions
205
                self.db.set_isolation_level(
206
                    psycopg2.extensions.ISOLATION_LEVEL_READ_COMMITTED)
207 2101 aaronmk
            if schemas != None:
208 2893 aaronmk
                search_path = [self.esc_name(s) for s in schemas.split(',')]
209
                search_path.append(value(run_query(self, 'SHOW search_path',
210
                    log_level=4)))
211
                run_query(self, 'SET search_path TO '+(','.join(search_path)),
212
                    log_level=3)
213 1869 aaronmk
214
        return self.__db
215 1889 aaronmk
216 1891 aaronmk
    class DbCursor(Proxy):
217 1927 aaronmk
        def __init__(self, outer):
218 1891 aaronmk
            Proxy.__init__(self, outer.db.cursor())
219 2191 aaronmk
            self.outer = outer
220 1927 aaronmk
            self.query_results = outer.query_results
221 1894 aaronmk
            self.query_lookup = None
222 1891 aaronmk
            self.result = []
223 1889 aaronmk
224 2802 aaronmk
        def execute(self, query):
225 2764 aaronmk
            self._is_insert = query.startswith('INSERT')
226 2797 aaronmk
            self.query_lookup = query
227 2148 aaronmk
            try:
228 2191 aaronmk
                try:
229 2802 aaronmk
                    cur = self.inner.execute(query)
230 2191 aaronmk
                    self.outer.do_autocommit()
231 2802 aaronmk
                finally: self.query = get_cur_query(self.inner, query)
232 1904 aaronmk
            except Exception, e:
233 2802 aaronmk
                _add_cursor_info(e, self, query)
234 1904 aaronmk
                self.result = e # cache the exception as the result
235
                self._cache_result()
236
                raise
237 2800 aaronmk
            if self.rowcount == 0 and query.startswith('SELECT'): # empty SELECT
238
                consume_rows(self) # fetch all rows so result will be cached
239 2762 aaronmk
            return cur
240 1894 aaronmk
241 1891 aaronmk
        def fetchone(self):
242
            row = self.inner.fetchone()
243 1899 aaronmk
            if row != None: self.result.append(row)
244
            # otherwise, fetched all rows
245 1904 aaronmk
            else: self._cache_result()
246
            return row
247
248
        def _cache_result(self):
249 2948 aaronmk
            # For inserts that return a result set, don't cache result set since
250
            # inserts are not idempotent. Other non-SELECT queries don't have
251
            # their result set read, so only exceptions will be cached (an
252
            # invalid query will always be invalid).
253 1930 aaronmk
            if self.query_results != None and (not self._is_insert
254 1906 aaronmk
                or isinstance(self.result, Exception)):
255
256 1894 aaronmk
                assert self.query_lookup != None
257 1916 aaronmk
                self.query_results[self.query_lookup] = self.CacheCursor(
258
                    util.dict_subset(dicts.AttrsDictView(self),
259
                    ['query', 'result', 'rowcount', 'description']))
260 1906 aaronmk
261 1916 aaronmk
        class CacheCursor:
262
            def __init__(self, cached_result): self.__dict__ = cached_result
263
264 1927 aaronmk
            def execute(self, *args, **kw_args):
265 1916 aaronmk
                if isinstance(self.result, Exception): raise self.result
266
                # otherwise, result is a rows list
267
                self.iter = iter(self.result)
268
269
            def fetchone(self):
270
                try: return self.iter.next()
271
                except StopIteration: return None
272 1891 aaronmk
273 2212 aaronmk
    def esc_value(self, value):
274 2663 aaronmk
        try: str_ = self.mogrify('%s', [value])
275
        except NotImplementedError, e:
276
            module = util.root_module(self.db)
277
            if module == 'MySQLdb':
278
                import _mysql
279
                str_ = _mysql.escape_string(value)
280
            else: raise e
281 2374 aaronmk
        return strings.to_unicode(str_)
282 2212 aaronmk
283 2347 aaronmk
    def esc_name(self, name): return esc_name(self, name) # calls global func
284
285 2814 aaronmk
    def std_code(self, str_):
286
        '''Standardizes SQL code.
287
        * Ensures that string literals are prefixed by `E`
288
        '''
289
        if str_.startswith("'"): str_ = 'E'+str_
290
        return str_
291
292 2665 aaronmk
    def can_mogrify(self):
293 2663 aaronmk
        module = util.root_module(self.db)
294 2665 aaronmk
        return module == 'psycopg2'
295 2663 aaronmk
296 2665 aaronmk
    def mogrify(self, query, params=None):
297
        if self.can_mogrify(): return self.db.cursor().mogrify(query, params)
298
        else: raise NotImplementedError("Can't mogrify query")
299
300 2671 aaronmk
    def print_notices(self):
301 2725 aaronmk
        if hasattr(self.db, 'notices'):
302
            for msg in self.db.notices:
303
                if msg not in self._notices_seen:
304
                    self._notices_seen.add(msg)
305
                    self.log_debug(msg, level=2)
306 2671 aaronmk
307 2793 aaronmk
    def run_query(self, query, cacheable=False, log_level=2,
308 2464 aaronmk
        debug_msg_ref=None):
309 2445 aaronmk
        '''
310 2464 aaronmk
        @param log_ignore_excs The log_level will be increased by 2 if the query
311
            throws one of these exceptions.
312 2664 aaronmk
        @param debug_msg_ref If specified, the log message will be returned in
313
            this instead of being output. This allows you to filter log messages
314
            depending on the result of the query.
315 2445 aaronmk
        '''
316 2167 aaronmk
        assert query != None
317
318 2047 aaronmk
        if not self.caching: cacheable = False
319 1903 aaronmk
        used_cache = False
320 2664 aaronmk
321
        def log_msg(query):
322
            if used_cache: cache_status = 'cache hit'
323
            elif cacheable: cache_status = 'cache miss'
324
            else: cache_status = 'non-cacheable'
325
            return 'DB query: '+cache_status+':\n'+strings.as_code(query, 'SQL')
326
327 1903 aaronmk
        try:
328 1927 aaronmk
            # Get cursor
329
            if cacheable:
330
                try:
331 2797 aaronmk
                    cur = self.query_results[query]
332 1927 aaronmk
                    used_cache = True
333
                except KeyError: cur = self.DbCursor(self)
334
            else: cur = self.db.cursor()
335
336 2664 aaronmk
            # Log query
337
            if self.debug and debug_msg_ref == None: # log before running
338
                self.log_debug(log_msg(query), log_level)
339
340 1927 aaronmk
            # Run query
341 2793 aaronmk
            cur.execute(query)
342 1903 aaronmk
        finally:
343 2671 aaronmk
            self.print_notices()
344 2664 aaronmk
            if self.debug and debug_msg_ref != None: # return after running
345 2793 aaronmk
                debug_msg_ref[0] = log_msg(str(get_cur_query(cur, query)))
346 1903 aaronmk
347
        return cur
348 1914 aaronmk
349 2797 aaronmk
    def is_cached(self, query): return query in self.query_results
350 2139 aaronmk
351 2907 aaronmk
    def with_autocommit(self, func):
352 2801 aaronmk
        import psycopg2.extensions
353
354
        prev_isolation_level = self.db.isolation_level
355 2907 aaronmk
        self.db.set_isolation_level(
356
            psycopg2.extensions.ISOLATION_LEVEL_AUTOCOMMIT)
357 2683 aaronmk
        try: return func()
358 2801 aaronmk
        finally: self.db.set_isolation_level(prev_isolation_level)
359 2683 aaronmk
360 2139 aaronmk
    def with_savepoint(self, func):
361 2171 aaronmk
        savepoint = 'level_'+str(self._savepoint)
362 2443 aaronmk
        self.run_query('SAVEPOINT '+savepoint, log_level=4)
363 2139 aaronmk
        self._savepoint += 1
364 2930 aaronmk
        try: return func()
365 2139 aaronmk
        except:
366 2443 aaronmk
            self.run_query('ROLLBACK TO SAVEPOINT '+savepoint, log_level=4)
367 2139 aaronmk
            raise
368 2930 aaronmk
        finally:
369
            # Always release savepoint, because after ROLLBACK TO SAVEPOINT,
370
            # "The savepoint remains valid and can be rolled back to again"
371
            # (http://www.postgresql.org/docs/8.3/static/sql-rollback-to.html).
372 2443 aaronmk
            self.run_query('RELEASE SAVEPOINT '+savepoint, log_level=4)
373 2930 aaronmk
374
            self._savepoint -= 1
375
            assert self._savepoint >= 0
376
377
            self.do_autocommit() # OK to do this after ROLLBACK TO SAVEPOINT
378 2191 aaronmk
379
    def do_autocommit(self):
380
        '''Autocommits if outside savepoint'''
381
        assert self._savepoint >= 0
382
        if self.autocommit and self._savepoint == 0:
383 2924 aaronmk
            self.log_debug('Autocommitting', level=4)
384 2191 aaronmk
            self.db.commit()
385 2643 aaronmk
386 2819 aaronmk
    def col_info(self, col):
387 2643 aaronmk
        table = sql_gen.Table('columns', 'information_schema')
388 2821 aaronmk
        cols = ['data_type', 'column_default',
389
            cast(self, 'boolean', 'is_nullable')]
390 2643 aaronmk
391
        conds = [('table_name', col.table.name), ('column_name', col.name)]
392
        schema = col.table.schema
393
        if schema != None: conds.append(('table_schema', schema))
394
395 2819 aaronmk
        type_, default, nullable = row(select(self, table, cols, conds,
396 2874 aaronmk
            order_by='table_schema', limit=1, log_level=4))
397 2643 aaronmk
            # TODO: order_by search_path schema order
398 2819 aaronmk
        default = sql_gen.as_Code(default, self)
399
400
        return sql_gen.TypedCol(col.name, type_, default, nullable)
401 2917 aaronmk
402
    def TempFunction(self, name):
403
        if self.debug_temp: schema = None
404
        else: schema = 'pg_temp'
405
        return sql_gen.Function(name, schema)
406 1849 aaronmk
407 1869 aaronmk
connect = DbConn
408
409 832 aaronmk
##### Recoverable querying
410 15 aaronmk
411 2139 aaronmk
def with_savepoint(db, func): return db.with_savepoint(func)
412 11 aaronmk
413 2791 aaronmk
def run_query(db, query, recover=None, cacheable=False, log_level=2,
414
    log_ignore_excs=None, **kw_args):
415 2794 aaronmk
    '''For params, see DbConn.run_query()'''
416 830 aaronmk
    if recover == None: recover = False
417 2464 aaronmk
    if log_ignore_excs == None: log_ignore_excs = ()
418
    log_ignore_excs = tuple(log_ignore_excs)
419 830 aaronmk
420 2666 aaronmk
    debug_msg_ref = None # usually, db.run_query() logs query before running it
421
    # But if filtering with log_ignore_excs, wait until after exception parsing
422 2984 aaronmk
    if log_ignore_excs != () or not db.can_mogrify(): debug_msg_ref = [None]
423 2666 aaronmk
424 2148 aaronmk
    try:
425 2464 aaronmk
        try:
426 2794 aaronmk
            def run(): return db.run_query(query, cacheable, log_level,
427 2793 aaronmk
                debug_msg_ref, **kw_args)
428 2796 aaronmk
            if recover and not db.is_cached(query):
429 2464 aaronmk
                return with_savepoint(db, run)
430
            else: return run() # don't need savepoint if cached
431
        except Exception, e:
432
            if not recover: raise # need savepoint to run index_cols()
433
            msg = exc.str_(e)
434
435
            match = re.search(r'duplicate key value violates unique constraint '
436 2493 aaronmk
                r'"((_?[^\W_]+)_.+?)"', msg)
437 2464 aaronmk
            if match:
438
                constraint, table = match.groups()
439
                try: cols = index_cols(db, table, constraint)
440
                except NotImplementedError: raise e
441
                else: raise DuplicateKeyException(constraint, cols, e)
442
443 2493 aaronmk
            match = re.search(r'null value in column "(.+?)" violates not-null'
444 2464 aaronmk
                r' constraint', msg)
445
            if match: raise NullValueException('NOT NULL', [match.group(1)], e)
446
447
            match = re.search(r'\b(?:invalid input (?:syntax|value)\b.*?'
448
                r'|date/time field value out of range): "(.+?)"\n'
449 2535 aaronmk
                r'(?:(?s).*?)\bfunction "(.+?)"', msg)
450 2464 aaronmk
            if match:
451
                value, name = match.groups()
452
                raise FunctionValueException(name, strings.to_unicode(value), e)
453
454 2526 aaronmk
            match = re.search(r'column "(.+?)" is of type (.+?) but expression '
455 2523 aaronmk
                r'is of type', msg)
456
            if match:
457
                col, type_ = match.groups()
458
                raise MissingCastException(type_, col, e)
459
460 2946 aaronmk
            match = re.search(r'\b(\S+) "(.+?)".*? already exists', msg)
461 2945 aaronmk
            if match:
462
                type_, name = match.groups()
463
                raise DuplicateException(type_, name, e)
464 2464 aaronmk
465
            raise # no specific exception raised
466
    except log_ignore_excs:
467
        log_level += 2
468
        raise
469
    finally:
470 2666 aaronmk
        if debug_msg_ref != None and debug_msg_ref[0] != None:
471
            db.log_debug(debug_msg_ref[0], log_level)
472 830 aaronmk
473 832 aaronmk
##### Basic queries
474
475 2153 aaronmk
def next_version(name):
476 2163 aaronmk
    version = 1 # first existing name was version 0
477 2586 aaronmk
    match = re.match(r'^(.*)#(\d+)$', name)
478 2153 aaronmk
    if match:
479 2586 aaronmk
        name, version = match.groups()
480
        version = int(version)+1
481 2932 aaronmk
    return sql_gen.concat(name, '#'+str(version))
482 2153 aaronmk
483 2899 aaronmk
def lock_table(db, table, mode):
484
    table = sql_gen.as_Table(table)
485
    run_query(db, 'LOCK TABLE '+table.to_str(db)+' IN '+mode+' MODE')
486
487 2789 aaronmk
def run_query_into(db, query, into=None, add_indexes_=False, **kw_args):
488 2085 aaronmk
    '''Outputs a query to a temp table.
489
    For params, see run_query().
490
    '''
491 2789 aaronmk
    if into == None: return run_query(db, query, **kw_args)
492 2790 aaronmk
493
    assert isinstance(into, sql_gen.Table)
494
495 2992 aaronmk
    into.is_temp = True
496
497 2790 aaronmk
    kw_args['recover'] = True
498 2945 aaronmk
    kw_args.setdefault('log_ignore_excs', (DuplicateException,))
499 2790 aaronmk
500 2916 aaronmk
    temp = not db.debug_temp # tables are permanent in debug_temp mode
501 2790 aaronmk
    # "temporary tables cannot specify a schema name", so remove schema
502
    if temp: into.schema = None
503
504
    # Create table
505
    while True:
506
        create_query = 'CREATE'
507
        if temp: create_query += ' TEMP'
508
        create_query += ' TABLE '+into.to_str(db)+' AS\n'+query
509 2385 aaronmk
510 2790 aaronmk
        try:
511
            cur = run_query(db, create_query, **kw_args)
512
                # CREATE TABLE AS sets rowcount to # rows in query
513
            break
514 2945 aaronmk
        except DuplicateException, e:
515 2790 aaronmk
            into.name = next_version(into.name)
516
            # try again with next version of name
517
518
    if add_indexes_: add_indexes(db, into)
519
520
    return cur
521 2085 aaronmk
522 2120 aaronmk
order_by_pkey = object() # tells mk_select() to order by the pkey
523
524 2199 aaronmk
distinct_on_all = object() # tells mk_select() to SELECT DISTINCT ON all columns
525
526 2233 aaronmk
def mk_select(db, tables, fields=None, conds=None, distinct_on=[], limit=None,
527 2293 aaronmk
    start=None, order_by=order_by_pkey, default_table=None):
528 1981 aaronmk
    '''
529 2121 aaronmk
    @param tables The single table to select from, or a list of tables to join
530 2280 aaronmk
        together, with tables after the first being sql_gen.Join objects
531 1981 aaronmk
    @param fields Use None to select all fields in the table
532 2377 aaronmk
    @param conds WHERE conditions: [(compare_left_side, compare_right_side),...]
533 2379 aaronmk
        * container can be any iterable type
534 2399 aaronmk
        * compare_left_side: sql_gen.Code|str (for col name)
535
        * compare_right_side: sql_gen.ValueCond|literal value
536 2199 aaronmk
    @param distinct_on The columns to SELECT DISTINCT ON, or distinct_on_all to
537
        use all columns
538 2786 aaronmk
    @return query
539 1981 aaronmk
    '''
540 2315 aaronmk
    # Parse tables param
541 2964 aaronmk
    tables = lists.mk_seq(tables)
542 2141 aaronmk
    tables = list(tables) # don't modify input! (list() copies input)
543 2315 aaronmk
    table0 = sql_gen.as_Table(tables.pop(0)) # first table is separate
544 2121 aaronmk
545 2315 aaronmk
    # Parse other params
546 2376 aaronmk
    if conds == None: conds = []
547 2650 aaronmk
    elif dicts.is_dict(conds): conds = conds.items()
548 2379 aaronmk
    conds = list(conds) # don't modify input! (list() copies input)
549 135 aaronmk
    assert limit == None or type(limit) == int
550 865 aaronmk
    assert start == None or type(start) == int
551 2315 aaronmk
    if order_by is order_by_pkey:
552
        if distinct_on != []: order_by = None
553
        else: order_by = pkey(db, table0, recover=True)
554 865 aaronmk
555 2315 aaronmk
    query = 'SELECT'
556 2056 aaronmk
557 2315 aaronmk
    def parse_col(col): return sql_gen.as_Col(col, default_table).to_str(db)
558 2056 aaronmk
559 2200 aaronmk
    # DISTINCT ON columns
560 2233 aaronmk
    if distinct_on != []:
561 2467 aaronmk
        query += '\nDISTINCT'
562 2254 aaronmk
        if distinct_on is not distinct_on_all:
563 2200 aaronmk
            query += ' ON ('+(', '.join(map(parse_col, distinct_on)))+')'
564
565
    # Columns
566 2467 aaronmk
    query += '\n'
567 1135 aaronmk
    if fields == None: query += '*'
568 2765 aaronmk
    else:
569
        assert fields != []
570
        query += '\n, '.join(map(parse_col, fields))
571 2200 aaronmk
572
    # Main table
573 2467 aaronmk
    query += '\nFROM '+table0.to_str(db)
574 865 aaronmk
575 2122 aaronmk
    # Add joins
576 2271 aaronmk
    left_table = table0
577 2263 aaronmk
    for join_ in tables:
578
        table = join_.table
579 2238 aaronmk
580 2343 aaronmk
        # Parse special values
581
        if join_.type_ is sql_gen.filter_out: # filter no match
582 2376 aaronmk
            conds.append((sql_gen.Col(table_not_null_col(db, table), table),
583 2853 aaronmk
                sql_gen.CompareCond(None, '~=')))
584 2343 aaronmk
585 2467 aaronmk
        query += '\n'+join_.to_str(db, left_table)
586 2122 aaronmk
587
        left_table = table
588
589 865 aaronmk
    missing = True
590 2376 aaronmk
    if conds != []:
591 2576 aaronmk
        if len(conds) == 1: whitespace = ' '
592
        else: whitespace = '\n'
593 2578 aaronmk
        query += '\n'+sql_gen.combine_conds([sql_gen.ColValueCond(l, r)
594
            .to_str(db) for l, r in conds], 'WHERE')
595 865 aaronmk
        missing = False
596 2227 aaronmk
    if order_by != None:
597 2467 aaronmk
        query += '\nORDER BY '+sql_gen.as_Col(order_by, table0).to_str(db)
598
    if limit != None: query += '\nLIMIT '+str(limit); missing = False
599 865 aaronmk
    if start != None:
600 2467 aaronmk
        if start != 0: query += '\nOFFSET '+str(start)
601 865 aaronmk
        missing = False
602
    if missing: warnings.warn(DbWarning(
603
        'SELECT statement missing a WHERE, LIMIT, or OFFSET clause: '+query))
604
605 2786 aaronmk
    return query
606 11 aaronmk
607 2054 aaronmk
def select(db, *args, **kw_args):
608
    '''For params, see mk_select() and run_query()'''
609
    recover = kw_args.pop('recover', None)
610
    cacheable = kw_args.pop('cacheable', True)
611 2442 aaronmk
    log_level = kw_args.pop('log_level', 2)
612 2054 aaronmk
613 2791 aaronmk
    return run_query(db, mk_select(db, *args, **kw_args), recover, cacheable,
614
        log_level=log_level)
615 2054 aaronmk
616 2788 aaronmk
def mk_insert_select(db, table, cols=None, select_query=None, returning=None,
617
    embeddable=False):
618 1960 aaronmk
    '''
619
    @param returning str|None An inserted column (such as pkey) to return
620 2070 aaronmk
    @param embeddable Whether the query should be embeddable as a nested SELECT.
621 2073 aaronmk
        Warning: If you set this and cacheable=True when the query is run, the
622
        query will be fully cached, not just if it raises an exception.
623 1960 aaronmk
    '''
624 2754 aaronmk
    table = sql_gen.remove_table_rename(sql_gen.as_Table(table))
625 2318 aaronmk
    if cols == []: cols = None # no cols (all defaults) = unknown col names
626 2646 aaronmk
    if cols != None:
627
        cols = [sql_gen.to_name_only_col(v, table).to_str(db) for v in cols]
628 2063 aaronmk
    if select_query == None: select_query = 'DEFAULT VALUES'
629 2327 aaronmk
    if returning != None: returning = sql_gen.as_Col(returning, table)
630 2063 aaronmk
631
    # Build query
632 2497 aaronmk
    first_line = 'INSERT INTO '+table.to_str(db)
633
    query = first_line
634 2467 aaronmk
    if cols != None: query += '\n('+', '.join(cols)+')'
635
    query += '\n'+select_query
636 2063 aaronmk
637
    if returning != None:
638 2743 aaronmk
        query += '\nRETURNING '+sql_gen.to_name_only_col(returning).to_str(db)
639 2063 aaronmk
640 2070 aaronmk
    if embeddable:
641 2327 aaronmk
        assert returning != None
642
643 2070 aaronmk
        # Create function
644 2513 aaronmk
        function_name = sql_gen.clean_name(first_line)
645 2327 aaronmk
        return_type = 'SETOF '+returning.to_str(db)+'%TYPE'
646 2189 aaronmk
        while True:
647
            try:
648 2918 aaronmk
                function = db.TempFunction(function_name)
649 2194 aaronmk
650 2189 aaronmk
                function_query = '''\
651 2698 aaronmk
CREATE FUNCTION '''+function.to_str(db)+'''()
652 2467 aaronmk
RETURNS '''+return_type+'''
653
LANGUAGE sql
654
AS $$
655 2788 aaronmk
'''+query+''';
656 2467 aaronmk
$$;
657 2070 aaronmk
'''
658 2446 aaronmk
                run_query(db, function_query, recover=True, cacheable=True,
659 2945 aaronmk
                    log_ignore_excs=(DuplicateException,))
660 2189 aaronmk
                break # this version was successful
661 2945 aaronmk
            except DuplicateException, e:
662 2189 aaronmk
                function_name = next_version(function_name)
663
                # try again with next version of name
664 2070 aaronmk
665 2337 aaronmk
        # Return query that uses function
666 2698 aaronmk
        func_table = sql_gen.NamedTable('f', sql_gen.FunctionCall(function),
667 2742 aaronmk
            [returning]) # AS clause requires function alias
668 2787 aaronmk
        return mk_select(db, func_table, start=0, order_by=None)
669 2070 aaronmk
670 2787 aaronmk
    return query
671 2066 aaronmk
672
def insert_select(db, *args, **kw_args):
673 2085 aaronmk
    '''For params, see mk_insert_select() and run_query_into()
674 2386 aaronmk
    @param into sql_gen.Table with suggested name of temp table to put RETURNING
675
        values in
676 2072 aaronmk
    '''
677 2386 aaronmk
    into = kw_args.pop('into', None)
678
    if into != None: kw_args['embeddable'] = True
679 2066 aaronmk
    recover = kw_args.pop('recover', None)
680
    cacheable = kw_args.pop('cacheable', True)
681 2673 aaronmk
    log_level = kw_args.pop('log_level', 2)
682 2066 aaronmk
683 2789 aaronmk
    return run_query_into(db, mk_insert_select(db, *args, **kw_args), into,
684 2787 aaronmk
        recover=recover, cacheable=cacheable, log_level=log_level)
685 2063 aaronmk
686 2738 aaronmk
default = sql_gen.default # tells insert() to use the default value for a column
687 2066 aaronmk
688 2063 aaronmk
def insert(db, table, row, *args, **kw_args):
689 2085 aaronmk
    '''For params, see insert_select()'''
690 1960 aaronmk
    if lists.is_seq(row): cols = None
691
    else:
692
        cols = row.keys()
693
        row = row.values()
694 2738 aaronmk
    row = list(row) # ensure that "== []" works
695 1960 aaronmk
696 2738 aaronmk
    if row == []: query = None
697
    else: query = sql_gen.Values(row).to_str(db)
698 1961 aaronmk
699 2788 aaronmk
    return insert_select(db, table, cols, query, *args, **kw_args)
700 11 aaronmk
701 2402 aaronmk
def mk_update(db, table, changes=None, cond=None):
702
    '''
703
    @param changes [(col, new_value),...]
704
        * container can be any iterable type
705
        * col: sql_gen.Code|str (for col name)
706
        * new_value: sql_gen.Code|literal value
707
    @param cond sql_gen.Code WHERE condition. e.g. use sql_gen.*Cond objects.
708
    @return str query
709
    '''
710
    query = 'UPDATE '+sql_gen.as_Table(table).to_str(db)+'\nSET\n'
711 2405 aaronmk
    query += ',\n'.join((sql_gen.to_name_only_col(col, table).to_str(db)+' = '
712 2402 aaronmk
        +sql_gen.as_Value(new_value).to_str(db) for col, new_value in changes))
713 2467 aaronmk
    if cond != None: query += '\nWHERE\n'+cond.to_str(db)
714 2402 aaronmk
715
    return query
716
717
def update(db, *args, **kw_args):
718
    '''For params, see mk_update() and run_query()'''
719
    recover = kw_args.pop('recover', None)
720
721 2791 aaronmk
    return run_query(db, mk_update(db, *args, **kw_args), recover)
722 2402 aaronmk
723 135 aaronmk
def last_insert_id(db):
724 1849 aaronmk
    module = util.root_module(db.db)
725 135 aaronmk
    if module == 'psycopg2': return value(run_query(db, 'SELECT lastval()'))
726
    elif module == 'MySQLdb': return db.insert_id()
727
    else: return None
728 13 aaronmk
729 2394 aaronmk
def mk_flatten_mapping(db, into, cols, preserve=[], as_items=False):
730 2383 aaronmk
    '''Creates a mapping from original column names (which may have collisions)
731 2415 aaronmk
    to names that will be distinct among the columns' tables.
732 2383 aaronmk
    This is meant to be used for several tables that are being joined together.
733 2415 aaronmk
    @param cols The columns to combine. Duplicates will be removed.
734
    @param into The table for the new columns.
735 2394 aaronmk
    @param preserve [sql_gen.Col...] Columns not to rename. Note that these
736
        columns will be included in the mapping even if they are not in cols.
737
        The tables of the provided Col objects will be changed to into, so make
738
        copies of them if you want to keep the original tables.
739
    @param as_items Whether to return a list of dict items instead of a dict
740 2383 aaronmk
    @return dict(orig_col=new_col, ...)
741
        * orig_col: sql_gen.Col(orig_col_name, orig_table)
742 2392 aaronmk
        * new_col: sql_gen.Col(orig_col_name, into)
743
        * All mappings use the into table so its name can easily be
744 2383 aaronmk
          changed for all columns at once
745
    '''
746 2415 aaronmk
    cols = lists.uniqify(cols)
747
748 2394 aaronmk
    items = []
749 2389 aaronmk
    for col in preserve:
750 2390 aaronmk
        orig_col = copy.copy(col)
751 2392 aaronmk
        col.table = into
752 2394 aaronmk
        items.append((orig_col, col))
753
    preserve = set(preserve)
754
    for col in cols:
755 2716 aaronmk
        if col not in preserve:
756
            items.append((col, sql_gen.Col(str(col), into, col.srcs)))
757 2394 aaronmk
758
    if not as_items: items = dict(items)
759
    return items
760 2383 aaronmk
761 2393 aaronmk
def flatten(db, into, joins, cols, limit=None, start=None, **kw_args):
762 2391 aaronmk
    '''For params, see mk_flatten_mapping()
763
    @return See return value of mk_flatten_mapping()
764
    '''
765 2394 aaronmk
    items = mk_flatten_mapping(db, into, cols, as_items=True, **kw_args)
766
    cols = [sql_gen.NamedCol(new.name, old) for old, new in items]
767 2786 aaronmk
    run_query_into(db, mk_select(db, joins, cols, limit=limit, start=start),
768 2846 aaronmk
        into=into, add_indexes_=True)
769 2394 aaronmk
    return dict(items)
770 2391 aaronmk
771 2751 aaronmk
def mk_track_data_error(db, errors_table, cols, value, error_code, error):
772 2810 aaronmk
    assert cols != ()
773
774 2751 aaronmk
    cols = map(sql_gen.to_name_only_col, cols)
775
776
    columns_cols = ['column']
777
    columns = sql_gen.NamedValues('columns', columns_cols,
778
        [[c.name] for c in cols])
779
    values_cols = ['value', 'error_code', 'error']
780
    values = sql_gen.NamedValues('values', values_cols,
781
        [value, error_code, error])
782
783
    select_cols = columns_cols+values_cols
784
    name_only_cols = map(sql_gen.to_name_only_col, select_cols)
785 2755 aaronmk
    errors_table = sql_gen.NamedTable('errors', errors_table)
786 2751 aaronmk
    joins = [columns, sql_gen.Join(values, type_='CROSS'),
787
        sql_gen.Join(errors_table, dict(zip(name_only_cols, select_cols)),
788
        sql_gen.filter_out)]
789
790
    return mk_insert_select(db, errors_table, name_only_cols,
791 2787 aaronmk
        mk_select(db, joins, select_cols, order_by=None))
792 2751 aaronmk
793 2810 aaronmk
def track_data_error(db, errors_table, cols, *args, **kw_args):
794 2758 aaronmk
    '''
795
    @param errors_table If None, does nothing.
796
    '''
797 2810 aaronmk
    if errors_table == None or cols == (): return
798
    run_query(db, mk_track_data_error(db, errors_table, cols, *args, **kw_args),
799 2758 aaronmk
        cacheable=True, log_level=4)
800
801 2704 aaronmk
def cast(db, type_, col, errors_table=None):
802
    '''Casts an (unrenamed) column or value.
803 2769 aaronmk
    If errors_table set and col has srcs, saves errors in errors_table (using
804
    col's srcs attr as the source columns) and converts errors to warnings.
805 2820 aaronmk
    @param col str|sql_gen.Col|sql_gen.Literal
806 2704 aaronmk
    @param errors_table None|sql_gen.Table|str
807
    '''
808 2820 aaronmk
    col = sql_gen.as_Col(col)
809 2769 aaronmk
    save_errors = (errors_table != None and isinstance(col, sql_gen.Col)
810
        and col.srcs != ())
811 2987 aaronmk
    if not save_errors: return sql_gen.Cast(type_, col) # can't save errors
812 2704 aaronmk
813
    assert not isinstance(col, sql_gen.NamedCol)
814
815 2770 aaronmk
    errors_table = sql_gen.as_Table(errors_table)
816
    srcs = map(sql_gen.to_name_only_col, col.srcs)
817
    function_name = str(sql_gen.FunctionCall(type_, *srcs))
818 2918 aaronmk
    function = db.TempFunction(function_name)
819 2704 aaronmk
820 2718 aaronmk
    while True:
821
        # Create function definition
822
        query = '''\
823 2704 aaronmk
CREATE FUNCTION '''+function.to_str(db)+'''(value text)
824
RETURNS '''+type_+'''
825
LANGUAGE plpgsql
826
STRICT
827
AS $$
828
BEGIN
829
    /* The explicit cast to the return type is needed to make the cast happen
830
    inside the try block. (Implicit casts to the return type happen at the end
831
    of the function, outside any block.) */
832
    RETURN value::'''+type_+''';
833
EXCEPTION
834
    WHEN data_exception THEN
835
        -- Save error in errors table.
836 2751 aaronmk
        -- Insert the value and error for *each* source column.
837
'''+mk_track_data_error(db, errors_table, srcs,
838
    *map(sql_gen.CustomCode, ['value', 'SQLSTATE', 'SQLERRM']))+''';
839 2704 aaronmk
840
        RAISE WARNING '%', SQLERRM;
841
        RETURN NULL;
842
END;
843
$$;
844
'''
845 2718 aaronmk
846
        # Create function
847
        try:
848
            run_query(db, query, recover=True, cacheable=True,
849 2945 aaronmk
                log_ignore_excs=(DuplicateException,))
850 2718 aaronmk
            break # successful
851 2945 aaronmk
        except DuplicateException:
852 2770 aaronmk
            function.name = next_version(function.name)
853
            # try again with next version of name
854 2704 aaronmk
855
    return sql_gen.FunctionCall(function, col)
856
857 2414 aaronmk
##### Database structure queries
858
859 2426 aaronmk
def table_row_count(db, table, recover=None):
860 2786 aaronmk
    return value(run_query(db, mk_select(db, table, [sql_gen.row_count],
861 2443 aaronmk
        order_by=None, start=0), recover=recover, log_level=3))
862 2426 aaronmk
863 2414 aaronmk
def table_cols(db, table, recover=None):
864
    return list(col_names(select(db, table, limit=0, order_by=None,
865 2443 aaronmk
        recover=recover, log_level=4)))
866 2414 aaronmk
867 2291 aaronmk
def pkey(db, table, recover=None):
868 832 aaronmk
    '''Assumed to be first column in table'''
869 2339 aaronmk
    return table_cols(db, table, recover)[0]
870 832 aaronmk
871 2559 aaronmk
not_null_col = 'not_null_col'
872 2340 aaronmk
873
def table_not_null_col(db, table, recover=None):
874
    '''Name assumed to be the value of not_null_col. If not found, uses pkey.'''
875
    if not_null_col in table_cols(db, table, recover): return not_null_col
876
    else: return pkey(db, table, recover)
877
878 853 aaronmk
def index_cols(db, table, index):
879
    '''Can also use this for UNIQUE constraints, because a UNIQUE index is
880
    automatically created. When you don't know whether something is a UNIQUE
881
    constraint or a UNIQUE index, use this function.'''
882 1909 aaronmk
    module = util.root_module(db.db)
883
    if module == 'psycopg2':
884
        return list(values(run_query(db, '''\
885 853 aaronmk
SELECT attname
886 866 aaronmk
FROM
887
(
888
        SELECT attnum, attname
889
        FROM pg_index
890
        JOIN pg_class index ON index.oid = indexrelid
891
        JOIN pg_class table_ ON table_.oid = indrelid
892
        JOIN pg_attribute ON attrelid = indrelid AND attnum = ANY (indkey)
893
        WHERE
894 2782 aaronmk
            table_.relname = '''+db.esc_value(table)+'''
895
            AND index.relname = '''+db.esc_value(index)+'''
896 866 aaronmk
    UNION
897
        SELECT attnum, attname
898
        FROM
899
        (
900
            SELECT
901
                indrelid
902
                , (regexp_matches(indexprs, E':varattno (\\\\d+)', 'g'))[1]::int
903
                    AS indkey
904
            FROM pg_index
905
            JOIN pg_class index ON index.oid = indexrelid
906
            JOIN pg_class table_ ON table_.oid = indrelid
907
            WHERE
908 2782 aaronmk
                table_.relname = '''+db.esc_value(table)+'''
909
                AND index.relname = '''+db.esc_value(index)+'''
910 866 aaronmk
        ) s
911
        JOIN pg_attribute ON attrelid = indrelid AND attnum = indkey
912
) s
913 853 aaronmk
ORDER BY attnum
914 2782 aaronmk
'''
915
            , cacheable=True, log_level=4)))
916 1909 aaronmk
    else: raise NotImplementedError("Can't list index columns for "+module+
917
        ' database')
918 853 aaronmk
919 464 aaronmk
def constraint_cols(db, table, constraint):
920 1849 aaronmk
    module = util.root_module(db.db)
921 464 aaronmk
    if module == 'psycopg2':
922
        return list(values(run_query(db, '''\
923
SELECT attname
924
FROM pg_constraint
925
JOIN pg_class ON pg_class.oid = conrelid
926
JOIN pg_attribute ON attrelid = conrelid AND attnum = ANY (conkey)
927
WHERE
928 2783 aaronmk
    relname = '''+db.esc_value(table)+'''
929
    AND conname = '''+db.esc_value(constraint)+'''
930 464 aaronmk
ORDER BY attnum
931 2783 aaronmk
'''
932
            )))
933 464 aaronmk
    else: raise NotImplementedError("Can't list constraint columns for "+module+
934
        ' database')
935
936 2096 aaronmk
row_num_col = '_row_num'
937
938 2847 aaronmk
def add_index(db, exprs, table=None, unique=False, ensure_not_null=True):
939 2688 aaronmk
    '''Adds an index on column(s) or expression(s) if it doesn't already exist.
940 2538 aaronmk
    Currently, only function calls are supported as expressions.
941 2847 aaronmk
    @param ensure_not_null If set, translates NULL values to sentinel values.
942
        This allows indexes to be used for comparisons where NULLs are equal.
943 2538 aaronmk
    '''
944 2964 aaronmk
    exprs = lists.mk_seq(exprs)
945 2538 aaronmk
946 2688 aaronmk
    # Parse exprs
947
    old_exprs = exprs[:]
948
    exprs = []
949
    cols = []
950
    for i, expr in enumerate(old_exprs):
951
        expr = copy.deepcopy(expr) # don't modify input!
952 2823 aaronmk
        expr = sql_gen.as_Col(expr, table)
953 2688 aaronmk
954 2823 aaronmk
        # Handle nullable columns
955 2860 aaronmk
        if ensure_not_null:
956
            try: expr = sql_gen.ensure_not_null(db, expr)
957
            except KeyError: pass # unknown type, so just create plain index
958 2823 aaronmk
959 2688 aaronmk
        # Extract col
960
        if isinstance(expr, sql_gen.FunctionCall):
961
            col = expr.args[0]
962
            expr = sql_gen.Expr(expr)
963
        else: col = expr
964 2823 aaronmk
        assert isinstance(col, sql_gen.Col)
965 2688 aaronmk
966
        # Extract table
967
        if table == None:
968
            assert sql_gen.is_table_col(col)
969
            table = col.table
970
971
        col.table = None
972
973
        exprs.append(expr)
974
        cols.append(col)
975 2408 aaronmk
976 2688 aaronmk
    table = sql_gen.as_Table(table)
977
    index = sql_gen.Table(str(sql_gen.Col(','.join(map(str, cols)), table)))
978
979
    str_ = 'CREATE'
980
    if unique: str_ += ' UNIQUE'
981
    str_ += ' INDEX '+index.to_str(db)+' ON '+table.to_str(db)+' ('+(
982
        ', '.join((v.to_str(db) for v in exprs)))+')'
983
984
    try: run_query(db, str_, recover=True, cacheable=True, log_level=3)
985 2945 aaronmk
    except DuplicateException: pass # index already existed
986 2408 aaronmk
987 2686 aaronmk
def add_pkey(db, table, cols=None, recover=None):
988
    '''Adds a primary key.
989
    @param cols [sql_gen.Col,...] The columns in the primary key.
990
        Defaults to the first column in the table.
991 2406 aaronmk
    @pre The table must not already have a primary key.
992
    '''
993
    table = sql_gen.as_Table(table)
994 2686 aaronmk
    if cols == None: cols = [pkey(db, table, recover)]
995
    col_strs = [sql_gen.to_name_only_col(v).to_str(db) for v in cols]
996 2406 aaronmk
997 2781 aaronmk
    run_query(db, 'ALTER TABLE '+table.to_str(db)+' ADD PRIMARY KEY ('
998
        +(', '.join(col_strs))+')', recover=True, cacheable=True, log_level=3,
999 2945 aaronmk
        log_ignore_excs=(DuplicateException,))
1000 2406 aaronmk
1001 2772 aaronmk
already_indexed = object() # tells add_indexes() the pkey has already been added
1002
1003
def add_indexes(db, table, has_pkey=True):
1004
    '''Adds an index on all columns in a table.
1005
    @param has_pkey bool|already_indexed Whether a pkey instead of a regular
1006
        index should be added on the first column.
1007
        * If already_indexed, the pkey is assumed to have already been added
1008
    '''
1009
    cols = table_cols(db, table)
1010
    if has_pkey:
1011
        if has_pkey is not already_indexed: add_pkey(db, table)
1012
        cols = cols[1:]
1013
    for col in cols: add_index(db, col, table)
1014
1015 2935 aaronmk
def add_col(db, table, col, **kw_args):
1016 2934 aaronmk
    assert isinstance(col, sql_gen.TypedCol)
1017 2947 aaronmk
    try: run_query(db, 'ALTER TABLE '+table.to_str(db)+' ADD COLUMN '
1018
        +col.to_str(db), recover=True, cacheable=True, **kw_args)
1019
    except DuplicateException: pass # column already existed
1020 2934 aaronmk
1021 2935 aaronmk
row_num_typed_col = sql_gen.TypedCol(row_num_col, 'serial', nullable=False,
1022
    constraints='PRIMARY KEY')
1023
1024 2086 aaronmk
def add_row_num(db, table):
1025 2117 aaronmk
    '''Adds a row number column to a table. Its name is in row_num_col. It will
1026
    be the primary key.'''
1027 2935 aaronmk
    add_col(db, table, row_num_typed_col, log_level=3)
1028 2086 aaronmk
1029 2936 aaronmk
def cast_temp_col(db, type_, col, errors_table=None):
1030
    '''Like cast(), but creates a new column with the cast values if the input
1031
    is a column.
1032
    @return The new column or cast value
1033
    '''
1034
    def cast_(col): return cast(db, type_, col, errors_table)
1035
1036
    try: col = sql_gen.underlying_col(col)
1037
    except sql_gen.NoUnderlyingTableException: return sql_gen.wrap(cast_, col)
1038
1039
    table = col.table
1040
    new_col = sql_gen.Col(sql_gen.concat(col.name, '::'+type_), table, col.srcs)
1041
    expr = cast_(col)
1042
    add_col(db, table, sql_gen.TypedCol(new_col.name, type_))
1043
    update(db, table, [(new_col, expr)])
1044 2955 aaronmk
    add_index(db, new_col)
1045 2936 aaronmk
1046
    return new_col
1047
1048 2889 aaronmk
def drop_table(db, table):
1049
    table = sql_gen.as_Table(table)
1050
    return run_query(db, 'DROP TABLE IF EXISTS '+table.to_str(db)+' CASCADE')
1051
1052 2760 aaronmk
def create_table(db, table, cols, has_pkey=True, col_indexes=True):
1053 2675 aaronmk
    '''Creates a table.
1054 2681 aaronmk
    @param cols [sql_gen.TypedCol,...] The column names and types
1055
    @param has_pkey If set, the first column becomes the primary key.
1056 2760 aaronmk
    @param col_indexes bool|[ref]
1057
        * If True, indexes will be added on all non-pkey columns.
1058
        * If a list reference, [0] will be set to a function to do this.
1059
          This can be used to delay index creation until the table is populated.
1060 2675 aaronmk
    '''
1061
    table = sql_gen.as_Table(table)
1062
1063 2681 aaronmk
    if has_pkey:
1064
        cols[0] = pkey = copy.copy(cols[0]) # don't modify input!
1065 2872 aaronmk
        pkey.constraints = 'PRIMARY KEY'
1066 2681 aaronmk
1067 2675 aaronmk
    str_ = 'CREATE TABLE '+table.to_str(db)+' (\n'
1068 2681 aaronmk
    str_ += '\n, '.join(v.to_str(db) for v in cols)
1069
    str_ += '\n);\n'
1070 2675 aaronmk
    run_query(db, str_, cacheable=True, log_level=2)
1071 2760 aaronmk
1072
    # Add indexes
1073 2773 aaronmk
    if has_pkey: has_pkey = already_indexed
1074
    def add_indexes_(): add_indexes(db, table, has_pkey)
1075
    if isinstance(col_indexes, list): col_indexes[0] = add_indexes_ # defer
1076
    elif col_indexes: add_indexes_() # add now
1077 2675 aaronmk
1078 2684 aaronmk
def vacuum(db, table):
1079
    table = sql_gen.as_Table(table)
1080
    db.with_autocommit(lambda: run_query(db, 'VACUUM ANALYZE '+table.to_str(db),
1081
        log_level=3))
1082
1083 2970 aaronmk
def truncate(db, table, schema='public', **kw_args):
1084
    '''For params, see run_query()'''
1085 2777 aaronmk
    table = sql_gen.as_Table(table, schema)
1086 2970 aaronmk
    return run_query(db, 'TRUNCATE '+table.to_str(db)+' CASCADE', **kw_args)
1087 2732 aaronmk
1088 2965 aaronmk
def empty_temp(db, tables):
1089 2972 aaronmk
    if db.debug_temp: return # leave temp tables there for debugging
1090 2965 aaronmk
    tables = lists.mk_seq(tables)
1091 2971 aaronmk
    for table in tables: truncate(db, table, log_level=3)
1092 2965 aaronmk
1093 2729 aaronmk
def tables(db, schema_like='public', table_like='%', exact=False):
1094
    if exact: compare = '='
1095
    else: compare = 'LIKE'
1096
1097 1849 aaronmk
    module = util.root_module(db.db)
1098 832 aaronmk
    if module == 'psycopg2':
1099 2779 aaronmk
        conds = [('schemaname', sql_gen.CompareCond(schema_like, compare)),
1100
            ('tablename', sql_gen.CompareCond(table_like, compare))]
1101
        return values(select(db, 'pg_tables', ['tablename'], conds,
1102
            order_by='tablename', log_level=4))
1103 1968 aaronmk
    elif module == 'MySQLdb':
1104 2779 aaronmk
        return values(run_query(db, 'SHOW TABLES LIKE '+db.esc_value(table_like)
1105
            , cacheable=True, log_level=4))
1106 832 aaronmk
    else: raise NotImplementedError("Can't list tables for "+module+' database')
1107 830 aaronmk
1108 2726 aaronmk
def table_exists(db, table):
1109
    table = sql_gen.as_Table(table)
1110 2730 aaronmk
    return list(tables(db, table.schema, table.name, exact=True)) != []
1111 2726 aaronmk
1112 2913 aaronmk
def function_exists(db, function):
1113
    function = sql_gen.as_Function(function)
1114
1115
    info_table = sql_gen.Table('routines', 'information_schema')
1116
    conds = [('routine_name', function.name)]
1117
    schema = function.schema
1118
    if schema != None: conds.append(('routine_schema', schema))
1119 2914 aaronmk
    # Exclude trigger functions, since they cannot be called directly
1120
    conds.append(('data_type', sql_gen.CompareCond('trigger', '!=')))
1121 2913 aaronmk
1122
    return list(values(select(db, info_table, ['routine_name'], conds,
1123
        order_by='routine_schema', limit=1, log_level=4))) != []
1124
        # TODO: order_by search_path schema order
1125
1126 2726 aaronmk
def errors_table(db, table, if_exists=True):
1127
    '''
1128
    @param if_exists If set, returns None if the errors table doesn't exist
1129
    @return None|sql_gen.Table
1130
    '''
1131
    table = sql_gen.as_Table(table)
1132
    if table.srcs != (): table = table.srcs[0]
1133
1134
    errors_table = sql_gen.suffixed_table(table, '.errors')
1135
    if if_exists and not table_exists(db, errors_table): return None
1136
    return errors_table
1137
1138 833 aaronmk
##### Database management
1139
1140 1968 aaronmk
def empty_db(db, schema='public', **kw_args):
1141
    '''For kw_args, see tables()'''
1142
    for table in tables(db, schema, **kw_args): truncate(db, table, schema)
1143 833 aaronmk
1144 832 aaronmk
##### Heuristic queries
1145
1146 2104 aaronmk
def put(db, table, row, pkey_=None, row_ct_ref=None):
1147 1554 aaronmk
    '''Recovers from errors.
1148 2077 aaronmk
    Only works under PostgreSQL (uses INSERT RETURNING).
1149
    '''
1150 2642 aaronmk
    row = sql_gen.ColDict(db, table, row)
1151 2104 aaronmk
    if pkey_ == None: pkey_ = pkey(db, table, recover=True)
1152
1153 471 aaronmk
    try:
1154 2149 aaronmk
        cur = insert(db, table, row, pkey_, recover=True)
1155 1554 aaronmk
        if row_ct_ref != None and cur.rowcount >= 0:
1156
            row_ct_ref[0] += cur.rowcount
1157
        return value(cur)
1158 471 aaronmk
    except DuplicateKeyException, e:
1159 2653 aaronmk
        row = sql_gen.ColDict(db, table,
1160
            util.dict_subset_right_join(row, e.cols))
1161
        return value(select(db, table, [pkey_], row, recover=True))
1162 471 aaronmk
1163 473 aaronmk
def get(db, table, row, pkey, row_ct_ref=None, create=False):
1164 830 aaronmk
    '''Recovers from errors'''
1165 2209 aaronmk
    try: return value(select(db, table, [pkey], row, limit=1, recover=True))
1166 14 aaronmk
    except StopIteration:
1167 40 aaronmk
        if not create: raise
1168 471 aaronmk
        return put(db, table, row, pkey, row_ct_ref) # insert new row
1169 2078 aaronmk
1170 2593 aaronmk
def is_func_result(col):
1171
    return col.table.name.find('(') >= 0 and col.name == 'result'
1172
1173 2592 aaronmk
def into_table_name(out_table, in_tables0, mapping, is_func):
1174 2615 aaronmk
    def in_col_str(in_col):
1175
        in_col = sql_gen.remove_col_rename(in_col)
1176
        if isinstance(in_col, sql_gen.Col):
1177
            table = in_col.table
1178
            if table == in_tables0:
1179
                in_col = sql_gen.to_name_only_col(in_col)
1180
            elif is_func_result(in_col): in_col = table # omit col name
1181
        return str(in_col)
1182
1183 2580 aaronmk
    str_ = str(out_table)
1184
    if is_func:
1185 2615 aaronmk
        str_ += '('
1186 2580 aaronmk
1187 2615 aaronmk
        try: value_in_col = mapping['value']
1188
        except KeyError:
1189 2654 aaronmk
            str_ += ', '.join((str(k)+'='+in_col_str(v)
1190 2615 aaronmk
                for k, v in mapping.iteritems()))
1191
        else: str_ += in_col_str(value_in_col)
1192
1193
        str_ += ')'
1194 2812 aaronmk
    else:
1195
        out_col = 'rank'
1196
        try: in_col = mapping[out_col]
1197
        except KeyError: str_ += '_pkeys'
1198
        else: # has a rank column, so hierarchical
1199
            str_ += '['+str(out_col)+'='+in_col_str(in_col)+']'
1200 2580 aaronmk
    return str_
1201
1202 2508 aaronmk
def put_table(db, out_table, in_tables, mapping, row_ct_ref=None, into=None,
1203 2805 aaronmk
    default=None, is_func=False, on_error=exc.raise_):
1204 2078 aaronmk
    '''Recovers from errors.
1205
    Only works under PostgreSQL (uses INSERT RETURNING).
1206 2909 aaronmk
    IMPORTANT: Must be run at the *beginning* of a transaction.
1207 2131 aaronmk
    @param in_tables The main input table to select from, followed by a list of
1208
        tables to join with it using the main input table's pkey
1209 2312 aaronmk
    @param mapping dict(out_table_col=in_table_col, ...)
1210 2612 aaronmk
        * out_table_col: str (*not* sql_gen.Col)
1211
        * in_table_col: sql_gen.Col|literal-value
1212 2489 aaronmk
    @param into The table to contain the output and input pkeys.
1213 2574 aaronmk
        Defaults to `out_table.name+'_pkeys'`.
1214 2509 aaronmk
    @param default The *output* column to use as the pkey for missing rows.
1215
        If this output column does not exist in the mapping, uses None.
1216 2552 aaronmk
    @param is_func Whether out_table is the name of a SQL function, not a table
1217 2312 aaronmk
    @return sql_gen.Col Where the output pkeys are made available
1218 2078 aaronmk
    '''
1219 2329 aaronmk
    out_table = sql_gen.as_Table(out_table)
1220 2312 aaronmk
1221 2450 aaronmk
    def log_debug(msg): db.log_debug(msg, level=1.5)
1222 2505 aaronmk
    def col_ustr(str_):
1223 2567 aaronmk
        return strings.repr_no_u(sql_gen.remove_col_rename(str_))
1224 2450 aaronmk
1225 2910 aaronmk
    log_debug('********** New iteration **********')
1226
    log_debug('Inserting these input columns into '+strings.as_tt(
1227
        out_table.to_str(db))+':\n'+strings.as_table(mapping, ustr=col_ustr))
1228
1229 2942 aaronmk
    is_function = function_exists(db, out_table)
1230
1231
    if is_function: out_pkey = 'result'
1232
    else: out_pkey = pkey(db, out_table, recover=True)
1233 2768 aaronmk
    out_pkey_col = sql_gen.as_Col(out_pkey, out_table)
1234
1235
    if mapping == {}: # need at least one column for INSERT SELECT
1236
        mapping = {out_pkey: None} # ColDict will replace with default value
1237
1238 2382 aaronmk
    # Create input joins from list of input tables
1239
    in_tables_ = in_tables[:] # don't modify input!
1240
    in_tables0 = in_tables_.pop(0) # first table is separate
1241 2735 aaronmk
    errors_table_ = errors_table(db, in_tables0)
1242 2279 aaronmk
    in_pkey = pkey(db, in_tables0, recover=True)
1243 2285 aaronmk
    in_pkey_col = sql_gen.as_Col(in_pkey, in_tables0)
1244 2460 aaronmk
    input_joins = [in_tables0]+[sql_gen.Join(v,
1245
        {in_pkey: sql_gen.join_same_not_null}) for v in in_tables_]
1246 2131 aaronmk
1247 2592 aaronmk
    if into == None:
1248
        into = into_table_name(out_table, in_tables0, mapping, is_func)
1249
    into = sql_gen.as_Table(into)
1250
1251 2702 aaronmk
    # Set column sources
1252
    in_cols = filter(sql_gen.is_table_col, mapping.values())
1253
    for col in in_cols:
1254
        if col.table == in_tables0: col.set_srcs(sql_gen.src_self)
1255
1256 2486 aaronmk
    log_debug('Joining together input tables into temp table')
1257 2395 aaronmk
    # Place in new table for speed and so don't modify input if values edited
1258 2584 aaronmk
    in_table = sql_gen.Table('in')
1259 2702 aaronmk
    mapping = dicts.join(mapping, flatten(db, in_table, input_joins, in_cols,
1260
        preserve=[in_pkey_col], start=0))
1261 2395 aaronmk
    input_joins = [in_table]
1262 2486 aaronmk
    db.log_debug('Temp table: '+strings.as_tt(in_table.to_str(db)), level=2)
1263 2395 aaronmk
1264 2656 aaronmk
    mapping = sql_gen.ColDict(db, out_table, mapping)
1265
        # after applying dicts.join() because that returns a plain dict
1266
1267 2509 aaronmk
    # Resolve default value column
1268 2841 aaronmk
    if default != None:
1269
        try: default = mapping[default]
1270
        except KeyError:
1271 2509 aaronmk
            db.log_debug('Default value column '
1272
                +strings.as_tt(strings.repr_no_u(default))
1273 2511 aaronmk
                +' does not exist in mapping, falling back to None', level=2.1)
1274 2509 aaronmk
            default = None
1275
1276 2387 aaronmk
    pkeys_names = [in_pkey, out_pkey]
1277 2236 aaronmk
    pkeys_cols = [in_pkey_col, out_pkey_col]
1278
1279 2201 aaronmk
    pkeys_table_exists_ref = [False]
1280 2960 aaronmk
    def insert_into_pkeys(joins, cols, distinct=False):
1281
        kw_args = {}
1282
        if distinct: kw_args.update(dict(distinct_on=[in_pkey_col]))
1283
        query = mk_select(db, joins, cols, order_by=None, start=0, **kw_args)
1284
1285 2201 aaronmk
        if pkeys_table_exists_ref[0]:
1286 2786 aaronmk
            insert_select(db, into, pkeys_names, query)
1287 2201 aaronmk
        else:
1288 2786 aaronmk
            run_query_into(db, query, into=into)
1289 2201 aaronmk
            pkeys_table_exists_ref[0] = True
1290
1291 2429 aaronmk
    limit_ref = [None]
1292 2380 aaronmk
    conds = set()
1293 2811 aaronmk
    distinct_on = sql_gen.ColDict(db, out_table)
1294 2325 aaronmk
    def mk_main_select(joins, cols):
1295 2811 aaronmk
        distinct_on_cols = [c.to_Col() for c in distinct_on.values()]
1296
        return mk_select(db, joins, cols, conds, distinct_on_cols,
1297 2429 aaronmk
            limit=limit_ref[0], start=0)
1298 2132 aaronmk
1299 2519 aaronmk
    exc_strs = set()
1300 2309 aaronmk
    def log_exc(e):
1301 2519 aaronmk
        e_str = exc.str_(e, first_line_only=True)
1302
        log_debug('Caught exception: '+e_str)
1303
        assert e_str not in exc_strs # avoid infinite loops
1304
        exc_strs.add(e_str)
1305 2735 aaronmk
1306 2451 aaronmk
    def remove_all_rows():
1307 2896 aaronmk
        log_debug('Ignoring all rows')
1308 2429 aaronmk
        limit_ref[0] = 0 # just create an empty pkeys table
1309 2735 aaronmk
1310 2758 aaronmk
    def ignore(in_col, value, e):
1311
        track_data_error(db, errors_table_, in_col.srcs, value, e.cause.pgcode,
1312
            e.cause.pgerror)
1313 2895 aaronmk
        log_debug('Ignoring rows with '+strings.as_tt(repr(in_col))+' = '
1314 2545 aaronmk
            +strings.as_tt(repr(value)))
1315 2895 aaronmk
1316 2758 aaronmk
    def remove_rows(in_col, value, e):
1317
        ignore(in_col, value, e)
1318 2378 aaronmk
        cond = (in_col, sql_gen.CompareCond(value, '!='))
1319
        assert cond not in conds # avoid infinite loops
1320 2380 aaronmk
        conds.add(cond)
1321 2895 aaronmk
1322 2758 aaronmk
    def invalid2null(in_col, value, e):
1323
        ignore(in_col, value, e)
1324 2403 aaronmk
        update(db, in_table, [(in_col, None)],
1325
            sql_gen.ColValueCond(in_col, value))
1326 2245 aaronmk
1327 2589 aaronmk
    def insert_pkeys_table(which):
1328 2932 aaronmk
        return sql_gen.Table(sql_gen.concat(in_table.name,
1329 2589 aaronmk
            '_insert_'+which+'_pkeys'))
1330
    insert_out_pkeys = insert_pkeys_table('out')
1331
    insert_in_pkeys = insert_pkeys_table('in')
1332
1333 2206 aaronmk
    # Do inserts and selects
1334 2642 aaronmk
    join_cols = sql_gen.ColDict(db, out_table)
1335 2206 aaronmk
    while True:
1336 2521 aaronmk
        if limit_ref[0] == 0: # special case
1337
            log_debug('Creating an empty pkeys table')
1338 2786 aaronmk
            cur = run_query_into(db, mk_select(db, out_table, [out_pkey],
1339 2521 aaronmk
                limit=limit_ref[0]), into=insert_out_pkeys)
1340
            break # don't do main case
1341
1342 2303 aaronmk
        has_joins = join_cols != {}
1343
1344 2903 aaronmk
        log_debug('Trying to insert new rows')
1345
1346 2305 aaronmk
        # Prepare to insert new rows
1347 2325 aaronmk
        insert_joins = input_joins[:] # don't modify original!
1348 2403 aaronmk
        insert_args = dict(recover=True, cacheable=False)
1349 2303 aaronmk
        if has_joins:
1350 2325 aaronmk
            insert_joins.append(sql_gen.Join(out_table, join_cols,
1351
                sql_gen.filter_out))
1352
        else:
1353 2404 aaronmk
            insert_args.update(dict(returning=out_pkey, into=insert_out_pkeys))
1354 2786 aaronmk
        main_select = mk_main_select(insert_joins, mapping.values())
1355 2303 aaronmk
1356 2929 aaronmk
        def main_insert():
1357 2942 aaronmk
            if is_function:
1358
                log_debug('Calling function on input rows')
1359
                args = dict(((k.name, v) for k, v in mapping.iteritems()))
1360
                func_call = sql_gen.NamedCol(out_pkey,
1361
                    sql_gen.FunctionCall(out_table, **args))
1362
                insert_into_pkeys(input_joins, [in_pkey_col, func_call])
1363
                return None
1364
            else:
1365
                lock_table(db, out_table, 'EXCLUSIVE')
1366
                return insert_select(db, out_table, mapping.keys(), main_select,
1367
                    **insert_args)
1368 2929 aaronmk
1369 2206 aaronmk
        try:
1370 2929 aaronmk
            cur = with_savepoint(db, main_insert)
1371 2357 aaronmk
            break # insert successful
1372 2931 aaronmk
        except MissingCastException, e:
1373
            log_exc(e)
1374
1375
            out_col = e.col
1376
            type_ = e.type
1377
1378
            log_debug('Casting '+strings.as_tt(out_col)+' input to '
1379
                +strings.as_tt(type_))
1380 2937 aaronmk
            mapping[out_col] = cast_temp_col(db, type_, mapping[out_col],
1381
                errors_table_)
1382 2206 aaronmk
        except DuplicateKeyException, e:
1383 2309 aaronmk
            log_exc(e)
1384
1385 2258 aaronmk
            old_join_cols = join_cols.copy()
1386 2811 aaronmk
            distinct_on.update(util.dict_subset(mapping, e.cols))
1387 2565 aaronmk
            join_cols.update(util.dict_subset_right_join(mapping, e.cols))
1388 2486 aaronmk
            log_debug('Ignoring existing rows, comparing on these columns:\n'
1389 2505 aaronmk
                +strings.as_inline_table(join_cols, ustr=col_ustr))
1390 2258 aaronmk
            assert join_cols != old_join_cols # avoid infinite loops
1391 2230 aaronmk
        except NullValueException, e:
1392 2309 aaronmk
            log_exc(e)
1393
1394 2230 aaronmk
            out_col, = e.cols
1395
            try: in_col = mapping[out_col]
1396 2356 aaronmk
            except KeyError:
1397 2486 aaronmk
                log_debug('Missing mapping for NOT NULL column '+out_col)
1398 2451 aaronmk
                remove_all_rows()
1399 2758 aaronmk
            else: remove_rows(in_col, None, e)
1400 2542 aaronmk
        except FunctionValueException, e:
1401
            log_exc(e)
1402
1403
            func_name = e.name
1404
            value = e.value
1405
            for out_col, in_col in mapping.iteritems():
1406 2758 aaronmk
                in_col = sql_gen.unwrap_func_call(in_col, func_name)
1407
                invalid2null(in_col, value, e)
1408 2429 aaronmk
        except DatabaseErrors, e:
1409
            log_exc(e)
1410
1411 2808 aaronmk
            log_debug('No handler for exception')
1412
            on_error(e)
1413 2451 aaronmk
            remove_all_rows()
1414 2358 aaronmk
        # after exception handled, rerun loop with additional constraints
1415 2132 aaronmk
1416 2942 aaronmk
    if cur != None and row_ct_ref != None and cur.rowcount >= 0:
1417 2357 aaronmk
        row_ct_ref[0] += cur.rowcount
1418
1419 2942 aaronmk
    if is_function: pass # pkeys table already created
1420
    elif has_joins:
1421 2357 aaronmk
        select_joins = input_joins+[sql_gen.Join(out_table, join_cols)]
1422 2486 aaronmk
        log_debug('Getting output table pkeys of existing/inserted rows')
1423 2960 aaronmk
        insert_into_pkeys(select_joins, pkeys_cols, distinct=True)
1424 2357 aaronmk
    else:
1425 2404 aaronmk
        add_row_num(db, insert_out_pkeys) # for joining with input pkeys
1426 2357 aaronmk
1427 2486 aaronmk
        log_debug('Getting input table pkeys of inserted rows')
1428 2786 aaronmk
        run_query_into(db, mk_main_select(input_joins, [in_pkey]),
1429 2404 aaronmk
            into=insert_in_pkeys)
1430
        add_row_num(db, insert_in_pkeys) # for joining with output pkeys
1431 2357 aaronmk
1432 2428 aaronmk
        assert table_row_count(db, insert_out_pkeys) == table_row_count(db,
1433
            insert_in_pkeys)
1434
1435 2486 aaronmk
        log_debug('Combining output and input pkeys in inserted order')
1436 2404 aaronmk
        pkey_joins = [insert_in_pkeys, sql_gen.Join(insert_out_pkeys,
1437 2357 aaronmk
            {row_num_col: sql_gen.join_same_not_null})]
1438 2420 aaronmk
        insert_into_pkeys(pkey_joins, pkeys_names)
1439 2959 aaronmk
1440 2974 aaronmk
        empty_temp(db, [insert_out_pkeys, insert_in_pkeys])
1441 2357 aaronmk
1442 2486 aaronmk
    db.log_debug('Adding pkey on pkeys table to enable fast joins', level=2.5)
1443 2594 aaronmk
    add_pkey(db, into)
1444 2407 aaronmk
1445 2508 aaronmk
    log_debug('Setting pkeys of missing rows to '+strings.as_tt(repr(default)))
1446 2489 aaronmk
    missing_rows_joins = input_joins+[sql_gen.Join(into,
1447 2357 aaronmk
        {in_pkey: sql_gen.join_same_not_null}, sql_gen.filter_out)]
1448
        # must use join_same_not_null or query will take forever
1449 2420 aaronmk
    insert_into_pkeys(missing_rows_joins,
1450 2508 aaronmk
        [in_pkey_col, sql_gen.NamedCol(out_pkey, default)])
1451 2357 aaronmk
1452 2489 aaronmk
    assert table_row_count(db, into) == table_row_count(db, in_table)
1453 2428 aaronmk
1454 2974 aaronmk
    empty_temp(db, in_table)
1455 2959 aaronmk
1456 2702 aaronmk
    srcs = []
1457
    if is_func: srcs = sql_gen.cols_srcs(in_cols)
1458
    return sql_gen.Col(out_pkey, into, srcs)
1459 2115 aaronmk
1460
##### Data cleanup
1461
1462 2290 aaronmk
def cleanup_table(db, table, cols):
1463 2677 aaronmk
    table = sql_gen.as_Table(table)
1464
    cols = map(sql_gen.as_Col, cols)
1465 2115 aaronmk
1466 2677 aaronmk
    expr = ('nullif(nullif(trim(both from %s), '+db.esc_value('')+'), '
1467
        +db.esc_value(r'\N')+')')
1468
    changes = [(v, sql_gen.CustomCode(expr % v.to_str(db)))
1469
        for v in cols]
1470 2115 aaronmk
1471 2677 aaronmk
    update(db, table, changes)