Project

General

Profile

1
# SQL code generation
2

    
3
import operator
4
import re
5

    
6
import objects
7
import strings
8
import util
9

    
10
##### Names
11

    
12
identifier_max_len = 63 # works for both PostgreSQL and MySQL
13

    
14
def add_suffix(str_, suffix):
15
    '''Preserves version so that it won't be truncated off the string, leading
16
    to collisions.'''
17
    # Preserve version
18
    before, sep, version = str_.rpartition('#')
19
    if sep != '': # found match
20
        str_ = before
21
        suffix = sep+version+suffix
22
    
23
    return strings.add_suffix(str_, suffix, identifier_max_len)
24

    
25
def is_safe_name(name):
26
    '''A name is safe *and unambiguous* if it:
27
    * contains only *lowercase* word (\w) characters
28
    * doesn't start with a digit
29
    * contains "_", so that it's not a keyword
30
    ''' 
31
    return re.match(r'^(?=.*_)(?!\d)[^\WA-Z]+$', name) 
32

    
33
def esc_name(name, quote='"'):
34
    return quote + name.replace(quote, quote+quote) + quote
35
        # doubling an embedded quote escapes it in both PostgreSQL and MySQL
36

    
37
def clean_name(name): return name.replace('"', '').replace('`', '')
38

    
39
##### SQL code objects
40

    
41
class MockDb:
42
    def esc_value(self, value): return strings.repr_no_u(value)
43
    
44
    def esc_name(self, name): return esc_name(name)
45
mockDb = MockDb()
46

    
47
class BasicObject(objects.BasicObject):
48
    def __init__(self, value): self.value = value
49
    
50
    def __str__(self): return clean_name(strings.repr_no_u(self))
51

    
52
class Code(BasicObject):
53
    def to_str(self, db): raise NotImplemented()
54
    
55
    def __repr__(self): return self.to_str(mockDb)
56

    
57
class CustomCode(Code):
58
    def __init__(self, str_): self.str_ = str_
59
    
60
    def to_str(self, db): return self.str_
61

    
62
class Expr(Code):
63
    def __init__(self, expr): self.expr = expr
64
    
65
    def to_str(self, db): return '('+self.expr.to_str(db)+')'
66

    
67
##### Literal values
68

    
69
class Literal(Code):
70
    def __init__(self, value): self.value = value
71
    
72
    def to_str(self, db): return db.esc_value(self.value)
73

    
74
def as_Value(value):
75
    if isinstance(value, Code): return value
76
    else: return Literal(value)
77

    
78
def is_null(value): return isinstance(value, Literal) and value.value == None
79

    
80
##### Tables
81

    
82
class Table(Code):
83
    def __init__(self, name, schema=None):
84
        '''
85
        @param schema str|None (for no schema)
86
        '''
87
        self.name = name
88
        self.schema = schema
89
    
90
    def to_str(self, db):
91
        str_ = ''
92
        if self.schema != None: str_ += db.esc_name(self.schema)+'.'
93
        str_ += db.esc_name(self.name)
94
        return str_
95
    
96
    def to_Table(self): return self
97

    
98
def as_Table(table):
99
    if table == None or isinstance(table, Code): return table
100
    else: return Table(table)
101

    
102
class NamedTable(Table):
103
    def __init__(self, name, code, cols=None):
104
        Table.__init__(self, name)
105
        
106
        if not isinstance(code, Code): code = Table(code)
107
        
108
        self.code = code
109
        self.cols = cols
110
    
111
    def to_str(self, db):
112
        str_ = self.code.to_str(db)+'\nAS '+Table.to_str(self, db)
113
        if self.cols != None: str_ += ' ('+(', '.join(self.cols))+')'
114
        return str_
115
    
116
    def to_Table(self): return Table(self.name)
117

    
118
##### Columns
119

    
120
class Col(Code):
121
    def __init__(self, name, table=None):
122
        '''
123
        @param table Table|None (for no table)
124
        '''
125
        if util.is_str(table): table = Table(table)
126
        assert table == None or isinstance(table, Table)
127
        
128
        self.name = name
129
        self.table = table
130
    
131
    def to_str(self, db):
132
        str_ = ''
133
        if self.table != None: str_ += self.table.to_str(db)+'.'
134
        str_ += db.esc_name(self.name)
135
        return str_
136
    
137
    def to_Col(self): return self
138

    
139
def is_table_col(col): return col.table != None
140

    
141
def as_Col(col, table=None, name=None):
142
    '''
143
    @param name If not None, any non-Col input will be renamed using NamedCol.
144
    '''
145
    if name != None:
146
        col = as_Value(col)
147
        if not isinstance(col, Col): col = NamedCol(name, col)
148
    
149
    if isinstance(col, Code): return col
150
    else: return Col(col, table)
151

    
152
def to_name_only_col(col, check_table=None):
153
    col = as_Col(col)
154
    if not isinstance(col, Col): return col
155
    
156
    if check_table != None:
157
        table = col.table
158
        assert table == None or table == check_table
159
    return Col(col.name)
160

    
161
class NamedCol(Col):
162
    def __init__(self, name, code):
163
        Col.__init__(self, name)
164
        
165
        if not isinstance(code, Code): code = Literal(code)
166
        
167
        self.code = code
168
    
169
    def to_str(self, db):
170
        return self.code.to_str(db)+' AS '+Col.to_str(self, db)
171
    
172
    def to_Col(self): return Col(self.name)
173

    
174
def remove_col_rename(col):
175
    if isinstance(col, NamedCol): col = col.code
176
    return col
177

    
178
class ColDict(dict):
179
    '''A dict that automatically makes inserted entries Col objects'''
180
    
181
    '''For params, see dict()'''
182
    def __init__(self, keys_table, *args, **kw_args):
183
        dict.__init__(self, *args, **kw_args)
184
        
185
        self.table = keys_table
186
    
187
    def __getitem__(self, key):
188
        return dict.__getitem__(self, self._key(key))
189
    
190
    def __setitem__(self, key, value):
191
        return dict.__setitem__(self, self._key(key), as_Col(value, name=key))
192
    
193
    def update(self, dict_):
194
        for key, value in dict_.iteritems(): self[key] = value
195
    
196
    def _key(self, key): return as_Col(key, self.table)
197

    
198
##### Functions
199

    
200
class Function(Table): pass
201

    
202
class FunctionCall(Code):
203
    def __init__(self, function, *args):
204
        '''
205
        @param args [Code...] The function's arguments
206
        '''
207
        if not isinstance(function, Code): function = Function(function)
208
        args = map(remove_col_rename, args)
209
        
210
        self.function = function
211
        self.args = args
212
    
213
    def to_str(self, db):
214
        args_str = ', '.join((v.to_str(db) for v in self.args))
215
        return self.function.to_str(db)+'('+args_str+')'
216

    
217
def wrap_in_func(function, value):
218
    '''Wraps a value inside a function call.
219
    Propagates any column renaming to the returned value.
220
    '''
221
    name = None
222
    if isinstance(value, NamedCol): name = value.name
223
    value = FunctionCall(function, value)
224
    if name != None: value = NamedCol(name, value)
225
    return value
226

    
227
def unwrap_func_call(func_call, check_name=None):
228
    '''Unwraps any function call to its first argument.
229
    Also removes any column renaming.
230
    '''
231
    func_call = remove_col_rename(func_call)
232
    if not isinstance(func_call, FunctionCall): return func_call
233
    
234
    if check_name != None:
235
        name = func_call.function.name
236
        assert name == None or name == check_name
237
    return func_call.args[0]
238

    
239
##### Conditions
240

    
241
class ColValueCond(Code):
242
    def __init__(self, col, value):
243
        value = as_ValueCond(value)
244
        
245
        self.col = col
246
        self.value = value
247
    
248
    def to_str(self, db): return self.value.to_str(db, self.col)
249

    
250
def combine_conds(conds, keyword=None):
251
    '''
252
    @param keyword The keyword to add before the conditions, if any
253
    '''
254
    str_ = ''
255
    if keyword != None:
256
        if conds == []: whitespace = ''
257
        elif len(conds) == 1: whitespace = ' '
258
        else: whitespace = '\n'
259
        str_ += keyword+whitespace
260
    
261
    str_ += '\nAND '.join(conds)
262
    return str_
263

    
264
##### Condition column comparisons
265

    
266
class ValueCond(BasicObject):
267
    def __init__(self, value):
268
        if not isinstance(value, Code): value = Literal(value)
269
        value = remove_col_rename(value)
270
        
271
        self.value = value
272
    
273
    def to_str(self, db, left_value):
274
        '''
275
        @param left_value The Code object that the condition is being applied on
276
        '''
277
        raise NotImplemented()
278
    
279
    def __repr__(self): return self.to_str(mockDb, '<left_value>')
280

    
281
class CompareCond(ValueCond):
282
    def __init__(self, value, operator='='):
283
        '''
284
        @param operator By default, compares NULL values literally. Use '~=' or
285
            '~!=' to pass NULLs through.
286
        '''
287
        ValueCond.__init__(self, value)
288
        self.operator = operator
289
    
290
    def to_str(self, db, left_value):
291
        if not isinstance(left_value, Code): left_value = Col(left_value)
292
        left_value = remove_col_rename(left_value)
293
        
294
        right_value = self.value
295
        left = left_value.to_str(db)
296
        right = right_value.to_str(db)
297
        
298
        # Parse operator
299
        operator = self.operator
300
        passthru_null_ref = [False]
301
        operator = strings.remove_prefix('~', operator, passthru_null_ref)
302
        neg_ref = [False]
303
        operator = strings.remove_prefix('!', operator, neg_ref)
304
        equals = operator.endswith('=')
305
        if equals and is_null(self.value): operator = 'IS'
306
        
307
        # Create str
308
        str_ = left+' '+operator+' '+right
309
        if equals and not passthru_null_ref[0] and isinstance(right_value, Col):
310
            str_ = '('+str_+' OR ('+left+' IS NULL AND '+right+' IS NULL))'
311
        if neg_ref[0]: str_ = 'NOT '+str_
312
        return str_
313

    
314
# Tells as_ValueCond() to assume a non-ValueCond is a literal value
315
assume_literal = object()
316

    
317
def as_ValueCond(value, default_table=assume_literal):
318
    if not isinstance(value, ValueCond):
319
        if default_table is not assume_literal:
320
            value = as_Col(value, default_table)
321
        return CompareCond(value)
322
    else: return value
323

    
324
##### Joins
325

    
326
join_same = object() # tells Join the left and right columns have the same name
327

    
328
# Tells Join the left and right columns have the same name and are never NULL
329
join_same_not_null = object()
330

    
331
filter_out = object() # tells Join to filter out rows that match the join
332

    
333
class Join(BasicObject):
334
    def __init__(self, table, mapping, type_=None):
335
        '''
336
        @param mapping dict(right_table_col=left_table_col, ...)
337
            * if left_table_col is join_same: left_table_col = right_table_col
338
              * Note that right_table_col must be a string
339
            * if left_table_col is join_same_not_null:
340
              left_table_col = right_table_col and both have NOT NULL constraint
341
              * Note that right_table_col must be a string
342
        @param type_ None (for plain join)|str (e.g. 'LEFT')|filter_out
343
            * filter_out: equivalent to 'LEFT' with the query filtered by
344
              `table_pkey IS NULL` (indicating no match)
345
        '''
346
        if util.is_str(table): table = Table(table)
347
        assert type_ == None or util.is_str(type_) or type_ is filter_out
348
        
349
        self.table = table
350
        self.mapping = mapping
351
        self.type_ = type_
352
    
353
    def to_str(self, db, left_table):
354
        # Switch order (left_table is on the right in the comparison)
355
        right_table = left_table
356
        left_table = self.table # note left_table is reassigned
357
        
358
        def join(entry):
359
            '''Parses non-USING joins'''
360
            right_table_col, left_table_col = entry
361
            
362
            # Switch order (right_table_col is on the left in the comparison)
363
            left = right_table_col
364
            right = left_table_col
365
            
366
            # Parse special values
367
            if right is join_same: right = left
368
            elif right is join_same_not_null:
369
                right = CompareCond(as_Col(left, right_table), '~=')
370
            
371
            right = as_ValueCond(right, right_table)
372
            return right.to_str(db, as_Col(left, left_table))
373
        
374
        # Create join condition
375
        type_ = self.type_
376
        joins = self.mapping
377
        if type_ is not filter_out and reduce(operator.and_,
378
            (v is join_same_not_null for v in joins.itervalues())):
379
            # all cols w/ USING, so can use simpler USING syntax
380
            cols = (as_Col(v).to_str(db) for v in joins.iterkeys())
381
            join_cond = 'USING ('+(', '.join(cols))+')'
382
        else:
383
            if len(joins) == 1: whitespace = ' '
384
            else: whitespace = '\n'
385
            join_cond = combine_conds(map(join, joins.iteritems()), 'ON')
386
        
387
        # Create join
388
        if type_ is filter_out: type_ = 'LEFT'
389
        str_ = ''
390
        if type_ != None: str_ += type_+' '
391
        str_ += 'JOIN '+left_table.to_str(db)+' '+join_cond
392
        return str_
393
    
394
    def __repr__(self): return self.to_str(mockDb, '<left_table>')
395

    
396
##### Value exprs
397

    
398
row_count = CustomCode('count(*)')
(25-25/36)