Parsing.py 126 KB
Newer Older
1
# cython: auto_cpdef=True, infer_types=True, language_level=3, py2_import=True
William Stein's avatar
William Stein committed
2
#
3
#   Parser
William Stein's avatar
William Stein committed
4 5
#

6 7
from __future__ import absolute_import

8 9
# This should be done automatically
import cython
10
cython.declare(Nodes=object, ExprNodes=object, EncodedString=object,
11
               bytes_literal=object, StringEncoding=object,
12
               FileSourceDescriptor=object, lookup_unicodechar=object, unicode_category=object,
13
               Future=object, Options=object, error=object, warning=object,
14 15
               Builtin=object, ModuleNode=object, Utils=object, _unicode=object, _bytes=object,
               re=object, sys=object, _parse_escape_sequences=object, _parse_escape_sequences_raw=object,
16 17
               partial=object, reduce=object, _IS_PY3=cython.bint, _IS_2BYTE_UNICODE=cython.bint,
               _CDEF_MODIFIERS=tuple)
18

19
from io import StringIO
20
import re
21
import sys
22
from unicodedata import lookup as lookup_unicodechar, category as unicode_category
23
from functools import partial, reduce
Lisandro Dalcin's avatar
Lisandro Dalcin committed
24

25
from .Scanning import PyrexScanner, FileSourceDescriptor, StringSourceDescriptor
26 27 28 29
from . import Nodes
from . import ExprNodes
from . import Builtin
from . import StringEncoding
30
from .StringEncoding import EncodedString, bytes_literal, _unicode, _bytes
31 32 33 34 35 36
from .ModuleNode import ModuleNode
from .Errors import error, warning
from .. import Utils
from . import Future
from . import Options

37
_IS_PY3 = sys.version_info[0] >= 3
38
_IS_2BYTE_UNICODE = sys.maxunicode == 0xffff
39
_CDEF_MODIFIERS = ('inline', 'nogil', 'api')
40

William Stein's avatar
William Stein committed
41

42 43 44 45 46 47 48 49 50
class Ctx(object):
    #  Parsing context
    level = 'other'
    visibility = 'private'
    cdef_flag = 0
    typedef_flag = 0
    api = 0
    overridable = 0
    nogil = 0
51
    namespace = None
Danilo Freitas's avatar
Danilo Freitas committed
52
    templates = None
53
    allow_struct_enum_decorator = False
54 55 56 57 58 59 60 61 62 63 64

    def __init__(self, **kwds):
        self.__dict__.update(kwds)

    def __call__(self, **kwds):
        ctx = Ctx()
        d = ctx.__dict__
        d.update(self.__dict__)
        d.update(kwds)
        return ctx

65

66
def p_ident(s, message="Expected an identifier"):
William Stein's avatar
William Stein committed
67 68 69 70 71 72 73 74 75 76 77 78
    if s.sy == 'IDENT':
        name = s.systring
        s.next()
        return name
    else:
        s.error(message)

def p_ident_list(s):
    names = []
    while s.sy == 'IDENT':
        names.append(s.systring)
        s.next()
Stefan Behnel's avatar
Stefan Behnel committed
79
        if s.sy != ',':
William Stein's avatar
William Stein committed
80 81 82 83 84 85 86 87 88 89
            break
        s.next()
    return names

#------------------------------------------
#
#   Expressions
#
#------------------------------------------

90 91 92 93 94 95
def p_binop_operator(s):
    pos = s.position()
    op = s.sy
    s.next()
    return op, pos

William Stein's avatar
William Stein committed
96 97 98
def p_binop_expr(s, ops, p_sub_expr):
    n1 = p_sub_expr(s)
    while s.sy in ops:
99
        op, pos = p_binop_operator(s)
William Stein's avatar
William Stein committed
100 101
        n2 = p_sub_expr(s)
        n1 = ExprNodes.binop_node(pos, op, n1, n2)
102 103 104 105 106
        if op == '/':
            if Future.division in s.context.future_directives:
                n1.truedivision = True
            else:
                n1.truedivision = None # unknown
William Stein's avatar
William Stein committed
107 108
    return n1

Stefan Behnel's avatar
Stefan Behnel committed
109 110 111 112 113 114 115 116 117 118
#lambdef: 'lambda' [varargslist] ':' test

def p_lambdef(s, allow_conditional=True):
    # s.sy == 'lambda'
    pos = s.position()
    s.next()
    if s.sy == ':':
        args = []
        star_arg = starstar_arg = None
    else:
119 120
        args, star_arg, starstar_arg = p_varargslist(
            s, terminator=':', annotated=False)
Stefan Behnel's avatar
Stefan Behnel committed
121 122
    s.expect(':')
    if allow_conditional:
123
        expr = p_test(s)
Stefan Behnel's avatar
Stefan Behnel committed
124 125 126 127 128 129 130 131 132 133 134 135
    else:
        expr = p_test_nocond(s)
    return ExprNodes.LambdaNode(
        pos, args = args,
        star_arg = star_arg, starstar_arg = starstar_arg,
        result_expr = expr)

#lambdef_nocond: 'lambda' [varargslist] ':' test_nocond

def p_lambdef_nocond(s):
    return p_lambdef(s, allow_conditional=False)

136
#test: or_test ['if' or_test 'else' test] | lambdef
William Stein's avatar
William Stein committed
137

Robert Bradshaw's avatar
Robert Bradshaw committed
138
def p_test(s):
139 140
    if s.sy == 'lambda':
        return p_lambdef(s)
Robert Bradshaw's avatar
Robert Bradshaw committed
141 142 143 144 145
    pos = s.position()
    expr = p_or_test(s)
    if s.sy == 'if':
        s.next()
        test = p_or_test(s)
Stefan Behnel's avatar
Stefan Behnel committed
146 147 148
        s.expect('else')
        other = p_test(s)
        return ExprNodes.CondExprNode(pos, test=test, true_val=expr, false_val=other)
Robert Bradshaw's avatar
Robert Bradshaw committed
149 150 151
    else:
        return expr

Stefan Behnel's avatar
Stefan Behnel committed
152 153 154 155 156 157 158
#test_nocond: or_test | lambdef_nocond

def p_test_nocond(s):
    if s.sy == 'lambda':
        return p_lambdef_nocond(s)
    else:
        return p_or_test(s)
Robert Bradshaw's avatar
Robert Bradshaw committed
159 160 161 162

#or_test: and_test ('or' and_test)*

def p_or_test(s):
William Stein's avatar
William Stein committed
163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
    return p_rassoc_binop_expr(s, ('or',), p_and_test)

def p_rassoc_binop_expr(s, ops, p_subexpr):
    n1 = p_subexpr(s)
    if s.sy in ops:
        pos = s.position()
        op = s.sy
        s.next()
        n2 = p_rassoc_binop_expr(s, ops, p_subexpr)
        n1 = ExprNodes.binop_node(pos, op, n1, n2)
    return n1

#and_test: not_test ('and' not_test)*

def p_and_test(s):
    #return p_binop_expr(s, ('and',), p_not_test)
    return p_rassoc_binop_expr(s, ('and',), p_not_test)

#not_test: 'not' not_test | comparison

def p_not_test(s):
    if s.sy == 'not':
        pos = s.position()
        s.next()
        return ExprNodes.NotNode(pos, operand = p_not_test(s))
    else:
        return p_comparison(s)

#comparison: expr (comp_op expr)*
#comp_op: '<'|'>'|'=='|'>='|'<='|'<>'|'!='|'in'|'not' 'in'|'is'|'is' 'not'

def p_comparison(s):
195
    n1 = p_starred_expr(s)
William Stein's avatar
William Stein committed
196 197 198
    if s.sy in comparison_ops:
        pos = s.position()
        op = p_cmp_op(s)
199
        n2 = p_starred_expr(s)
200
        n1 = ExprNodes.PrimaryCmpNode(pos,
William Stein's avatar
William Stein committed
201 202 203 204 205
            operator = op, operand1 = n1, operand2 = n2)
        if s.sy in comparison_ops:
            n1.cascade = p_cascaded_cmp(s)
    return n1

206 207 208 209 210 211
def p_test_or_starred_expr(s):
    if s.sy == '*':
        return p_starred_expr(s)
    else:
        return p_test(s)

212
def p_starred_expr(s):
213
    pos = s.position()
214 215 216 217 218 219
    if s.sy == '*':
        starred = True
        s.next()
    else:
        starred = False
    expr = p_bit_expr(s)
220
    if starred:
221
        expr = ExprNodes.StarredUnpackingNode(pos, expr)
222 223
    return expr

William Stein's avatar
William Stein committed
224 225 226
def p_cascaded_cmp(s):
    pos = s.position()
    op = p_cmp_op(s)
227
    n2 = p_starred_expr(s)
228
    result = ExprNodes.CascadedCmpNode(pos,
William Stein's avatar
William Stein committed
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251
        operator = op, operand2 = n2)
    if s.sy in comparison_ops:
        result.cascade = p_cascaded_cmp(s)
    return result

def p_cmp_op(s):
    if s.sy == 'not':
        s.next()
        s.expect('in')
        op = 'not_in'
    elif s.sy == 'is':
        s.next()
        if s.sy == 'not':
            s.next()
            op = 'is_not'
        else:
            op = 'is'
    else:
        op = s.sy
        s.next()
    if op == '<>':
        op = '!='
    return op
252

253
comparison_ops = cython.declare(set, set([
254
    '<', '>', '==', '>=', '<=', '<>', '!=',
William Stein's avatar
William Stein committed
255
    'in', 'is', 'not'
256
]))
William Stein's avatar
William Stein committed
257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282

#expr: xor_expr ('|' xor_expr)*

def p_bit_expr(s):
    return p_binop_expr(s, ('|',), p_xor_expr)

#xor_expr: and_expr ('^' and_expr)*

def p_xor_expr(s):
    return p_binop_expr(s, ('^',), p_and_expr)

#and_expr: shift_expr ('&' shift_expr)*

def p_and_expr(s):
    return p_binop_expr(s, ('&',), p_shift_expr)

#shift_expr: arith_expr (('<<'|'>>') arith_expr)*

def p_shift_expr(s):
    return p_binop_expr(s, ('<<', '>>'), p_arith_expr)

#arith_expr: term (('+'|'-') term)*

def p_arith_expr(s):
    return p_binop_expr(s, ('+', '-'), p_term)

283
#term: factor (('*'|'@'|'/'|'%'|'//') factor)*
William Stein's avatar
William Stein committed
284 285

def p_term(s):
286
    return p_binop_expr(s, ('*', '@', '/', '%', '//'), p_factor)
William Stein's avatar
William Stein committed
287 288 289 290

#factor: ('+'|'-'|'~'|'&'|typecast|sizeof) factor | power

def p_factor(s):
291 292 293 294
    # little indirection for C-ification purposes
    return _p_factor(s)

def _p_factor(s):
William Stein's avatar
William Stein committed
295 296 297 298 299 300
    sy = s.sy
    if sy in ('+', '-', '~'):
        op = s.sy
        pos = s.position()
        s.next()
        return ExprNodes.unop_node(pos, op, p_factor(s))
301 302 303 304 305 306 307 308 309 310 311
    elif not s.in_python_file:
        if sy == '&':
            pos = s.position()
            s.next()
            arg = p_factor(s)
            return ExprNodes.AmpersandNode(pos, operand = arg)
        elif sy == "<":
            return p_typecast(s)
        elif sy == 'IDENT' and s.systring == "sizeof":
            return p_sizeof(s)
    return p_power(s)
William Stein's avatar
William Stein committed
312 313 314 315 316 317

def p_typecast(s):
    # s.sy == "<"
    pos = s.position()
    s.next()
    base_type = p_c_base_type(s)
318
    is_memslice = isinstance(base_type, Nodes.MemoryViewSliceTypeNode)
Robert Bradshaw's avatar
Robert Bradshaw committed
319 320 321 322
    is_template = isinstance(base_type, Nodes.TemplatedTypeNode)
    is_const = isinstance(base_type, Nodes.CConstTypeNode)
    if (not is_memslice and not is_template and not is_const
        and base_type.name is None):
323
        s.error("Unknown type")
William Stein's avatar
William Stein committed
324
    declarator = p_c_declarator(s, empty = 1)
325 326 327 328 329
    if s.sy == '?':
        s.next()
        typecheck = 1
    else:
        typecheck = 0
William Stein's avatar
William Stein committed
330 331
    s.expect(">")
    operand = p_factor(s)
332 333 334 335
    if is_memslice:
        return ExprNodes.CythonArrayNode(pos, base_type_node=base_type,
                                         operand=operand)

336 337
    return ExprNodes.TypecastNode(pos,
        base_type = base_type,
William Stein's avatar
William Stein committed
338
        declarator = declarator,
339 340
        operand = operand,
        typecheck = typecheck)
William Stein's avatar
William Stein committed
341 342 343 344 345 346

def p_sizeof(s):
    # s.sy == ident "sizeof"
    pos = s.position()
    s.next()
    s.expect('(')
347
    # Here we decide if we are looking at an expression or type
348 349
    # If it is actually a type, but parsable as an expression,
    # we treat it as an expression here.
350
    if looking_at_expr(s):
351
        operand = p_test(s)
352 353
        node = ExprNodes.SizeofVarNode(pos, operand = operand)
    else:
William Stein's avatar
William Stein committed
354 355
        base_type = p_c_base_type(s)
        declarator = p_c_declarator(s, empty = 1)
356
        node = ExprNodes.SizeofTypeNode(pos,
William Stein's avatar
William Stein committed
357 358 359 360
            base_type = base_type, declarator = declarator)
    s.expect(')')
    return node

361

362 363 364 365
def p_yield_expression(s):
    # s.sy == "yield"
    pos = s.position()
    s.next()
366 367 368 369
    is_yield_from = False
    if s.sy == 'from':
        is_yield_from = True
        s.next()
370
    if s.sy != ')' and s.sy not in statement_terminators:
371 372
        # "yield from" does not support implicit tuples, but "yield" does ("yield 1,2")
        arg = p_test(s) if is_yield_from else p_testlist(s)
373
    else:
374
        if is_yield_from:
375 376
            s.error("'yield from' requires a source argument",
                    pos=pos, fatal=False)
377
        arg = None
378 379 380 381
    if is_yield_from:
        return ExprNodes.YieldFromExprNode(pos, arg=arg)
    else:
        return ExprNodes.YieldExprNode(pos, arg=arg)
382

383

384 385 386 387
def p_yield_statement(s):
    # s.sy == "yield"
    yield_expr = p_yield_expression(s)
    return Nodes.ExprStatNode(yield_expr.pos, expr=yield_expr)
388

389 390 391 392 393 394 395 396 397 398 399 400

def p_async_statement(s, ctx, decorators):
    # s.sy >> 'async' ...
    if s.sy == 'def':
        # 'async def' statements aren't allowed in pxd files
        if 'pxd' in ctx.level:
            s.error('def statement not allowed here')
        s.level = ctx.level
        return p_def_statement(s, decorators, is_async_def=True)
    elif decorators:
        s.error("Decorators can only be followed by functions or classes")
    elif s.sy == 'for':
401
        return p_for_statement(s, is_async=True)
402
    elif s.sy == 'with':
403 404
        s.next()
        return p_with_items(s, is_async=True)
405 406 407 408 409 410
    else:
        s.error("expected one of 'def', 'for', 'with' after 'async'")


#power: atom_expr ('**' factor)*
#atom_expr: ['await'] atom trailer*
William Stein's avatar
William Stein committed
411 412

def p_power(s):
413
    if s.systring == 'new' and s.peek()[0] == 'IDENT':
Danilo Freitas's avatar
Danilo Freitas committed
414
        return p_new_expr(s)
415 416 417 418
    await_pos = None
    if s.sy == 'await':
        await_pos = s.position()
        s.next()
William Stein's avatar
William Stein committed
419 420 421
    n1 = p_atom(s)
    while s.sy in ('(', '[', '.'):
        n1 = p_trailer(s, n1)
422 423
    if await_pos:
        n1 = ExprNodes.AwaitExprNode(await_pos, arg=n1)
William Stein's avatar
William Stein committed
424 425 426 427 428 429 430
    if s.sy == '**':
        pos = s.position()
        s.next()
        n2 = p_factor(s)
        n1 = ExprNodes.binop_node(pos, '**', n1, n2)
    return n1

431

Danilo Freitas's avatar
Danilo Freitas committed
432
def p_new_expr(s):
Danilo Freitas's avatar
Danilo Freitas committed
433
    # s.systring == 'new'.
Danilo Freitas's avatar
Danilo Freitas committed
434 435
    pos = s.position()
    s.next()
Robert Bradshaw's avatar
Robert Bradshaw committed
436 437
    cppclass = p_c_base_type(s)
    return p_call(s, ExprNodes.NewExprNode(pos, cppclass = cppclass))
Danilo Freitas's avatar
Danilo Freitas committed
438

William Stein's avatar
William Stein committed
439 440 441 442 443 444 445 446 447 448
#trailer: '(' [arglist] ')' | '[' subscriptlist ']' | '.' NAME

def p_trailer(s, node1):
    pos = s.position()
    if s.sy == '(':
        return p_call(s, node1)
    elif s.sy == '[':
        return p_index(s, node1)
    else: # s.sy == '.'
        s.next()
449
        name = p_ident(s)
450
        return ExprNodes.AttributeNode(pos,
451
            obj=node1, attribute=name)
William Stein's avatar
William Stein committed
452

453

William Stein's avatar
William Stein committed
454 455 456
# arglist:  argument (',' argument)* [',']
# argument: [test '='] test       # Really [keyword '='] test

457 458 459 460 461 462 463
# since PEP 448:
# argument: ( test [comp_for] |
#             test '=' test |
#             '**' expr |
#             star_expr )

def p_call_parse_args(s, allow_genexp=True):
William Stein's avatar
William Stein committed
464 465 466 467 468
    # s.sy == '('
    pos = s.position()
    s.next()
    positional_args = []
    keyword_args = []
469 470 471
    starstar_seen = False
    last_was_tuple_unpack = False
    while s.sy != ')':
472
        if s.sy == '*':
473 474 475 476 477 478
            if starstar_seen:
                s.error("Non-keyword arg following keyword arg", pos=s.position())
            s.next()
            positional_args.append(p_test(s))
            last_was_tuple_unpack = True
        elif s.sy == '**':
William Stein's avatar
William Stein committed
479
            s.next()
480 481
            keyword_args.append(p_test(s))
            starstar_seen = True
William Stein's avatar
William Stein committed
482
        else:
483
            arg = p_test(s)
484 485 486 487
            if s.sy == '=':
                s.next()
                if not arg.is_name:
                    s.error("Expected an identifier before '='",
488
                            pos=arg.pos)
489
                encoded_name = s.context.intern_ustring(arg.name)
490 491
                keyword = ExprNodes.IdentifierStringNode(
                    arg.pos, value=encoded_name)
492
                arg = p_test(s)
493 494 495
                keyword_args.append((keyword, arg))
            else:
                if keyword_args:
496 497 498 499 500 501
                    s.error("Non-keyword arg following keyword arg", pos=arg.pos)
                if positional_args and not last_was_tuple_unpack:
                    positional_args[-1].append(arg)
                else:
                    positional_args.append([arg])
                last_was_tuple_unpack = False
Stefan Behnel's avatar
Stefan Behnel committed
502
        if s.sy != ',':
William Stein's avatar
William Stein committed
503 504
            break
        s.next()
505

506
    if s.sy in ('for', 'async'):
507 508
        if not keyword_args and not last_was_tuple_unpack:
            if len(positional_args) == 1 and len(positional_args[0]) == 1:
509
                positional_args = [[p_genexp(s, positional_args[0][0])]]
William Stein's avatar
William Stein committed
510
    s.expect(')')
511
    return positional_args or [[]], keyword_args
512

513 514

def p_call_build_packed_args(pos, positional_args, keyword_args):
515
    keyword_dict = None
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550

    subtuples = [
        ExprNodes.TupleNode(pos, args=arg) if isinstance(arg, list) else ExprNodes.AsTupleNode(pos, arg=arg)
        for arg in positional_args
    ]
    # TODO: implement a faster way to join tuples than creating each one and adding them
    arg_tuple = reduce(partial(ExprNodes.binop_node, pos, '+'), subtuples)

    if keyword_args:
        kwargs = []
        dict_items = []
        for item in keyword_args:
            if isinstance(item, tuple):
                key, value = item
                dict_items.append(ExprNodes.DictItemNode(pos=key.pos, key=key, value=value))
            elif item.is_dict_literal:
                # unpack "**{a:b}" directly
                dict_items.extend(item.key_value_pairs)
            else:
                if dict_items:
                    kwargs.append(ExprNodes.DictNode(
                        dict_items[0].pos, key_value_pairs=dict_items, reject_duplicates=True))
                    dict_items = []
                kwargs.append(item)

        if dict_items:
            kwargs.append(ExprNodes.DictNode(
                dict_items[0].pos, key_value_pairs=dict_items, reject_duplicates=True))

        if kwargs:
            if len(kwargs) == 1 and kwargs[0].is_dict_literal:
                # only simple keyword arguments found -> one dict
                keyword_dict = kwargs[0]
            else:
                # at least one **kwargs
551
                keyword_dict = ExprNodes.MergedDictNode(pos, keyword_args=kwargs)
552

553 554
    return arg_tuple, keyword_dict

555

556 557 558
def p_call(s, function):
    # s.sy == '('
    pos = s.position()
559
    positional_args, keyword_args = p_call_parse_args(s)
560

561 562
    if not keyword_args and len(positional_args) == 1 and isinstance(positional_args[0], list):
        return ExprNodes.SimpleCallNode(pos, function=function, args=positional_args[0])
William Stein's avatar
William Stein committed
563
    else:
564 565 566 567
        arg_tuple, keyword_dict = p_call_build_packed_args(pos, positional_args, keyword_args)
        return ExprNodes.GeneralCallNode(
            pos, function=function, positional_args=arg_tuple, keyword_args=keyword_dict)

William Stein's avatar
William Stein committed
568 569 570 571 572 573 574 575 576

#lambdef: 'lambda' [varargslist] ':' test

#subscriptlist: subscript (',' subscript)* [',']

def p_index(s, base):
    # s.sy == '['
    pos = s.position()
    s.next()
577 578
    subscripts, is_single_value = p_subscript_list(s)
    if is_single_value and len(subscripts[0]) == 2:
William Stein's avatar
William Stein committed
579
        start, stop = subscripts[0]
580
        result = ExprNodes.SliceIndexNode(pos,
William Stein's avatar
William Stein committed
581 582 583
            base = base, start = start, stop = stop)
    else:
        indexes = make_slice_nodes(pos, subscripts)
584
        if is_single_value:
William Stein's avatar
William Stein committed
585 586 587 588 589 590 591 592 593
            index = indexes[0]
        else:
            index = ExprNodes.TupleNode(pos, args = indexes)
        result = ExprNodes.IndexNode(pos,
            base = base, index = index)
    s.expect(']')
    return result

def p_subscript_list(s):
594
    is_single_value = True
William Stein's avatar
William Stein committed
595 596
    items = [p_subscript(s)]
    while s.sy == ',':
597
        is_single_value = False
William Stein's avatar
William Stein committed
598 599 600 601
        s.next()
        if s.sy == ']':
            break
        items.append(p_subscript(s))
602
    return items, is_single_value
William Stein's avatar
William Stein committed
603 604 605 606 607 608 609 610

#subscript: '.' '.' '.' | test | [test] ':' [test] [':' [test]]

def p_subscript(s):
    # Parse a subscript and return a list of
    # 1, 2 or 3 ExprNodes, depending on how
    # many slice elements were encountered.
    pos = s.position()
611 612 613 614 615 616 617 618 619 620
    start = p_slice_element(s, (':',))
    if s.sy != ':':
        return [start]
    s.next()
    stop = p_slice_element(s, (':', ',', ']'))
    if s.sy != ':':
        return [start, stop]
    s.next()
    step = p_slice_element(s, (':', ',', ']'))
    return [start, stop, step]
William Stein's avatar
William Stein committed
621 622 623 624 625

def p_slice_element(s, follow_set):
    # Simple expression which may be missing iff
    # it is followed by something in follow_set.
    if s.sy not in follow_set:
626
        return p_test(s)
William Stein's avatar
William Stein committed
627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657
    else:
        return None

def expect_ellipsis(s):
    s.expect('.')
    s.expect('.')
    s.expect('.')

def make_slice_nodes(pos, subscripts):
    # Convert a list of subscripts as returned
    # by p_subscript_list into a list of ExprNodes,
    # creating SliceNodes for elements with 2 or
    # more components.
    result = []
    for subscript in subscripts:
        if len(subscript) == 1:
            result.append(subscript[0])
        else:
            result.append(make_slice_node(pos, *subscript))
    return result

def make_slice_node(pos, start, stop = None, step = None):
    if not start:
        start = ExprNodes.NoneNode(pos)
    if not stop:
        stop = ExprNodes.NoneNode(pos)
    if not step:
        step = ExprNodes.NoneNode(pos)
    return ExprNodes.SliceNode(pos,
        start = start, stop = stop, step = step)

658
#atom: '(' [yield_expr|testlist_comp] ')' | '[' [listmaker] ']' | '{' [dict_or_set_maker] '}' | '`' testlist '`' | NAME | NUMBER | STRING+
William Stein's avatar
William Stein committed
659 660 661 662 663 664 665 666

def p_atom(s):
    pos = s.position()
    sy = s.sy
    if sy == '(':
        s.next()
        if s.sy == ')':
            result = ExprNodes.TupleNode(pos, args = [])
667 668
        elif s.sy == 'yield':
            result = p_yield_expression(s)
William Stein's avatar
William Stein committed
669
        else:
670
            result = p_testlist_comp(s)
William Stein's avatar
William Stein committed
671 672 673 674 675
        s.expect(')')
        return result
    elif sy == '[':
        return p_list_maker(s)
    elif sy == '{':
676
        return p_dict_or_set_maker(s)
William Stein's avatar
William Stein committed
677 678
    elif sy == '`':
        return p_backquote_expr(s)
679 680 681
    elif sy == '.':
        expect_ellipsis(s)
        return ExprNodes.EllipsisNode(pos)
William Stein's avatar
William Stein committed
682
    elif sy == 'INT':
683
        return p_int_literal(s)
William Stein's avatar
William Stein committed
684 685 686 687 688 689 690 691
    elif sy == 'FLOAT':
        value = s.systring
        s.next()
        return ExprNodes.FloatNode(pos, value = value)
    elif sy == 'IMAG':
        value = s.systring[:-1]
        s.next()
        return ExprNodes.ImagNode(pos, value = value)
692
    elif sy == 'BEGIN_STRING':
693
        kind, bytes_value, unicode_value = p_cat_string_literal(s)
William Stein's avatar
William Stein committed
694
        if kind == 'c':
695
            return ExprNodes.CharNode(pos, value = bytes_value)
696
        elif kind == 'u':
697
            return ExprNodes.UnicodeNode(pos, value = unicode_value, bytes_value = bytes_value)
698
        elif kind == 'b':
699
            return ExprNodes.BytesNode(pos, value = bytes_value)
700 701 702
        elif kind == 'f':
            return ExprNodes.JoinedStrNode(pos, values = unicode_value)
        elif kind == '':
703
            return ExprNodes.StringNode(pos, value = bytes_value, unicode_value = unicode_value)
704 705
        else:
            s.error("invalid string kind '%s'" % kind)
William Stein's avatar
William Stein committed
706
    elif sy == 'IDENT':
707
        name = s.systring
William Stein's avatar
William Stein committed
708
        if name == "None":
709
            result = ExprNodes.NoneNode(pos)
710
        elif name == "True":
711
            result = ExprNodes.BoolNode(pos, value=True)
712
        elif name == "False":
713
            result = ExprNodes.BoolNode(pos, value=False)
714
        elif name == "NULL" and not s.in_python_file:
715
            result = ExprNodes.NullNode(pos)
William Stein's avatar
William Stein committed
716
        else:
717 718 719
            result = p_name(s, name)
        s.next()
        return result
William Stein's avatar
William Stein committed
720 721 722
    else:
        s.error("Expected an identifier or literal")

723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753
def p_int_literal(s):
    pos = s.position()
    value = s.systring
    s.next()
    unsigned = ""
    longness = ""
    while value[-1] in u"UuLl":
        if value[-1] in u"Ll":
            longness += "L"
        else:
            unsigned += "U"
        value = value[:-1]
    # '3L' is ambiguous in Py2 but not in Py3.  '3U' and '3LL' are
    # illegal in Py2 Python files.  All suffixes are illegal in Py3
    # Python files.
    is_c_literal = None
    if unsigned:
        is_c_literal = True
    elif longness:
        if longness == 'LL' or s.context.language_level >= 3:
            is_c_literal = True
    if s.in_python_file:
        if is_c_literal:
            error(pos, "illegal integer literal syntax in Python source file")
        is_c_literal = False
    return ExprNodes.IntNode(pos,
                             is_c_literal = is_c_literal,
                             value = value,
                             unsigned = unsigned,
                             longness = longness)

754

755 756
def p_name(s, name):
    pos = s.position()
757 758
    if not s.compile_time_expr and name in s.compile_time_env:
        value = s.compile_time_env.lookup_here(name)
759 760 761 762 763 764 765 766
        node = wrap_compile_time_constant(pos, value)
        if node is not None:
            return node
    return ExprNodes.NameNode(pos, name=name)


def wrap_compile_time_constant(pos, value):
    rep = repr(value)
767 768
    if value is None:
        return ExprNodes.NoneNode(pos)
769 770
    elif value is Ellipsis:
        return ExprNodes.EllipsisNode(pos)
771
    elif isinstance(value, bool):
772 773
        return ExprNodes.BoolNode(pos, value=value)
    elif isinstance(value, int):
774
        return ExprNodes.IntNode(pos, value=rep, constant_result=value)
775
    elif isinstance(value, float):
776
        return ExprNodes.FloatNode(pos, value=rep, constant_result=value)
777 778 779 780 781 782 783 784 785
    elif isinstance(value, complex):
        node = ExprNodes.ImagNode(pos, value=repr(value.imag), constant_result=complex(0.0, value.imag))
        if value.real:
            # FIXME: should we care about -0.0 ?
            # probably not worth using the '-' operator for negative imag values
            node = ExprNodes.binop_node(
                pos, '+', ExprNodes.FloatNode(pos, value=repr(value.real), constant_result=value.real), node,
                constant_result=value)
        return node
786
    elif isinstance(value, _unicode):
787
        return ExprNodes.UnicodeNode(pos, value=EncodedString(value))
788
    elif isinstance(value, _bytes):
789 790
        bvalue = bytes_literal(value, 'ascii')  # actually: unknown encoding, but BytesLiteral requires one
        return ExprNodes.BytesNode(pos, value=bvalue, constant_result=value)
791 792 793 794 795
    elif isinstance(value, tuple):
        args = [wrap_compile_time_constant(pos, arg)
                for arg in value]
        if None not in args:
            return ExprNodes.TupleNode(pos, args=args)
796
        else:
797 798
            # error already reported
            return None
799
    elif not _IS_PY3 and isinstance(value, long):
800
        return ExprNodes.IntNode(pos, value=rep.rstrip('L'), constant_result=value)
801 802
    error(pos, "Invalid type for compile-time constant: %r (type %s)"
               % (value, value.__class__.__name__))
803 804
    return None

805

William Stein's avatar
William Stein committed
806 807
def p_cat_string_literal(s):
    # A sequence of one or more adjacent string literals.
808
    # Returns (kind, bytes_value, unicode_value)
809 810
    # where kind in ('b', 'c', 'u', 'f', '')
    pos = s.position()
811 812 813
    kind, bytes_value, unicode_value = p_string_literal(s)
    if kind == 'c' or s.sy != 'BEGIN_STRING':
        return kind, bytes_value, unicode_value
814
    bstrings, ustrings, positions = [bytes_value], [unicode_value], [pos]
815 816 817 818 819 820
    bytes_value = unicode_value = None
    while s.sy == 'BEGIN_STRING':
        pos = s.position()
        next_kind, next_bytes_value, next_unicode_value = p_string_literal(s)
        if next_kind == 'c':
            error(pos, "Cannot concatenate char literal with another string or char literal")
821
            continue
822
        elif next_kind != kind:
823
            # concatenating f strings and normal strings is allowed and leads to an f string
Stefan Behnel's avatar
Stefan Behnel committed
824
            if set([kind, next_kind]) in (set(['f', 'u']), set(['f', ''])):
825 826
                kind = 'f'
            else:
827 828
                error(pos, "Cannot mix string literals of different types, expected %s'', got %s''" % (
                    kind, next_kind))
829 830 831 832
                continue
        bstrings.append(next_bytes_value)
        ustrings.append(next_unicode_value)
        positions.append(pos)
833
    # join and rewrap the partial literals
834
    if kind in ('b', 'c', '') or kind == 'u' and None not in bstrings:
835
        # Py3 enforced unicode literals are parsed as bytes/unicode combination
836
        bytes_value = bytes_literal(StringEncoding.join_bytes(bstrings), s.source_encoding)
837
    if kind in ('u', ''):
Stefan Behnel's avatar
Stefan Behnel committed
838
        unicode_value = EncodedString(u''.join([u for u in ustrings if u is not None]))
839 840 841 842 843 844 845
    if kind == 'f':
        unicode_value = []
        for u, pos in zip(ustrings, positions):
            if isinstance(u, list):
                unicode_value += u
            else:
                # non-f-string concatenated into the f-string
Stefan Behnel's avatar
Stefan Behnel committed
846
                unicode_value.append(ExprNodes.UnicodeNode(pos, value=EncodedString(u)))
847 848
    return kind, bytes_value, unicode_value

Stefan Behnel's avatar
Stefan Behnel committed
849

850
def p_opt_string_literal(s, required_type='u'):
851
    if s.sy != 'BEGIN_STRING':
William Stein's avatar
William Stein committed
852
        return None
853 854 855 856 857 858 859 860 861 862
    pos = s.position()
    kind, bytes_value, unicode_value = p_string_literal(s, required_type)
    if required_type == 'u':
        if kind == 'f':
            s.error("f-string not allowed here", pos)
        return unicode_value
    elif required_type == 'b':
        return bytes_value
    else:
        s.error("internal parser configuration error")
William Stein's avatar
William Stein committed
863

Stefan Behnel's avatar
Stefan Behnel committed
864

865 866 867 868 869 870
def check_for_non_ascii_characters(string):
    for c in string:
        if c >= u'\x80':
            return True
    return False

Stefan Behnel's avatar
Stefan Behnel committed
871

872
def p_string_literal(s, kind_override=None):
873
    # A single string or char literal.  Returns (kind, bvalue, uvalue)
874
    # where kind in ('b', 'c', 'u', 'f', '').  The 'bvalue' is the source
875 876 877
    # code byte sequence of the string literal, 'uvalue' is the
    # decoded Unicode string.  Either of the two may be None depending
    # on the 'kind' of string, only unprefixed strings have both
878 879
    # representations. In f-strings, the uvalue is a list of the Unicode
    # strings and f-string expressions that make up the f-string.
880

William Stein's avatar
William Stein committed
881 882
    # s.sy == 'BEGIN_STRING'
    pos = s.position()
883
    is_python3_source = s.context.language_level >= 3
884
    has_non_ascii_literal_characters = False
885
    kind_string = s.systring.rstrip('"\'').lower()
886 887 888 889 890 891 892 893 894
    if len(kind_string) > 1:
        if len(set(kind_string)) != len(kind_string):
            error(pos, 'Duplicate string prefix character')
        if 'b' in kind_string and 'u' in kind_string:
            error(pos, 'String prefixes b and u cannot be combined')
        if 'b' in kind_string and 'f' in kind_string:
            error(pos, 'String prefixes b and f cannot be combined')
        if 'u' in kind_string and 'f' in kind_string:
            error(pos, 'String prefixes u and f cannot be combined')
895 896 897 898 899 900 901

    is_raw = 'r' in kind_string

    if 'c' in kind_string:
        # this should never happen, since the lexer does not allow combining c
        # with other prefix characters
        if len(kind_string) != 1:
902
            error(pos, 'Invalid string prefix for character literal')
903 904
        kind = 'c'
    elif 'f' in kind_string:
905 906
        kind = 'f'     # u is ignored
        is_raw = True  # postpone the escape resolution
907 908 909 910 911
    elif 'b' in kind_string:
        kind = 'b'
    elif 'u' in kind_string:
        kind = 'u'
    else:
William Stein's avatar
William Stein committed
912
        kind = ''
913

914 915 916 917 918 919
    if kind == '' and kind_override is None and Future.unicode_literals in s.context.future_directives:
        chars = StringEncoding.StrLiteralBuilder(s.source_encoding)
        kind = 'u'
    else:
        if kind_override is not None and kind_override in 'ub':
            kind = kind_override
920
        if kind in ('u', 'f'):  # f-strings are scanned exactly like Unicode literals, but are parsed further later
921 922 923 924 925
            chars = StringEncoding.UnicodeLiteralBuilder()
        elif kind == '':
            chars = StringEncoding.StrLiteralBuilder(s.source_encoding)
        else:
            chars = StringEncoding.BytesLiteralBuilder(s.source_encoding)
926

William Stein's avatar
William Stein committed
927 928 929
    while 1:
        s.next()
        sy = s.sy
930
        systr = s.systring
931
        # print "p_string_literal: sy =", sy, repr(s.systring) ###
William Stein's avatar
William Stein committed
932
        if sy == 'CHARS':
933
            chars.append(systr)
934 935
            if is_python3_source and not has_non_ascii_literal_characters and check_for_non_ascii_characters(systr):
                has_non_ascii_literal_characters = True
William Stein's avatar
William Stein committed
936
        elif sy == 'ESCAPE':
937 938
            # in Py2, 'ur' raw unicode strings resolve unicode escapes but nothing else
            if is_raw and (is_python3_source or kind != 'u' or systr[1] not in u'Uu'):
939
                chars.append(systr)
940 941
                if is_python3_source and not has_non_ascii_literal_characters and check_for_non_ascii_characters(systr):
                    has_non_ascii_literal_characters = True
William Stein's avatar
William Stein committed
942
            else:
943
                _append_escape_sequence(kind, chars, systr, s)
William Stein's avatar
William Stein committed
944
        elif sy == 'NEWLINE':
945
            chars.append(u'\n')
William Stein's avatar
William Stein committed
946 947 948
        elif sy == 'END_STRING':
            break
        elif sy == 'EOF':
949
            s.error("Unclosed string literal", pos=pos)
William Stein's avatar
William Stein committed
950
        else:
951 952
            s.error("Unexpected token %r:%r in string literal" % (
                sy, s.systring))
953

954
    if kind == 'c':
955 956 957 958
        unicode_value = None
        bytes_value = chars.getchar()
        if len(bytes_value) != 1:
            error(pos, u"invalid character literal: %r" % bytes_value)
959
    else:
960
        bytes_value, unicode_value = chars.getstrings()
961
        if is_python3_source and has_non_ascii_literal_characters:
962
            # Python 3 forbids literal non-ASCII characters in byte strings
963
            if kind == 'b':
964
                s.error("bytes can only contain ASCII literal characters.", pos=pos)
965
            bytes_value = None
966
    if kind == 'f':
967
        unicode_value = p_f_string(s, unicode_value, pos, is_raw='r' in kind_string)
William Stein's avatar
William Stein committed
968
    s.next()
969
    return (kind, bytes_value, unicode_value)
William Stein's avatar
William Stein committed
970

971

972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989
def _append_escape_sequence(kind, builder, escape_sequence, s):
    c = escape_sequence[1]
    if c in u"01234567":
        builder.append_charval(int(escape_sequence[1:], 8))
    elif c in u"'\"\\":
        builder.append(c)
    elif c in u"abfnrtv":
        builder.append(StringEncoding.char_from_escape_sequence(escape_sequence))
    elif c == u'\n':
        pass  # line continuation
    elif c == u'x':  # \xXX
        if len(escape_sequence) == 4:
            builder.append_charval(int(escape_sequence[2:], 16))
        else:
            s.error("Invalid hex escape '%s'" % escape_sequence, fatal=False)
    elif c in u'NUu' and kind in ('u', 'f', ''):  # \uxxxx, \Uxxxxxxxx, \N{...}
        chrval = -1
        if c == u'N':
990
            uchar = None
991
            try:
992 993
                uchar = lookup_unicodechar(escape_sequence[3:-1])
                chrval = ord(uchar)
994 995 996
            except KeyError:
                s.error("Unknown Unicode character name %s" %
                        repr(escape_sequence[3:-1]).lstrip('u'), fatal=False)
997 998 999 1000 1001 1002 1003 1004
            except TypeError:
                # 2-byte unicode build of CPython?
                if (uchar is not None and _IS_2BYTE_UNICODE and len(uchar) == 2 and
                        unicode_category(uchar[0]) == 'Cs' and unicode_category(uchar[1]) == 'Cs'):
                    # surrogate pair instead of single character
                    chrval = 0x10000 + (ord(uchar[0]) - 0xd800) >> 10 + (ord(uchar[1]) - 0xdc00)
                else:
                    raise
1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
        elif len(escape_sequence) in (6, 10):
            chrval = int(escape_sequence[2:], 16)
            if chrval > 1114111:  # sys.maxunicode:
                s.error("Invalid unicode escape '%s'" % escape_sequence)
                chrval = -1
        else:
            s.error("Invalid unicode escape '%s'" % escape_sequence, fatal=False)
        if chrval >= 0:
            builder.append_uescape(chrval, escape_sequence)
    else:
        builder.append(escape_sequence)


1018
_parse_escape_sequences_raw, _parse_escape_sequences = [re.compile((
1019
    # escape sequences:
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029
    br'(\\(?:' +
    (br'\\?' if is_raw else (
        br'[\\abfnrtv"\'{]|'
        br'[0-7]{2,3}|'
        br'N\{[^}]*\}|'
        br'x[0-9a-fA-F]{2}|'
        br'u[0-9a-fA-F]{4}|'
        br'U[0-9a-fA-F]{8}|'
        br'[NxuU]|'  # detect invalid escape sequences that do not match above
    )) +
1030 1031 1032 1033
    br')?|'
    # non-escape sequences:
    br'\{\{?|'
    br'\}\}?|'
1034 1035 1036
    br'[^\\{}]+)'
    ).decode('us-ascii')).match
    for is_raw in (True, False)]
1037 1038 1039


def p_f_string(s, unicode_value, pos, is_raw):
1040 1041 1042
    # Parses a PEP 498 f-string literal into a list of nodes. Nodes are either UnicodeNodes
    # or FormattedValueNodes.
    values = []
1043
    next_start = 0
1044
    size = len(unicode_value)
1045
    builder = StringEncoding.UnicodeLiteralBuilder()
1046 1047
    error_pos = list(pos)  # [src, line, column]
    _parse_seq = _parse_escape_sequences_raw if is_raw else _parse_escape_sequences
1048 1049 1050

    while next_start < size:
        end = next_start
1051 1052
        error_pos[2] = pos[2] + end  # FIXME: handle newlines in string
        match = _parse_seq(unicode_value, next_start)
1053
        if match is None:
1054
            error(tuple(error_pos), "Invalid escape sequence")
1055 1056 1057 1058 1059 1060 1061

        next_start = match.end()
        part = match.group()
        c = part[0]
        if c == '\\':
            if not is_raw and len(part) > 1:
                _append_escape_sequence('f', builder, part, s)
1062
            else:
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072
                builder.append(part)
        elif c == '{':
            if part == '{{':
                builder.append('{')
            else:
                # start of an expression
                if builder.chars:
                    values.append(ExprNodes.UnicodeNode(pos, value=builder.getstring()))
                    builder = StringEncoding.UnicodeLiteralBuilder()
                next_start, expr_node = p_f_string_expr(s, unicode_value, pos, next_start, is_raw)
1073
                values.append(expr_node)
1074 1075 1076 1077
        elif c == '}':
            if part == '}}':
                builder.append('}')
            else:
1078
                s.error("f-string: single '}' is not allowed", pos=tuple(error_pos))
1079
        else:
1080
            builder.append(part)
1081

1082 1083
    if builder.chars:
        values.append(ExprNodes.UnicodeNode(pos, value=builder.getstring()))
1084 1085 1086
    return values


1087
def p_f_string_expr(s, unicode_value, pos, starting_index, is_raw):
1088 1089 1090 1091
    # Parses a {}-delimited expression inside an f-string. Returns a FormattedValueNode
    # and the index in the string that follows the expression.
    i = starting_index
    size = len(unicode_value)
1092
    conversion_char = terminal_char = format_spec = None
1093
    format_spec_str = None
1094
    NO_CHAR = 2**30
1095 1096

    nested_depth = 0
1097
    quote_char = NO_CHAR
1098
    in_triple_quotes = False
1099 1100 1101 1102 1103 1104

    while True:
        if i >= size:
            s.error("missing '}' in format string expression")
        c = unicode_value[i]

1105
        if quote_char != NO_CHAR:
1106
            if c == '\\':
1107 1108
                error_pos = (pos[0], pos[1] + i, pos[2])  # FIXME: handle newlines in string
                error(error_pos, "backslashes not allowed in f-strings")
1109 1110 1111 1112
            elif c == quote_char:
                if in_triple_quotes:
                    if i + 2 < size and unicode_value[i + 1] == c and unicode_value[i + 2] == c:
                        in_triple_quotes = False
1113
                        quote_char = NO_CHAR
1114 1115
                        i += 2
                else:
1116
                    quote_char = NO_CHAR
1117 1118 1119 1120 1121 1122
        elif c in '\'"':
            quote_char = c
            if i + 2 < size and unicode_value[i + 1] == c and unicode_value[i + 2] == c:
                in_triple_quotes = True
                i += 2
        elif c in '{[(':
1123 1124 1125 1126 1127 1128 1129 1130
            nested_depth += 1
        elif nested_depth != 0 and c in '}])':
            nested_depth -= 1
        elif c == '#':
            s.error("format string cannot include #")
        elif nested_depth == 0 and c in '!:}':
            # allow != as a special case
            if c == '!' and i + 1 < size and unicode_value[i + 1] == '=':
Jelle Zijlstra's avatar
Jelle Zijlstra committed
1131
                i += 1
1132 1133 1134 1135 1136 1137
                continue

            terminal_char = c
            break
        i += 1

1138 1139 1140 1141 1142
    # normalise line endings as the parser expects that
    expr_str = unicode_value[starting_index:i].replace('\r\n', '\n').replace('\r', '\n')
    expr_pos = (pos[0], pos[1], pos[2] + starting_index + 2)  # TODO: find exact code position (concat, multi-line, ...)

    if not expr_str.strip():
1143
        error(expr_pos, "empty expression not allowed in f-string")
1144 1145 1146

    if terminal_char == '!':
        i += 1
1147
        if i + 2 > size:
1148
            error(expr_pos, "invalid conversion char at end of string")
1149 1150 1151 1152
        else:
            conversion_char = unicode_value[i]
            i += 1
            terminal_char = unicode_value[i]
1153 1154

    if terminal_char == ':':
1155 1156
        in_triple_quotes = False
        in_string = False
1157 1158 1159 1160
        nested_depth = 0
        start_format_spec = i + 1
        while True:
            if i >= size:
1161
                s.error("missing '}' in format specifier", pos=expr_pos)
1162
            c = unicode_value[i]
1163 1164 1165 1166
            if not in_triple_quotes and not in_string:
                if c == '{':
                    nested_depth += 1
                elif c == '}':
1167 1168 1169 1170 1171
                    if nested_depth > 0:
                        nested_depth -= 1
                    else:
                        terminal_char = c
                        break
1172 1173 1174 1175 1176 1177
            if c in '\'"':
                if not in_string and i + 2 < size and unicode_value[i + 1] == c and unicode_value[i + 2] == c:
                    in_triple_quotes = not in_triple_quotes
                    i += 2
                elif not in_triple_quotes:
                    in_string = not in_string
1178 1179 1180 1181 1182
            i += 1

        format_spec_str = unicode_value[start_format_spec:i]

    if terminal_char != '}':
1183
        s.error("missing '}' in format string expression', found '%s'" % terminal_char)
1184

1185 1186 1187
    # parse the expression as if it was surrounded by parentheses
    buf = StringIO('(%s)' % expr_str)
    scanner = PyrexScanner(buf, expr_pos[0], parent_scanner=s, source_encoding=s.source_encoding, initial_pos=expr_pos)
1188 1189 1190
    expr = p_testlist(scanner)  # TODO is testlist right here?

    # validate the conversion char
1191
    if conversion_char is not None and not ExprNodes.FormattedValueNode.find_conversion_func(conversion_char):
1192
        error(pos, "invalid conversion character '%s'" % conversion_char)
1193 1194

    # the format spec is itself treated like an f-string
1195
    if format_spec_str:
1196
        format_spec = ExprNodes.JoinedStrNode(pos, values=p_f_string(s, format_spec_str, pos, is_raw))
1197 1198

    return i + 1, ExprNodes.FormattedValueNode(
1199
        pos, value=expr, conversion_char=conversion_char, format_spec=format_spec)
1200 1201


1202 1203 1204
# since PEP 448:
# list_display  ::=     "[" [listmaker] "]"
# listmaker     ::=     (test|star_expr) ( comp_for | (',' (test|star_expr))* [','] )
Stefan Behnel's avatar
Stefan Behnel committed
1205
# comp_iter     ::=     comp_for | comp_if
1206
# comp_for      ::=     ["async"] "for" expression_list "in" testlist [comp_iter]
1207
# comp_if       ::=     "if" test [comp_iter]
1208

William Stein's avatar
William Stein committed
1209 1210 1211 1212
def p_list_maker(s):
    # s.sy == '['
    pos = s.position()
    s.next()
Robert Bradshaw's avatar
Robert Bradshaw committed
1213 1214
    if s.sy == ']':
        s.expect(']')
1215 1216 1217
        return ExprNodes.ListNode(pos, args=[])

    expr = p_test_or_starred_expr(s)
1218
    if s.sy in ('for', 'async'):
1219 1220
        if expr.is_starred:
            s.error("iterable unpacking cannot be used in comprehension")
1221
        append = ExprNodes.ComprehensionAppendNode(pos, expr=expr)
1222
        loop = p_comp_for(s, append)
Robert Bradshaw's avatar
Robert Bradshaw committed
1223
        s.expect(']')
1224
        return ExprNodes.ComprehensionNode(
1225
            pos, loop=loop, append=append, type=Builtin.list_type,
1226
            # list comprehensions leak their loop variable in Py2
1227 1228 1229 1230 1231 1232
            has_local_scope=s.context.language_level >= 3)

    # (merged) list literal
    if s.sy == ',':
        s.next()
        exprs = p_test_or_starred_expr_list(s, expr)
Robert Bradshaw's avatar
Robert Bradshaw committed
1233
    else:
1234 1235 1236 1237
        exprs = [expr]
    s.expect(']')
    return ExprNodes.ListNode(pos, args=exprs)

1238

Stefan Behnel's avatar
Stefan Behnel committed
1239
def p_comp_iter(s, body):
1240
    if s.sy in ('for', 'async'):
1241
        return p_comp_for(s, body)
Robert Bradshaw's avatar
Robert Bradshaw committed
1242
    elif s.sy == 'if':
Stefan Behnel's avatar
Stefan Behnel committed
1243
        return p_comp_if(s, body)
Robert Bradshaw's avatar
Robert Bradshaw committed
1244
    else:
1245 1246
        # insert the 'append' operation into the loop
        return body
William Stein's avatar
William Stein committed
1247

1248
def p_comp_for(s, body):
Robert Bradshaw's avatar
Robert Bradshaw committed
1249
    pos = s.position()
1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
    # [async] for ...
    is_async = False
    if s.sy == 'async':
        is_async = True
        s.next()

    # s.sy == 'for'
    s.expect('for')
    kw = p_for_bounds(s, allow_testlist=False, is_async=is_async)
    kw.update(else_clause=None, body=p_comp_iter(s, body), is_async=is_async)
Robert Bradshaw's avatar
Robert Bradshaw committed
1260
    return Nodes.ForStatNode(pos, **kw)
1261

Stefan Behnel's avatar
Stefan Behnel committed
1262
def p_comp_if(s, body):
Robert Bradshaw's avatar
Robert Bradshaw committed
1263 1264 1265
    # s.sy == 'if'
    pos = s.position()
    s.next()
Stefan Behnel's avatar
Stefan Behnel committed
1266
    test = p_test_nocond(s)
1267
    return Nodes.IfStatNode(pos,
1268
        if_clauses = [Nodes.IfClauseNode(pos, condition = test,
Stefan Behnel's avatar
Stefan Behnel committed
1269
                                         body = p_comp_iter(s, body))],
Robert Bradshaw's avatar
Robert Bradshaw committed
1270
        else_clause = None )
1271

1272 1273 1274 1275 1276 1277

# since PEP 448:
#dictorsetmaker: ( ((test ':' test | '**' expr)
#                   (comp_for | (',' (test ':' test | '**' expr))* [','])) |
#                  ((test | star_expr)
#                   (comp_for | (',' (test | star_expr))* [','])) )
William Stein's avatar
William Stein committed
1278

1279
def p_dict_or_set_maker(s):
William Stein's avatar
William Stein committed
1280 1281 1282
    # s.sy == '{'
    pos = s.position()
    s.next()
1283
    if s.sy == '}':
William Stein's avatar
William Stein committed
1284
        s.next()
1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298
        return ExprNodes.DictNode(pos, key_value_pairs=[])

    parts = []
    target_type = 0
    last_was_simple_item = False
    while True:
        if s.sy in ('*', '**'):
            # merged set/dict literal
            if target_type == 0:
                target_type = 1 if s.sy == '*' else 2  # 'stars'
            elif target_type != len(s.sy):
                s.error("unexpected %sitem found in %s literal" % (
                    s.sy, 'set' if target_type == 1 else 'dict'))
            s.next()
1299 1300 1301
            if s.sy == '*':
                s.error("expected expression, found '*'")
            item = p_starred_expr(s)
1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320
            parts.append(item)
            last_was_simple_item = False
        else:
            item = p_test(s)
            if target_type == 0:
                target_type = 2 if s.sy == ':' else 1  # dict vs. set
            if target_type == 2:
                # dict literal
                s.expect(':')
                key = item
                value = p_test(s)
                item = ExprNodes.DictItemNode(key.pos, key=key, value=value)
            if last_was_simple_item:
                parts[-1].append(item)
            else:
                parts.append([item])
                last_was_simple_item = True

        if s.sy == ',':
1321
            s.next()
1322 1323
            if s.sy == '}':
                break
1324 1325 1326
        else:
            break

1327
    if s.sy in ('for', 'async'):
1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338
        # dict/set comprehension
        if len(parts) == 1 and isinstance(parts[0], list) and len(parts[0]) == 1:
            item = parts[0][0]
            if target_type == 2:
                assert isinstance(item, ExprNodes.DictItemNode), type(item)
                comprehension_type = Builtin.dict_type
                append = ExprNodes.DictComprehensionAppendNode(
                    item.pos, key_expr=item.key, value_expr=item.value)
            else:
                comprehension_type = Builtin.set_type
                append = ExprNodes.ComprehensionAppendNode(item.pos, expr=item)
1339
            loop = p_comp_for(s, append)
1340
            s.expect('}')
1341
            return ExprNodes.ComprehensionNode(pos, loop=loop, append=append, type=comprehension_type)
1342
        else:
1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367
            # syntax error, try to find a good error message
            if len(parts) == 1 and not isinstance(parts[0], list):
                s.error("iterable unpacking cannot be used in comprehension")
            else:
                # e.g. "{1,2,3 for ..."
                s.expect('}')
            return ExprNodes.DictNode(pos, key_value_pairs=[])

    s.expect('}')
    if target_type == 1:
        # (merged) set literal
        items = []
        set_items = []
        for part in parts:
            if isinstance(part, list):
                set_items.extend(part)
            else:
                if set_items:
                    items.append(ExprNodes.SetNode(set_items[0].pos, args=set_items))
                    set_items = []
                items.append(part)
        if set_items:
            items.append(ExprNodes.SetNode(set_items[0].pos, args=set_items))
        if len(items) == 1 and items[0].is_set_literal:
            return items[0]
1368
        return ExprNodes.MergedSequenceNode(pos, args=items, type=Builtin.set_type)
1369
    else:
1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
        # (merged) dict literal
        items = []
        dict_items = []
        for part in parts:
            if isinstance(part, list):
                dict_items.extend(part)
            else:
                if dict_items:
                    items.append(ExprNodes.DictNode(dict_items[0].pos, key_value_pairs=dict_items))
                    dict_items = []
                items.append(part)
        if dict_items:
            items.append(ExprNodes.DictNode(dict_items[0].pos, key_value_pairs=dict_items))
        if len(items) == 1 and items[0].is_dict_literal:
            return items[0]
        return ExprNodes.MergedDictNode(pos, keyword_args=items, reject_duplicates=False)

William Stein's avatar
William Stein committed
1387

1388
# NOTE: no longer in Py3 :)
William Stein's avatar
William Stein committed
1389 1390 1391 1392
def p_backquote_expr(s):
    # s.sy == '`'
    pos = s.position()
    s.next()
1393 1394 1395 1396
    args = [p_test(s)]
    while s.sy == ',':
        s.next()
        args.append(p_test(s))
William Stein's avatar
William Stein committed
1397
    s.expect('`')
1398 1399 1400 1401
    if len(args) == 1:
        arg = args[0]
    else:
        arg = ExprNodes.TupleNode(pos, args = args)
William Stein's avatar
William Stein committed
1402 1403
    return ExprNodes.BackquoteNode(pos, arg = arg)

1404 1405
def p_simple_expr_list(s, expr=None):
    exprs = expr is not None and [expr] or []
William Stein's avatar
William Stein committed
1406
    while s.sy not in expr_terminators:
1407
        exprs.append( p_test(s) )
Stefan Behnel's avatar
Stefan Behnel committed
1408
        if s.sy != ',':
William Stein's avatar
William Stein committed
1409 1410 1411 1412
            break
        s.next()
    return exprs

1413

1414 1415 1416
def p_test_or_starred_expr_list(s, expr=None):
    exprs = expr is not None and [expr] or []
    while s.sy not in expr_terminators:
1417
        exprs.append(p_test_or_starred_expr(s))
1418 1419 1420 1421
        if s.sy != ',':
            break
        s.next()
    return exprs
1422

1423 1424 1425 1426

#testlist: test (',' test)* [',']

def p_testlist(s):
William Stein's avatar
William Stein committed
1427
    pos = s.position()
1428
    expr = p_test(s)
William Stein's avatar
William Stein committed
1429 1430
    if s.sy == ',':
        s.next()
1431
        exprs = p_simple_expr_list(s, expr)
William Stein's avatar
William Stein committed
1432 1433 1434 1435
        return ExprNodes.TupleNode(pos, args = exprs)
    else:
        return expr

1436
# testlist_star_expr: (test|star_expr) ( comp_for | (',' (test|star_expr))* [','] )
Robert Bradshaw's avatar
Robert Bradshaw committed
1437

1438
def p_testlist_star_expr(s):
Robert Bradshaw's avatar
Robert Bradshaw committed
1439
    pos = s.position()
1440
    expr = p_test_or_starred_expr(s)
Robert Bradshaw's avatar
Robert Bradshaw committed
1441
    if s.sy == ',':
1442
        s.next()
1443 1444
        exprs = p_test_or_starred_expr_list(s, expr)
        return ExprNodes.TupleNode(pos, args = exprs)
Robert Bradshaw's avatar
Robert Bradshaw committed
1445 1446 1447
    else:
        return expr

1448 1449 1450
# testlist_comp: (test|star_expr) ( comp_for | (',' (test|star_expr))* [','] )

def p_testlist_comp(s):
Robert Bradshaw's avatar
Robert Bradshaw committed
1451
    pos = s.position()
1452
    expr = p_test_or_starred_expr(s)
Robert Bradshaw's avatar
Robert Bradshaw committed
1453
    if s.sy == ',':
1454
        s.next()
1455
        exprs = p_test_or_starred_expr_list(s, expr)
Robert Bradshaw's avatar
Robert Bradshaw committed
1456
        return ExprNodes.TupleNode(pos, args = exprs)
1457
    elif s.sy in ('for', 'async'):
1458
        return p_genexp(s, expr)
Robert Bradshaw's avatar
Robert Bradshaw committed
1459 1460
    else:
        return expr
1461 1462

def p_genexp(s, expr):
1463
    # s.sy == 'async' | 'for'
1464 1465
    loop = p_comp_for(s, Nodes.ExprStatNode(
        expr.pos, expr = ExprNodes.YieldExprNode(expr.pos, arg=expr)))
1466 1467
    return ExprNodes.GeneratorExpressionNode(expr.pos, loop=loop)

1468 1469
expr_terminators = cython.declare(set, set([
    ')', ']', '}', ':', '=', 'NEWLINE']))
William Stein's avatar
William Stein committed
1470

1471

William Stein's avatar
William Stein committed
1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484
#-------------------------------------------------------
#
#   Statements
#
#-------------------------------------------------------

def p_global_statement(s):
    # assume s.sy == 'global'
    pos = s.position()
    s.next()
    names = p_ident_list(s)
    return Nodes.GlobalNode(pos, names = names)

1485

1486 1487 1488 1489 1490 1491
def p_nonlocal_statement(s):
    pos = s.position()
    s.next()
    names = p_ident_list(s)
    return Nodes.NonlocalNode(pos, names = names)

1492

William Stein's avatar
William Stein committed
1493
def p_expression_or_assignment(s):
1494 1495 1496 1497 1498
    expr = p_testlist_star_expr(s)
    if s.sy == ':' and (expr.is_name or expr.is_subscript or expr.is_attribute):
        s.next()
        expr.annotation = p_test(s)
    if s.sy == '=' and expr.is_starred:
1499 1500 1501 1502
        # This is a common enough error to make when learning Cython to let
        # it fail as early as possible and give a very clear error message.
        s.error("a starred assignment target must be in a list or tuple"
                " - maybe you meant to use an index assignment: var[0] = ...",
1503 1504
                pos=expr.pos)
    expr_list = [expr]
William Stein's avatar
William Stein committed
1505 1506
    while s.sy == '=':
        s.next()
1507 1508 1509 1510 1511
        if s.sy == 'yield':
            expr = p_yield_expression(s)
        else:
            expr = p_testlist_star_expr(s)
        expr_list.append(expr)
William Stein's avatar
William Stein committed
1512
    if len(expr_list) == 1:
1513
        if re.match(r"([-+*/%^&|]|<<|>>|\*\*|//|@)=", s.sy):
1514
            lhs = expr_list[0]
1515 1516 1517 1518 1519 1520
            if isinstance(lhs, ExprNodes.SliceIndexNode):
                # implementation requires IndexNode
                lhs = ExprNodes.IndexNode(
                    lhs.pos,
                    base=lhs.base,
                    index=make_slice_node(lhs.pos, lhs.start, lhs.stop))
1521
            elif not isinstance(lhs, (ExprNodes.AttributeNode, ExprNodes.IndexNode, ExprNodes.NameNode)):
1522
                error(lhs.pos, "Illegal operand for inplace operation.")
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
1523
            operator = s.sy[:-1]
1524
            s.next()
1525 1526 1527 1528
            if s.sy == 'yield':
                rhs = p_yield_expression(s)
            else:
                rhs = p_testlist(s)
1529
            return Nodes.InPlaceAssignmentNode(lhs.pos, operator=operator, lhs=lhs, rhs=rhs)
1530
        expr = expr_list[0]
1531
        return Nodes.ExprStatNode(expr.pos, expr=expr)
1532

1533 1534
    rhs = expr_list[-1]
    if len(expr_list) == 2:
1535
        return Nodes.SingleAssignmentNode(rhs.pos, lhs=expr_list[0], rhs=rhs)
William Stein's avatar
William Stein committed
1536
    else:
1537 1538
        return Nodes.CascadedAssignmentNode(rhs.pos, lhs_list=expr_list[:-1], rhs=rhs)

William Stein's avatar
William Stein committed
1539 1540 1541 1542

def p_print_statement(s):
    # s.sy == 'print'
    pos = s.position()
1543
    ends_with_comma = 0
William Stein's avatar
William Stein committed
1544 1545
    s.next()
    if s.sy == '>>':
1546
        s.next()
1547
        stream = p_test(s)
1548 1549 1550 1551 1552
        if s.sy == ',':
            s.next()
            ends_with_comma = s.sy in ('NEWLINE', 'EOF')
    else:
        stream = None
William Stein's avatar
William Stein committed
1553 1554
    args = []
    if s.sy not in ('NEWLINE', 'EOF'):
1555
        args.append(p_test(s))
William Stein's avatar
William Stein committed
1556 1557 1558
        while s.sy == ',':
            s.next()
            if s.sy in ('NEWLINE', 'EOF'):
1559
                ends_with_comma = 1
William Stein's avatar
William Stein committed
1560
                break
1561
            args.append(p_test(s))
1562
    arg_tuple = ExprNodes.TupleNode(pos, args=args)
1563
    return Nodes.PrintStatNode(pos,
1564 1565 1566
        arg_tuple=arg_tuple, stream=stream,
        append_newline=not ends_with_comma)

William Stein's avatar
William Stein committed
1567

1568 1569 1570 1571
def p_exec_statement(s):
    # s.sy == 'exec'
    pos = s.position()
    s.next()
1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583
    code = p_bit_expr(s)
    if isinstance(code, ExprNodes.TupleNode):
        # Py3 compatibility syntax
        tuple_variant = True
        args = code.args
        if len(args) not in (2, 3):
            s.error("expected tuple of length 2 or 3, got length %d" % len(args),
                    pos=pos, fatal=False)
            args = [code]
    else:
        tuple_variant = False
        args = [code]
1584
    if s.sy == 'in':
1585 1586 1587
        if tuple_variant:
            s.error("tuple variant of exec does not support additional 'in' arguments",
                    fatal=False)
1588
        s.next()
1589
        args.append(p_test(s))
1590 1591
        if s.sy == ',':
            s.next()
1592
            args.append(p_test(s))
1593
    return Nodes.ExecStatNode(pos, args=args)
1594

William Stein's avatar
William Stein committed
1595 1596 1597 1598
def p_del_statement(s):
    # s.sy == 'del'
    pos = s.position()
    s.next()
1599
    # FIXME: 'exprlist' in Python
William Stein's avatar
William Stein committed
1600 1601 1602 1603 1604 1605 1606
    args = p_simple_expr_list(s)
    return Nodes.DelStatNode(pos, args = args)

def p_pass_statement(s, with_newline = 0):
    pos = s.position()
    s.expect('pass')
    if with_newline:
1607
        s.expect_newline("Expected a newline", ignore_semicolon=True)
William Stein's avatar
William Stein committed
1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626
    return Nodes.PassStatNode(pos)

def p_break_statement(s):
    # s.sy == 'break'
    pos = s.position()
    s.next()
    return Nodes.BreakStatNode(pos)

def p_continue_statement(s):
    # s.sy == 'continue'
    pos = s.position()
    s.next()
    return Nodes.ContinueStatNode(pos)

def p_return_statement(s):
    # s.sy == 'return'
    pos = s.position()
    s.next()
    if s.sy not in statement_terminators:
1627
        value = p_testlist(s)
William Stein's avatar
William Stein committed
1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638
    else:
        value = None
    return Nodes.ReturnStatNode(pos, value = value)

def p_raise_statement(s):
    # s.sy == 'raise'
    pos = s.position()
    s.next()
    exc_type = None
    exc_value = None
    exc_tb = None
Haoyu Bai's avatar
Haoyu Bai committed
1639
    cause = None
William Stein's avatar
William Stein committed
1640
    if s.sy not in statement_terminators:
1641
        exc_type = p_test(s)
William Stein's avatar
William Stein committed
1642 1643
        if s.sy == ',':
            s.next()
1644
            exc_value = p_test(s)
William Stein's avatar
William Stein committed
1645 1646
            if s.sy == ',':
                s.next()
1647
                exc_tb = p_test(s)
Haoyu Bai's avatar
Haoyu Bai committed
1648 1649 1650
        elif s.sy == 'from':
            s.next()
            cause = p_test(s)
1651
    if exc_type or exc_value or exc_tb:
1652
        return Nodes.RaiseStatNode(pos,
1653 1654
            exc_type = exc_type,
            exc_value = exc_value,
Haoyu Bai's avatar
Haoyu Bai committed
1655 1656
            exc_tb = exc_tb,
            cause = cause)
1657 1658
    else:
        return Nodes.ReraiseStatNode(pos)
William Stein's avatar
William Stein committed
1659

1660

William Stein's avatar
William Stein committed
1661 1662 1663 1664 1665
def p_import_statement(s):
    # s.sy in ('import', 'cimport')
    pos = s.position()
    kind = s.sy
    s.next()
1666
    items = [p_dotted_name(s, as_allowed=1)]
William Stein's avatar
William Stein committed
1667 1668
    while s.sy == ',':
        s.next()
1669
        items.append(p_dotted_name(s, as_allowed=1))
William Stein's avatar
William Stein committed
1670
    stats = []
1671
    is_absolute = Future.absolute_import in s.context.future_directives
William Stein's avatar
William Stein committed
1672 1673
    for pos, target_name, dotted_name, as_name in items:
        if kind == 'cimport':
1674 1675 1676 1677 1678
            stat = Nodes.CImportStatNode(
                pos,
                module_name=dotted_name,
                as_name=as_name,
                is_absolute=is_absolute)
William Stein's avatar
William Stein committed
1679
        else:
1680
            if as_name and "." in dotted_name:
1681
                name_list = ExprNodes.ListNode(pos, args=[
1682
                    ExprNodes.IdentifierStringNode(pos, value=s.context.intern_ustring("*"))])
1683 1684
            else:
                name_list = None
1685 1686 1687 1688 1689 1690 1691 1692
            stat = Nodes.SingleAssignmentNode(
                pos,
                lhs=ExprNodes.NameNode(pos, name=as_name or target_name),
                rhs=ExprNodes.ImportNode(
                    pos,
                    module_name=ExprNodes.IdentifierStringNode(pos, value=dotted_name),
                    level=0 if is_absolute else None,
                    name_list=name_list))
William Stein's avatar
William Stein committed
1693
        stats.append(stat)
1694 1695
    return Nodes.StatListNode(pos, stats=stats)

William Stein's avatar
William Stein committed
1696

Stefan Behnel's avatar
Stefan Behnel committed
1697
def p_from_import_statement(s, first_statement = 0):
William Stein's avatar
William Stein committed
1698 1699 1700
    # s.sy == 'from'
    pos = s.position()
    s.next()
Haoyu Bai's avatar
Haoyu Bai committed
1701 1702 1703 1704 1705 1706 1707
    if s.sy == '.':
        # count relative import level
        level = 0
        while s.sy == '.':
            level += 1
            s.next()
    else:
1708
        level = None
1709
    if level is not None and s.sy in ('import', 'cimport'):
Haoyu Bai's avatar
Haoyu Bai committed
1710
        # we are dealing with "from .. import foo, bar"
1711
        dotted_name_pos, dotted_name = s.position(), s.context.intern_ustring('')
William Stein's avatar
William Stein committed
1712
    else:
1713 1714 1715 1716
        if level is None and Future.absolute_import in s.context.future_directives:
            level = 0
        (dotted_name_pos, _, dotted_name, _) = p_dotted_name(s, as_allowed=False)
    if s.sy not in ('import', 'cimport'):
William Stein's avatar
William Stein committed
1717
        s.error("Expected 'import' or 'cimport'")
1718 1719
    kind = s.sy
    s.next()
Haoyu Bai's avatar
Haoyu Bai committed
1720

1721
    is_cimport = kind == 'cimport'
1722
    is_parenthesized = False
William Stein's avatar
William Stein committed
1723
    if s.sy == '*':
1724
        imported_names = [(s.position(), s.context.intern_ustring("*"), None, None)]
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
1725 1726
        s.next()
    else:
1727 1728 1729
        if s.sy == '(':
            is_parenthesized = True
            s.next()
1730
        imported_names = [p_imported_name(s, is_cimport)]
William Stein's avatar
William Stein committed
1731 1732
    while s.sy == ',':
        s.next()
1733 1734
        if is_parenthesized and s.sy == ')':
            break
1735
        imported_names.append(p_imported_name(s, is_cimport))
1736 1737
    if is_parenthesized:
        s.expect(')')
Stefan Behnel's avatar
Stefan Behnel committed
1738 1739 1740
    if dotted_name == '__future__':
        if not first_statement:
            s.error("from __future__ imports must occur at the beginning of the file")
1741
        elif level:
Haoyu Bai's avatar
Haoyu Bai committed
1742
            s.error("invalid syntax")
Stefan Behnel's avatar
Stefan Behnel committed
1743
        else:
1744
            for (name_pos, name, as_name, kind) in imported_names:
1745 1746 1747
                if name == "braces":
                    s.error("not a chance", name_pos)
                    break
Stefan Behnel's avatar
Stefan Behnel committed
1748 1749 1750
                try:
                    directive = getattr(Future, name)
                except AttributeError:
1751
                    s.error("future feature %s is not defined" % name, name_pos)
Stefan Behnel's avatar
Stefan Behnel committed
1752 1753 1754 1755
                    break
                s.context.future_directives.add(directive)
        return Nodes.PassStatNode(pos)
    elif kind == 'cimport':
1756 1757 1758 1759
        return Nodes.FromCImportStatNode(
            pos, module_name=dotted_name,
            relative_level=level,
            imported_names=imported_names)
William Stein's avatar
William Stein committed
1760 1761 1762
    else:
        imported_name_strings = []
        items = []
1763
        for (name_pos, name, as_name, kind) in imported_names:
William Stein's avatar
William Stein committed
1764
            imported_name_strings.append(
1765
                ExprNodes.IdentifierStringNode(name_pos, value=name))
William Stein's avatar
William Stein committed
1766
            items.append(
1767
                (name, ExprNodes.NameNode(name_pos, name=as_name or name)))
William Stein's avatar
William Stein committed
1768
        import_list = ExprNodes.ListNode(
1769
            imported_names[0][0], args=imported_name_strings)
William Stein's avatar
William Stein committed
1770 1771
        return Nodes.FromImportStatNode(pos,
            module = ExprNodes.ImportNode(dotted_name_pos,
1772
                module_name = ExprNodes.IdentifierStringNode(pos, value = dotted_name),
Haoyu Bai's avatar
Haoyu Bai committed
1773
                level = level,
William Stein's avatar
William Stein committed
1774 1775 1776
                name_list = import_list),
            items = items)

1777 1778

imported_name_kinds = cython.declare(set, set(['class', 'struct', 'union']))
1779 1780

def p_imported_name(s, is_cimport):
William Stein's avatar
William Stein committed
1781
    pos = s.position()
1782 1783 1784 1785
    kind = None
    if is_cimport and s.systring in imported_name_kinds:
        kind = s.systring
        s.next()
William Stein's avatar
William Stein committed
1786 1787
    name = p_ident(s)
    as_name = p_as_name(s)
1788
    return (pos, name, as_name, kind)
William Stein's avatar
William Stein committed
1789

1790

William Stein's avatar
William Stein committed
1791 1792 1793 1794 1795 1796 1797 1798 1799 1800
def p_dotted_name(s, as_allowed):
    pos = s.position()
    target_name = p_ident(s)
    as_name = None
    names = [target_name]
    while s.sy == '.':
        s.next()
        names.append(p_ident(s))
    if as_allowed:
        as_name = p_as_name(s)
1801 1802
    return (pos, target_name, s.context.intern_ustring(u'.'.join(names)), as_name)

William Stein's avatar
William Stein committed
1803 1804 1805 1806 1807 1808 1809 1810

def p_as_name(s):
    if s.sy == 'IDENT' and s.systring == 'as':
        s.next()
        return p_ident(s)
    else:
        return None

1811

William Stein's avatar
William Stein committed
1812 1813 1814 1815
def p_assert_statement(s):
    # s.sy == 'assert'
    pos = s.position()
    s.next()
1816
    cond = p_test(s)
William Stein's avatar
William Stein committed
1817 1818
    if s.sy == ',':
        s.next()
1819
        value = p_test(s)
William Stein's avatar
William Stein committed
1820 1821 1822 1823
    else:
        value = None
    return Nodes.AssertStatNode(pos, cond = cond, value = value)

1824

1825
statement_terminators = cython.declare(set, set([';', 'NEWLINE', 'EOF']))
William Stein's avatar
William Stein committed
1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840

def p_if_statement(s):
    # s.sy == 'if'
    pos = s.position()
    s.next()
    if_clauses = [p_if_clause(s)]
    while s.sy == 'elif':
        s.next()
        if_clauses.append(p_if_clause(s))
    else_clause = p_else_clause(s)
    return Nodes.IfStatNode(pos,
        if_clauses = if_clauses, else_clause = else_clause)

def p_if_clause(s):
    pos = s.position()
1841
    test = p_test(s)
William Stein's avatar
William Stein committed
1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856
    body = p_suite(s)
    return Nodes.IfClauseNode(pos,
        condition = test, body = body)

def p_else_clause(s):
    if s.sy == 'else':
        s.next()
        return p_suite(s)
    else:
        return None

def p_while_statement(s):
    # s.sy == 'while'
    pos = s.position()
    s.next()
1857
    test = p_test(s)
William Stein's avatar
William Stein committed
1858 1859
    body = p_suite(s)
    else_clause = p_else_clause(s)
1860 1861
    return Nodes.WhileStatNode(pos,
        condition = test, body = body,
William Stein's avatar
William Stein committed
1862 1863
        else_clause = else_clause)

1864 1865

def p_for_statement(s, is_async=False):
William Stein's avatar
William Stein committed
1866 1867 1868
    # s.sy == 'for'
    pos = s.position()
    s.next()
1869
    kw = p_for_bounds(s, allow_testlist=True, is_async=is_async)
1870 1871
    body = p_suite(s)
    else_clause = p_else_clause(s)
1872
    kw.update(body=body, else_clause=else_clause, is_async=is_async)
Robert Bradshaw's avatar
Robert Bradshaw committed
1873
    return Nodes.ForStatNode(pos, **kw)
1874

1875 1876

def p_for_bounds(s, allow_testlist=True, is_async=False):
William Stein's avatar
William Stein committed
1877 1878 1879
    target = p_for_target(s)
    if s.sy == 'in':
        s.next()
1880 1881 1882
        iterator = p_for_iterator(s, allow_testlist, is_async=is_async)
        return dict(target=target, iterator=iterator)
    elif not s.in_python_file and not is_async:
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
1883 1884 1885 1886 1887 1888
        if s.sy == 'from':
            s.next()
            bound1 = p_bit_expr(s)
        else:
            # Support shorter "for a <= x < b" syntax
            bound1, target = target, None
William Stein's avatar
William Stein committed
1889 1890 1891 1892 1893 1894
        rel1 = p_for_from_relation(s)
        name2_pos = s.position()
        name2 = p_ident(s)
        rel2_pos = s.position()
        rel2 = p_for_from_relation(s)
        bound2 = p_bit_expr(s)
1895
        step = p_for_from_step(s)
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
1896 1897 1898 1899
        if target is None:
            target = ExprNodes.NameNode(name2_pos, name = name2)
        else:
            if not target.is_name:
1900
                error(target.pos,
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
1901 1902 1903 1904
                    "Target of for-from statement must be a variable name")
            elif name2 != target.name:
                error(name2_pos,
                    "Variable name in for-from range does not match target")
Stefan Behnel's avatar
Stefan Behnel committed
1905
        if rel1[0] != rel2[0]:
William Stein's avatar
William Stein committed
1906 1907
            error(rel2_pos,
                "Relation directions in for-from do not match")
1908 1909 1910
        return dict(target = target,
                    bound1 = bound1,
                    relation1 = rel1,
1911 1912 1913 1914
                    relation2 = rel2,
                    bound2 = bound2,
                    step = step,
                    )
1915 1916 1917
    else:
        s.expect('in')
        return {}
William Stein's avatar
William Stein committed
1918 1919 1920 1921 1922 1923 1924 1925

def p_for_from_relation(s):
    if s.sy in inequality_relations:
        op = s.sy
        s.next()
        return op
    else:
        s.error("Expected one of '<', '<=', '>' '>='")
1926

1927
def p_for_from_step(s):
1928
    if s.sy == 'IDENT' and s.systring == 'by':
1929 1930 1931 1932 1933
        s.next()
        step = p_bit_expr(s)
        return step
    else:
        return None
William Stein's avatar
William Stein committed
1934

1935
inequality_relations = cython.declare(set, set(['<', '<=', '>', '>=']))
William Stein's avatar
William Stein committed
1936

1937
def p_target(s, terminator):
William Stein's avatar
William Stein committed
1938
    pos = s.position()
1939
    expr = p_starred_expr(s)
William Stein's avatar
William Stein committed
1940 1941 1942
    if s.sy == ',':
        s.next()
        exprs = [expr]
1943
        while s.sy != terminator:
1944
            exprs.append(p_starred_expr(s))
Stefan Behnel's avatar
Stefan Behnel committed
1945
            if s.sy != ',':
William Stein's avatar
William Stein committed
1946 1947 1948 1949 1950 1951
                break
            s.next()
        return ExprNodes.TupleNode(pos, args = exprs)
    else:
        return expr

1952

1953 1954 1955
def p_for_target(s):
    return p_target(s, 'in')

1956 1957

def p_for_iterator(s, allow_testlist=True, is_async=False):
William Stein's avatar
William Stein committed
1958
    pos = s.position()
1959 1960 1961 1962
    if allow_testlist:
        expr = p_testlist(s)
    else:
        expr = p_or_test(s)
1963 1964
    return (ExprNodes.AsyncIteratorNode if is_async else ExprNodes.IteratorNode)(pos, sequence=expr)

William Stein's avatar
William Stein committed
1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978

def p_try_statement(s):
    # s.sy == 'try'
    pos = s.position()
    s.next()
    body = p_suite(s)
    except_clauses = []
    else_clause = None
    if s.sy in ('except', 'else'):
        while s.sy == 'except':
            except_clauses.append(p_except_clause(s))
        if s.sy == 'else':
            s.next()
            else_clause = p_suite(s)
1979
        body = Nodes.TryExceptStatNode(pos,
William Stein's avatar
William Stein committed
1980 1981
            body = body, except_clauses = except_clauses,
            else_clause = else_clause)
1982 1983 1984 1985
        if s.sy != 'finally':
            return body
        # try-except-finally is equivalent to nested try-except/try-finally
    if s.sy == 'finally':
William Stein's avatar
William Stein committed
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
        s.next()
        finally_clause = p_suite(s)
        return Nodes.TryFinallyStatNode(pos,
            body = body, finally_clause = finally_clause)
    else:
        s.error("Expected 'except' or 'finally'")

def p_except_clause(s):
    # s.sy == 'except'
    pos = s.position()
    s.next()
    exc_type = None
    exc_value = None
1999
    is_except_as = False
Stefan Behnel's avatar
Stefan Behnel committed
2000
    if s.sy != ':':
2001
        exc_type = p_test(s)
2002 2003 2004 2005 2006
        # normalise into list of single exception tests
        if isinstance(exc_type, ExprNodes.TupleNode):
            exc_type = exc_type.args
        else:
            exc_type = [exc_type]
2007 2008
        if s.sy == ',' or (s.sy == 'IDENT' and s.systring == 'as'
                           and s.context.language_level == 2):
William Stein's avatar
William Stein committed
2009
            s.next()
2010
            exc_value = p_test(s)
2011
        elif s.sy == 'IDENT' and s.systring == 'as':
2012
            # Py3 syntax requires a name here
2013
            s.next()
2014 2015 2016
            pos2 = s.position()
            name = p_ident(s)
            exc_value = ExprNodes.NameNode(pos2, name = name)
2017
            is_except_as = True
William Stein's avatar
William Stein committed
2018 2019
    body = p_suite(s)
    return Nodes.ExceptClauseNode(pos,
2020 2021
        pattern = exc_type, target = exc_value,
        body = body, is_except_as=is_except_as)
William Stein's avatar
William Stein committed
2022

2023
def p_include_statement(s, ctx):
William Stein's avatar
William Stein committed
2024 2025
    pos = s.position()
    s.next() # 'include'
2026
    unicode_include_file_name = p_string_literal(s, 'u')[2]
William Stein's avatar
William Stein committed
2027
    s.expect_newline("Syntax error in include statement")
2028
    if s.compile_time_eval:
2029
        include_file_name = unicode_include_file_name
2030 2031
        include_file_path = s.context.find_include_file(include_file_name, pos)
        if include_file_path:
2032
            s.included_files.append(include_file_name)
2033 2034 2035
            with Utils.open_source_file(include_file_path) as f:
                source_desc = FileSourceDescriptor(include_file_path)
                s2 = PyrexScanner(f, source_desc, s, source_encoding=f.encoding, parse_comments=s.parse_comments)
2036
                tree = p_statement_list(s2, ctx)
2037 2038 2039
            return tree
        else:
            return None
William Stein's avatar
William Stein committed
2040
    else:
2041 2042
        return Nodes.PassStatNode(pos)

2043

2044
def p_with_statement(s):
2045
    s.next()  # 'with'
2046
    if s.systring == 'template' and not s.in_python_file:
2047 2048 2049 2050 2051
        node = p_with_template(s)
    else:
        node = p_with_items(s)
    return node

2052 2053

def p_with_items(s, is_async=False):
2054
    pos = s.position()
2055
    if not s.in_python_file and s.sy == 'IDENT' and s.systring in ('nogil', 'gil'):
2056 2057
        if is_async:
            s.error("with gil/nogil cannot be async")
2058 2059
        state = s.systring
        s.next()
2060
        if s.sy == ',':
Danilo Freitas's avatar
Danilo Freitas committed
2061
            s.next()
2062
            body = p_with_items(s)
Danilo Freitas's avatar
Danilo Freitas committed
2063
        else:
2064
            body = p_suite(s)
2065
        return Nodes.GILStatNode(pos, state=state, body=body)
2066
    else:
2067
        manager = p_test(s)
2068 2069 2070
        target = None
        if s.sy == 'IDENT' and s.systring == 'as':
            s.next()
2071 2072 2073
            target = p_starred_expr(s)
        if s.sy == ',':
            s.next()
2074
            body = p_with_items(s, is_async=is_async)
2075 2076
        else:
            body = p_suite(s)
2077 2078
    return Nodes.WithStatNode(pos, manager=manager, target=target, body=body, is_async=is_async)

2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103

def p_with_template(s):
    pos = s.position()
    templates = []
    s.next()
    s.expect('[')
    templates.append(s.systring)
    s.next()
    while s.systring == ',':
        s.next()
        templates.append(s.systring)
        s.next()
    s.expect(']')
    if s.sy == ':':
        s.next()
        s.expect_newline("Syntax error in template function declaration")
        s.expect_indent()
        body_ctx = Ctx()
        body_ctx.templates = templates
        func_or_var = p_c_func_or_var_declaration(s, pos, body_ctx)
        s.expect_dedent()
        return func_or_var
    else:
        error(pos, "Syntax error in template function declaration")

Stefan Behnel's avatar
Stefan Behnel committed
2104
def p_simple_statement(s, first_statement = 0):
William Stein's avatar
William Stein committed
2105 2106 2107
    #print "p_simple_statement:", s.sy, s.systring ###
    if s.sy == 'global':
        node = p_global_statement(s)
2108 2109
    elif s.sy == 'nonlocal':
        node = p_nonlocal_statement(s)
William Stein's avatar
William Stein committed
2110 2111
    elif s.sy == 'print':
        node = p_print_statement(s)
2112 2113
    elif s.sy == 'exec':
        node = p_exec_statement(s)
William Stein's avatar
William Stein committed
2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126
    elif s.sy == 'del':
        node = p_del_statement(s)
    elif s.sy == 'break':
        node = p_break_statement(s)
    elif s.sy == 'continue':
        node = p_continue_statement(s)
    elif s.sy == 'return':
        node = p_return_statement(s)
    elif s.sy == 'raise':
        node = p_raise_statement(s)
    elif s.sy in ('import', 'cimport'):
        node = p_import_statement(s)
    elif s.sy == 'from':
Stefan Behnel's avatar
Stefan Behnel committed
2127
        node = p_from_import_statement(s, first_statement = first_statement)
2128
    elif s.sy == 'yield':
2129
        node = p_yield_statement(s)
William Stein's avatar
William Stein committed
2130 2131 2132 2133 2134 2135 2136 2137
    elif s.sy == 'assert':
        node = p_assert_statement(s)
    elif s.sy == 'pass':
        node = p_pass_statement(s)
    else:
        node = p_expression_or_assignment(s)
    return node

2138
def p_simple_statement_list(s, ctx, first_statement = 0):
William Stein's avatar
William Stein committed
2139 2140
    # Parse a series of simple statements on one line
    # separated by semicolons.
Stefan Behnel's avatar
Stefan Behnel committed
2141
    stat = p_simple_statement(s, first_statement = first_statement)
2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162
    pos = stat.pos
    stats = []
    if not isinstance(stat, Nodes.PassStatNode):
        stats.append(stat)
    while s.sy == ';':
        #print "p_simple_statement_list: maybe more to follow" ###
        s.next()
        if s.sy in ('NEWLINE', 'EOF'):
            break
        stat = p_simple_statement(s, first_statement = first_statement)
        if isinstance(stat, Nodes.PassStatNode):
            continue
        stats.append(stat)
        first_statement = False

    if not stats:
        stat = Nodes.PassStatNode(pos)
    elif len(stats) == 1:
        stat = stats[0]
    else:
        stat = Nodes.StatListNode(pos, stats = stats)
2163 2164 2165 2166 2167 2168

    if s.sy not in ('NEWLINE', 'EOF'):
        # provide a better error message for users who accidentally write Cython code in .py files
        if isinstance(stat, Nodes.ExprStatNode):
            if stat.expr.is_name and stat.expr.name == 'cdef':
                s.error("The 'cdef' keyword is only allowed in Cython files (pyx/pxi/pxd)", pos)
William Stein's avatar
William Stein committed
2169
    s.expect_newline("Syntax error in simple statement list")
2170

William Stein's avatar
William Stein committed
2171 2172
    return stat

2173 2174 2175
def p_compile_time_expr(s):
    old = s.compile_time_expr
    s.compile_time_expr = 1
2176
    expr = p_testlist(s)
2177 2178 2179 2180 2181 2182 2183 2184 2185 2186
    s.compile_time_expr = old
    return expr

def p_DEF_statement(s):
    pos = s.position()
    denv = s.compile_time_env
    s.next() # 'DEF'
    name = p_ident(s)
    s.expect('=')
    expr = p_compile_time_expr(s)
2187 2188 2189 2190
    if s.compile_time_eval:
        value = expr.compile_time_value(denv)
        #print "p_DEF_statement: %s = %r" % (name, value) ###
        denv.declare(name, value)
2191
    s.expect_newline("Expected a newline", ignore_semicolon=True)
2192 2193
    return Nodes.PassStatNode(pos)

2194
def p_IF_statement(s, ctx):
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
2195
    pos = s.position()
2196 2197 2198 2199 2200 2201 2202 2203
    saved_eval = s.compile_time_eval
    current_eval = saved_eval
    denv = s.compile_time_env
    result = None
    while 1:
        s.next() # 'IF' or 'ELIF'
        expr = p_compile_time_expr(s)
        s.compile_time_eval = current_eval and bool(expr.compile_time_value(denv))
2204
        body = p_suite(s, ctx)
2205 2206 2207
        if s.compile_time_eval:
            result = body
            current_eval = 0
Stefan Behnel's avatar
Stefan Behnel committed
2208
        if s.sy != 'ELIF':
2209 2210 2211 2212
            break
    if s.sy == 'ELSE':
        s.next()
        s.compile_time_eval = current_eval
2213
        body = p_suite(s, ctx)
2214 2215 2216
        if current_eval:
            result = body
    if not result:
Stefan Behnel's avatar
Stefan Behnel committed
2217
        result = Nodes.PassStatNode(pos)
2218 2219 2220
    s.compile_time_eval = saved_eval
    return result

2221 2222
def p_statement(s, ctx, first_statement = 0):
    cdef_flag = ctx.cdef_flag
Robert Bradshaw's avatar
Robert Bradshaw committed
2223
    decorators = None
William Stein's avatar
William Stein committed
2224
    if s.sy == 'ctypedef':
2225
        if ctx.level not in ('module', 'module_pxd'):
William Stein's avatar
William Stein committed
2226
            s.error("ctypedef statement not allowed here")
2227 2228
        #if ctx.api:
        #    error(s.position(), "'api' not allowed with 'ctypedef'")
2229
        return p_ctypedef_statement(s, ctx)
2230 2231 2232
    elif s.sy == 'DEF':
        return p_DEF_statement(s)
    elif s.sy == 'IF':
2233
        return p_IF_statement(s, ctx)
2234
    elif s.sy == '@':
Haoyu Bai's avatar
Haoyu Bai committed
2235
        if ctx.level not in ('module', 'class', 'c_class', 'function', 'property', 'module_pxd', 'c_class_pxd', 'other'):
2236 2237 2238
            s.error('decorator not allowed here')
        s.level = ctx.level
        decorators = p_decorators(s)
2239 2240 2241 2242 2243
        if not ctx.allow_struct_enum_decorator and s.sy not in ('def', 'cdef', 'cpdef', 'class'):
            if s.sy == 'IDENT' and s.systring == 'async':
                pass  # handled below
            else:
                s.error("Decorators can only be followed by functions or classes")
2244 2245
    elif s.sy == 'pass' and cdef_flag:
        # empty cdef block
2246
        return p_pass_statement(s, with_newline=1)
2247 2248 2249 2250 2251

    overridable = 0
    if s.sy == 'cdef':
        cdef_flag = 1
        s.next()
2252
    elif s.sy == 'cpdef':
2253 2254 2255 2256 2257 2258 2259
        cdef_flag = 1
        overridable = 1
        s.next()
    if cdef_flag:
        if ctx.level not in ('module', 'module_pxd', 'function', 'c_class', 'c_class_pxd'):
            s.error('cdef statement not allowed here')
        s.level = ctx.level
2260
        node = p_cdef_statement(s, ctx(overridable=overridable))
2261
        if decorators is not None:
2262
            tup = (Nodes.CFuncDefNode, Nodes.CVarDefNode, Nodes.CClassDefNode)
2263
            if ctx.allow_struct_enum_decorator:
2264
                tup += (Nodes.CStructOrUnionDefNode, Nodes.CEnumDefNode)
2265
            if not isinstance(node, tup):
2266
                s.error("Decorators can only be followed by functions or classes")
2267 2268
            node.decorators = decorators
        return node
William Stein's avatar
William Stein committed
2269
    else:
2270
        if ctx.api:
2271
            s.error("'api' not allowed with this statement", fatal=False)
2272
        elif s.sy == 'def':
2273 2274 2275
            # def statements aren't allowed in pxd files, except
            # as part of a cdef class
            if ('pxd' in ctx.level) and (ctx.level != 'c_class_pxd'):
2276
                s.error('def statement not allowed here')
2277
            s.level = ctx.level
2278 2279
            return p_def_statement(s, decorators)
        elif s.sy == 'class':
2280
            if ctx.level not in ('module', 'function', 'class', 'other'):
2281
                s.error("class definition not allowed here")
2282
            return p_class_statement(s, decorators)
2283 2284 2285 2286 2287 2288 2289
        elif s.sy == 'include':
            if ctx.level not in ('module', 'module_pxd'):
                s.error("include statement not allowed here")
            return p_include_statement(s, ctx)
        elif ctx.level == 'c_class' and s.sy == 'IDENT' and s.systring == 'property':
            return p_property_decl(s)
        elif s.sy == 'pass' and ctx.level != 'property':
2290
            return p_pass_statement(s, with_newline=True)
William Stein's avatar
William Stein committed
2291
        else:
2292
            if ctx.level in ('c_class_pxd', 'property'):
2293 2294 2295
                node = p_ignorable_statement(s)
                if node is not None:
                    return node
2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306
                s.error("Executable statement not allowed here")
            if s.sy == 'if':
                return p_if_statement(s)
            elif s.sy == 'while':
                return p_while_statement(s)
            elif s.sy == 'for':
                return p_for_statement(s)
            elif s.sy == 'try':
                return p_try_statement(s)
            elif s.sy == 'with':
                return p_with_statement(s)
2307 2308 2309
            elif s.sy == 'async':
                s.next()
                return p_async_statement(s, ctx, decorators)
2310
            else:
2311
                if s.sy == 'IDENT' and s.systring == 'async':
2312
                    ident_name = s.systring
2313 2314 2315
                    # PEP 492 enables the async/await keywords when it spots "async def ..."
                    s.next()
                    if s.sy == 'def':
2316
                        return p_async_statement(s, ctx, decorators)
2317 2318
                    elif decorators:
                        s.error("Decorators can only be followed by functions or classes")
2319
                    s.put_back('IDENT', ident_name)  # re-insert original token
2320 2321
                return p_simple_statement_list(s, ctx, first_statement=first_statement)

William Stein's avatar
William Stein committed
2322

2323
def p_statement_list(s, ctx, first_statement = 0):
William Stein's avatar
William Stein committed
2324 2325 2326 2327
    # Parse a series of statements separated by newlines.
    pos = s.position()
    stats = []
    while s.sy not in ('DEDENT', 'EOF'):
2328 2329 2330 2331 2332 2333 2334 2335
        stat = p_statement(s, ctx, first_statement = first_statement)
        if isinstance(stat, Nodes.PassStatNode):
            continue
        stats.append(stat)
        first_statement = False
    if not stats:
        return Nodes.PassStatNode(pos)
    elif len(stats) == 1:
2336 2337 2338
        return stats[0]
    else:
        return Nodes.StatListNode(pos, stats = stats)
William Stein's avatar
William Stein committed
2339

2340 2341 2342 2343 2344 2345

def p_suite(s, ctx=Ctx()):
    return p_suite_with_docstring(s, ctx, with_doc_only=False)[1]


def p_suite_with_docstring(s, ctx, with_doc_only=False):
William Stein's avatar
William Stein committed
2346 2347 2348 2349 2350
    s.expect(':')
    doc = None
    if s.sy == 'NEWLINE':
        s.next()
        s.expect_indent()
2351
        if with_doc_only:
2352
            doc = p_doc_string(s)
2353
        body = p_statement_list(s, ctx)
William Stein's avatar
William Stein committed
2354 2355
        s.expect_dedent()
    else:
2356
        if ctx.api:
2357
            s.error("'api' not allowed with this statement", fatal=False)
2358 2359
        if ctx.level in ('module', 'class', 'function', 'other'):
            body = p_simple_statement_list(s, ctx)
William Stein's avatar
William Stein committed
2360 2361
        else:
            body = p_pass_statement(s)
2362
            s.expect_newline("Syntax error in declarations", ignore_semicolon=True)
2363 2364 2365 2366
    if not with_doc_only:
        doc, body = _extract_docstring(body)
    return doc, body

William Stein's avatar
William Stein committed
2367

2368
def p_positional_and_keyword_args(s, end_sy_set, templates = None):
2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381
    """
    Parses positional and keyword arguments. end_sy_set
    should contain any s.sy that terminate the argument list.
    Argument expansion (* and **) are not allowed.

    Returns: (positional_args, keyword_args)
    """
    positional_args = []
    keyword_args = []
    pos_idx = 0

    while s.sy not in end_sy_set:
        if s.sy == '*' or s.sy == '**':
2382
            s.error('Argument expansion not allowed here.', fatal=False)
2383 2384

        parsed_type = False
2385
        if s.sy == 'IDENT' and s.peek()[0] == '=':
2386
            ident = s.systring
2387
            s.next() # s.sy is '='
2388
            s.next()
2389
            if looking_at_expr(s):
2390
                arg = p_test(s)
2391 2392
            else:
                base_type = p_c_base_type(s, templates = templates)
2393
                declarator = p_c_declarator(s, empty = 1)
2394
                arg = Nodes.CComplexBaseTypeNode(base_type.pos,
2395 2396
                    base_type = base_type, declarator = declarator)
                parsed_type = True
2397
            keyword_node = ExprNodes.IdentifierStringNode(arg.pos, value=ident)
2398 2399
            keyword_args.append((keyword_node, arg))
            was_keyword = True
2400

2401
        else:
2402
            if looking_at_expr(s):
2403
                arg = p_test(s)
2404 2405
            else:
                base_type = p_c_base_type(s, templates = templates)
2406
                declarator = p_c_declarator(s, empty = 1)
2407
                arg = Nodes.CComplexBaseTypeNode(base_type.pos,
2408
                    base_type = base_type, declarator = declarator)
2409 2410 2411 2412 2413
                parsed_type = True
            positional_args.append(arg)
            pos_idx += 1
            if len(keyword_args) > 0:
                s.error("Non-keyword arg following keyword arg",
2414
                        pos=arg.pos)
2415 2416 2417 2418

        if s.sy != ',':
            if s.sy not in end_sy_set:
                if parsed_type:
2419
                    s.error("Unmatched %s" % " or ".join(end_sy_set))
2420 2421 2422 2423
            break
        s.next()
    return positional_args, keyword_args

Danilo Freitas's avatar
Danilo Freitas committed
2424
def p_c_base_type(s, self_flag = 0, nonempty = 0, templates = None):
William Stein's avatar
William Stein committed
2425 2426 2427
    # If self_flag is true, this is the base type for the
    # self argument of a C method of an extension type.
    if s.sy == '(':
2428
        return p_c_complex_base_type(s, templates = templates)
William Stein's avatar
William Stein committed
2429
    else:
Danilo Freitas's avatar
Danilo Freitas committed
2430
        return p_c_simple_base_type(s, self_flag, nonempty = nonempty, templates = templates)
William Stein's avatar
William Stein committed
2431

2432 2433 2434 2435 2436 2437 2438 2439
def p_calling_convention(s):
    if s.sy == 'IDENT' and s.systring in calling_convention_words:
        result = s.systring
        s.next()
        return result
    else:
        return ""

Stefan Behnel's avatar
Stefan Behnel committed
2440

2441 2442
calling_convention_words = cython.declare(
    set, set(["__stdcall", "__cdecl", "__fastcall"]))
2443

Stefan Behnel's avatar
Stefan Behnel committed
2444

2445
def p_c_complex_base_type(s, templates = None):
William Stein's avatar
William Stein committed
2446 2447 2448
    # s.sy == '('
    pos = s.position()
    s.next()
Stefan Behnel's avatar
Stefan Behnel committed
2449 2450 2451 2452
    base_type = p_c_base_type(s, templates=templates)
    declarator = p_c_declarator(s, empty=True)
    type_node = Nodes.CComplexBaseTypeNode(
        pos, base_type=base_type, declarator=declarator)
2453 2454 2455 2456 2457 2458
    if s.sy == ',':
        components = [type_node]
        while s.sy == ',':
            s.next()
            if s.sy == ')':
                break
Stefan Behnel's avatar
Stefan Behnel committed
2459 2460 2461 2462
            base_type = p_c_base_type(s, templates=templates)
            declarator = p_c_declarator(s, empty=True)
            components.append(Nodes.CComplexBaseTypeNode(
                pos, base_type=base_type, declarator=declarator))
2463 2464 2465
        type_node = Nodes.CTupleBaseTypeNode(pos, components = components)

    s.expect(')')
2466 2467 2468 2469 2470 2471
    if s.sy == '[':
        if is_memoryviewslice_access(s):
            type_node = p_memoryviewslice_access(s, type_node)
        else:
            type_node = p_buffer_or_template(s, type_node, templates)
    return type_node
2472

William Stein's avatar
William Stein committed
2473

Danilo Freitas's avatar
Danilo Freitas committed
2474
def p_c_simple_base_type(s, self_flag, nonempty, templates = None):
2475
    #print "p_c_simple_base_type: self_flag =", self_flag, nonempty
William Stein's avatar
William Stein committed
2476 2477 2478
    is_basic = 0
    signed = 1
    longness = 0
2479
    complex = 0
William Stein's avatar
William Stein committed
2480
    module_path = []
2481
    pos = s.position()
2482 2483
    if not s.sy == 'IDENT':
        error(pos, "Expected an identifier, found '%s'" % s.sy)
Robert Bradshaw's avatar
Robert Bradshaw committed
2484 2485
    if s.systring == 'const':
        s.next()
2486 2487 2488 2489 2490 2491
        base_type = p_c_base_type(s, self_flag=self_flag, nonempty=nonempty, templates=templates)
        if isinstance(base_type, Nodes.MemoryViewSliceTypeNode):
            # reverse order to avoid having to write "(const int)[:]"
            base_type.base_type_node = Nodes.CConstTypeNode(pos, base_type=base_type.base_type_node)
            return base_type
        return Nodes.CConstTypeNode(pos, base_type=base_type)
William Stein's avatar
William Stein committed
2492 2493 2494
    if looking_at_base_type(s):
        #print "p_c_simple_base_type: looking_at_base_type at", s.position()
        is_basic = 1
2495 2496
        if s.sy == 'IDENT' and s.systring in special_basic_c_types:
            signed, longness = special_basic_c_types[s.systring]
William Stein's avatar
William Stein committed
2497 2498 2499
            name = s.systring
            s.next()
        else:
2500 2501 2502 2503 2504
            signed, longness = p_sign_and_longness(s)
            if s.sy == 'IDENT' and s.systring in basic_c_type_names:
                name = s.systring
                s.next()
            else:
Stefan Behnel's avatar
Stefan Behnel committed
2505
                name = 'int'  # long [int], short [int], long [int] complex, etc.
2506 2507 2508
        if s.sy == 'IDENT' and s.systring == 'complex':
            complex = 1
            s.next()
2509 2510
    elif looking_at_dotted_name(s):
        #print "p_c_simple_base_type: looking_at_type_name at", s.position()
2511
        name = s.systring
2512 2513 2514 2515 2516
        s.next()
        while s.sy == '.':
            module_path.append(name)
            s.next()
            name = p_ident(s)
2517
    else:
2518 2519 2520
        name = s.systring
        s.next()
        if nonempty and s.sy != 'IDENT':
2521
            # Make sure this is not a declaration of a variable or function.
2522 2523
            if s.sy == '(':
                s.next()
2524 2525
                if (s.sy == '*' or s.sy == '**' or s.sy == '&'
                        or (s.sy == 'IDENT' and s.systring in calling_convention_words)):
2526 2527 2528 2529 2530
                    s.put_back('(', '(')
                else:
                    s.put_back('(', '(')
                    s.put_back('IDENT', name)
                    name = None
Robert Bradshaw's avatar
Robert Bradshaw committed
2531
            elif s.sy not in ('*', '**', '[', '&'):
2532 2533
                s.put_back('IDENT', name)
                name = None
Danilo Freitas's avatar
Danilo Freitas committed
2534

2535
    type_node = Nodes.CSimpleBaseTypeNode(pos,
William Stein's avatar
William Stein committed
2536 2537
        name = name, module_path = module_path,
        is_basic_c_type = is_basic, signed = signed,
2538
        complex = complex, longness = longness,
Danilo Freitas's avatar
Danilo Freitas committed
2539
        is_self_arg = self_flag, templates = templates)
William Stein's avatar
William Stein committed
2540

2541
    #    declarations here.
2542
    if s.sy == '[':
2543
        if is_memoryviewslice_access(s):
Dag Sverre Seljebotn's avatar
Dag Sverre Seljebotn committed
2544
            type_node = p_memoryviewslice_access(s, type_node)
2545 2546
        else:
            type_node = p_buffer_or_template(s, type_node, templates)
2547

Robert Bradshaw's avatar
Robert Bradshaw committed
2548 2549 2550 2551
    if s.sy == '.':
        s.next()
        name = p_ident(s)
        type_node = Nodes.CNestedBaseTypeNode(pos, base_type = type_node, name = name)
2552

Robert Bradshaw's avatar
Robert Bradshaw committed
2553
    return type_node
2554

2555
def p_buffer_or_template(s, base_type_node, templates):
2556 2557 2558
    # s.sy == '['
    pos = s.position()
    s.next()
2559 2560
    # Note that buffer_positional_options_count=1, so the only positional argument is dtype.
    # For templated types, all parameters are types.
2561
    positional_args, keyword_args = (
2562
        p_positional_and_keyword_args(s, (']',), templates)
2563 2564
    )
    s.expect(']')
2565

2566 2567
    if s.sy == '[':
        base_type_node = p_buffer_or_template(s, base_type_node, templates)
2568 2569 2570 2571 2572 2573

    keyword_dict = ExprNodes.DictNode(pos,
        key_value_pairs = [
            ExprNodes.DictItemNode(pos=key.pos, key=key, value=value)
            for key, value in keyword_args
        ])
2574
    result = Nodes.TemplatedTypeNode(pos,
2575 2576
        positional_args = positional_args,
        keyword_args = keyword_dict,
2577
        base_type_node = base_type_node)
2578
    return result
2579

2580 2581 2582 2583 2584 2585
def p_bracketed_base_type(s, base_type_node, nonempty, empty):
    # s.sy == '['
    if empty and not nonempty:
        # sizeof-like thing.  Only anonymous C arrays allowed (int[SIZE]).
        return base_type_node
    elif not empty and nonempty:
2586 2587 2588
        # declaration of either memoryview slice or buffer.
        if is_memoryviewslice_access(s):
            return p_memoryviewslice_access(s, base_type_node)
2589
        else:
Mark Florisson's avatar
Mark Florisson committed
2590 2591
            return p_buffer_or_template(s, base_type_node, None)
            # return p_buffer_access(s, base_type_node)
2592
    elif not empty and not nonempty:
2593 2594 2595 2596 2597
        # only anonymous C arrays and memoryview slice arrays here.  We
        # disallow buffer declarations for now, due to ambiguity with anonymous
        # C arrays.
        if is_memoryviewslice_access(s):
            return p_memoryviewslice_access(s, base_type_node)
2598 2599 2600
        else:
            return base_type_node

2601
def is_memoryviewslice_access(s):
2602
    # s.sy == '['
2603
    # a memoryview slice declaration is distinguishable from a buffer access
2604
    # declaration by the first entry in the bracketed list.  The buffer will
2605
    # not have an unnested colon in the first entry; the memoryview slice will.
2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616
    saved = [(s.sy, s.systring)]
    s.next()
    retval = False
    if s.systring == ':':
        retval = True
    elif s.sy == 'INT':
        saved.append((s.sy, s.systring))
        s.next()
        if s.sy == ':':
            retval = True

2617
    for sv in saved[::-1]:
2618 2619 2620 2621
        s.put_back(*sv)

    return retval

2622
def p_memoryviewslice_access(s, base_type_node):
2623 2624 2625
    # s.sy == '['
    pos = s.position()
    s.next()
2626
    subscripts, _ = p_subscript_list(s)
2627 2628 2629 2630 2631 2632
    # make sure each entry in subscripts is a slice
    for subscript in subscripts:
        if len(subscript) < 2:
            s.error("An axis specification in memoryview declaration does not have a ':'.")
    s.expect(']')
    indexes = make_slice_nodes(pos, subscripts)
2633
    result = Nodes.MemoryViewSliceTypeNode(pos,
2634 2635 2636
            base_type_node = base_type_node,
            axes = indexes)
    return result
2637

2638
def looking_at_name(s):
2639 2640
    return s.sy == 'IDENT' and not s.systring in calling_convention_words

2641
def looking_at_expr(s):
2642
    if s.systring in base_type_start_words:
2643
        return False
2644 2645 2646 2647 2648
    elif s.sy == 'IDENT':
        is_type = False
        name = s.systring
        dotted_path = []
        s.next()
2649

2650 2651 2652 2653
        while s.sy == '.':
            s.next()
            dotted_path.append(s.systring)
            s.expect('IDENT')
2654

2655
        saved = s.sy, s.systring
2656 2657 2658
        if s.sy == 'IDENT':
            is_type = True
        elif s.sy == '*' or s.sy == '**':
2659
            s.next()
2660
            is_type = s.sy in (')', ']')
2661 2662 2663 2664 2665 2666 2667 2668 2669
            s.put_back(*saved)
        elif s.sy == '(':
            s.next()
            is_type = s.sy == '*'
            s.put_back(*saved)
        elif s.sy == '[':
            s.next()
            is_type = s.sy == ']'
            s.put_back(*saved)
2670

2671 2672 2673 2674
        dotted_path.reverse()
        for p in dotted_path:
            s.put_back('IDENT', p)
            s.put_back('.', '.')
2675

2676
        s.put_back('IDENT', name)
2677
        return not is_type and saved[0]
2678
    else:
2679
        return True
William Stein's avatar
William Stein committed
2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693

def looking_at_base_type(s):
    #print "looking_at_base_type?", s.sy, s.systring, s.position()
    return s.sy == 'IDENT' and s.systring in base_type_start_words

def looking_at_dotted_name(s):
    if s.sy == 'IDENT':
        name = s.systring
        s.next()
        result = s.sy == '.'
        s.put_back('IDENT', name)
        return result
    else:
        return 0
2694

2695 2696 2697 2698 2699 2700
def looking_at_call(s):
    "See if we're looking at a.b.c("
    # Don't mess up the original position, so save and restore it.
    # Unfortunately there's no good way to handle this, as a subsequent call
    # to next() will not advance the position until it reads a new token.
    position = s.start_line, s.start_col
Mark Florisson's avatar
Mark Florisson committed
2701
    result = looking_at_expr(s) == u'('
2702 2703 2704 2705
    if not result:
        s.start_line, s.start_col = position
    return result

2706 2707
basic_c_type_names = cython.declare(
    set, set(["void", "char", "int", "float", "double", "bint"]))
2708

2709
special_basic_c_types = cython.declare(dict, {
2710
    # name : (signed, longness)
2711
    "Py_UNICODE" : (0, 0),
Stefan Behnel's avatar
Stefan Behnel committed
2712
    "Py_UCS4"    : (0, 0),
Stefan Behnel's avatar
Stefan Behnel committed
2713
    "Py_hash_t"  : (2, 0),
2714
    "Py_ssize_t" : (2, 0),
2715
    "ssize_t"    : (2, 0),
2716
    "size_t"     : (0, 0),
2717
    "ptrdiff_t"  : (2, 0),
2718
    "Py_tss_t"   : (1, 0),
2719
})
William Stein's avatar
William Stein committed
2720

2721 2722
sign_and_longness_words = cython.declare(
    set, set(["short", "long", "signed", "unsigned"]))
William Stein's avatar
William Stein committed
2723

2724 2725 2726 2727 2728
base_type_start_words = cython.declare(
    set,
    basic_c_type_names
    | sign_and_longness_words
    | set(special_basic_c_types))
William Stein's avatar
William Stein committed
2729

2730 2731
struct_enum_union = cython.declare(
    set, set(["struct", "union", "enum", "packed"]))
Mark Florisson's avatar
Mark Florisson committed
2732

William Stein's avatar
William Stein committed
2733 2734 2735 2736 2737 2738
def p_sign_and_longness(s):
    signed = 1
    longness = 0
    while s.sy == 'IDENT' and s.systring in sign_and_longness_words:
        if s.systring == 'unsigned':
            signed = 0
2739 2740
        elif s.systring == 'signed':
            signed = 2
William Stein's avatar
William Stein committed
2741 2742 2743 2744 2745 2746 2747 2748
        elif s.systring == 'short':
            longness = -1
        elif s.systring == 'long':
            longness += 1
        s.next()
    return signed, longness

def p_opt_cname(s):
2749 2750 2751
    literal = p_opt_string_literal(s, 'u')
    if literal is not None:
        cname = EncodedString(literal)
Stefan Behnel's avatar
Stefan Behnel committed
2752
        cname.encoding = s.source_encoding
William Stein's avatar
William Stein committed
2753 2754 2755 2756
    else:
        cname = None
    return cname

2757 2758 2759
def p_c_declarator(s, ctx = Ctx(), empty = 0, is_type = 0, cmethod_flag = 0,
                   assignable = 0, nonempty = 0,
                   calling_convention_allowed = 0):
2760 2761
    # If empty is true, the declarator must be empty. If nonempty is true,
    # the declarator must be nonempty. Otherwise we don't care.
William Stein's avatar
William Stein committed
2762 2763 2764
    # If cmethod_flag is true, then if this declarator declares
    # a function, it's a C method of an extension type.
    pos = s.position()
2765 2766
    if s.sy == '(':
        s.next()
2767
        if s.sy == ')' or looking_at_name(s):
2768
            base = Nodes.CNameDeclaratorNode(pos, name=s.context.intern_ustring(u""), cname=None)
2769
            result = p_c_func_declarator(s, pos, ctx, base, cmethod_flag)
2770
        else:
2771 2772 2773 2774
            result = p_c_declarator(s, ctx, empty = empty, is_type = is_type,
                                    cmethod_flag = cmethod_flag,
                                    nonempty = nonempty,
                                    calling_convention_allowed = 1)
2775 2776
            s.expect(')')
    else:
2777 2778
        result = p_c_simple_declarator(s, ctx, empty, is_type, cmethod_flag,
                                       assignable, nonempty)
Stefan Behnel's avatar
Stefan Behnel committed
2779
    if not calling_convention_allowed and result.calling_convention and s.sy != '(':
2780 2781 2782 2783 2784 2785 2786 2787
        error(s.position(), "%s on something that is not a function"
            % result.calling_convention)
    while s.sy in ('[', '('):
        pos = s.position()
        if s.sy == '[':
            result = p_c_array_declarator(s, result)
        else: # sy == '('
            s.next()
2788
            result = p_c_func_declarator(s, pos, ctx, result, cmethod_flag)
2789 2790 2791 2792 2793 2794
        cmethod_flag = 0
    return result

def p_c_array_declarator(s, base):
    pos = s.position()
    s.next() # '['
Stefan Behnel's avatar
Stefan Behnel committed
2795
    if s.sy != ']':
2796
        dim = p_testlist(s)
2797 2798 2799 2800 2801
    else:
        dim = None
    s.expect(']')
    return Nodes.CArrayDeclaratorNode(pos, base = base, dimension = dim)

root's avatar
root committed
2802
def p_c_func_declarator(s, pos, ctx, base, cmethod_flag):
2803
    #  Opening paren has already been skipped
2804 2805
    args = p_c_arg_list(s, ctx, cmethod_flag = cmethod_flag,
                        nonempty_declarators = 0)
2806 2807 2808 2809 2810
    ellipsis = p_optional_ellipsis(s)
    s.expect(')')
    nogil = p_nogil(s)
    exc_val, exc_check = p_exception_value_clause(s)
    with_gil = p_with_gil(s)
2811
    return Nodes.CFuncDeclaratorNode(pos,
2812 2813
        base = base, args = args, has_varargs = ellipsis,
        exception_value = exc_val, exception_check = exc_check,
2814
        nogil = nogil or ctx.nogil or with_gil, with_gil = with_gil)
2815

2816
supported_overloaded_operators = cython.declare(set, set([
2817
    '+', '-', '*', '/', '%',
Robert Bradshaw's avatar
Robert Bradshaw committed
2818
    '++', '--', '~', '|', '&', '^', '<<', '>>', ',',
Robert Bradshaw's avatar
Robert Bradshaw committed
2819
    '==', '!=', '>=', '>', '<=', '<',
2820 2821
    '[]', '()', '!', '=',
    'bool',
2822
]))
2823

2824 2825
def p_c_simple_declarator(s, ctx, empty, is_type, cmethod_flag,
                          assignable, nonempty):
2826 2827
    pos = s.position()
    calling_convention = p_calling_convention(s)
William Stein's avatar
William Stein committed
2828 2829
    if s.sy == '*':
        s.next()
2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842
        if s.systring == 'const':
            const_pos = s.position()
            s.next()
            const_base = p_c_declarator(s, ctx, empty = empty,
                                       is_type = is_type,
                                       cmethod_flag = cmethod_flag,
                                       assignable = assignable,
                                       nonempty = nonempty)
            base = Nodes.CConstDeclaratorNode(const_pos, base = const_base)
        else:
            base = p_c_declarator(s, ctx, empty = empty, is_type = is_type,
                                  cmethod_flag = cmethod_flag,
                                  assignable = assignable, nonempty = nonempty)
2843
        result = Nodes.CPtrDeclaratorNode(pos,
William Stein's avatar
William Stein committed
2844 2845 2846
            base = base)
    elif s.sy == '**': # scanner returns this as a single token
        s.next()
2847 2848 2849
        base = p_c_declarator(s, ctx, empty = empty, is_type = is_type,
                              cmethod_flag = cmethod_flag,
                              assignable = assignable, nonempty = nonempty)
William Stein's avatar
William Stein committed
2850 2851 2852
        result = Nodes.CPtrDeclaratorNode(pos,
            base = Nodes.CPtrDeclaratorNode(pos,
                base = base))
Robert Bradshaw's avatar
merge  
Robert Bradshaw committed
2853
    elif s.sy == '&':
2854 2855 2856 2857 2858
        s.next()
        base = p_c_declarator(s, ctx, empty = empty, is_type = is_type,
                              cmethod_flag = cmethod_flag,
                              assignable = assignable, nonempty = nonempty)
        result = Nodes.CReferenceDeclaratorNode(pos, base = base)
William Stein's avatar
William Stein committed
2859
    else:
2860 2861
        rhs = None
        if s.sy == 'IDENT':
2862
            name = s.systring
2863 2864
            if empty:
                error(s.position(), "Declarator should be empty")
William Stein's avatar
William Stein committed
2865
            s.next()
2866
            cname = p_opt_cname(s)
2867
            if name != 'operator' and s.sy == '=' and assignable:
2868
                s.next()
2869
                rhs = p_test(s)
William Stein's avatar
William Stein committed
2870
        else:
2871 2872 2873 2874
            if nonempty:
                error(s.position(), "Empty declarator")
            name = ""
            cname = None
2875
        if cname is None and ctx.namespace is not None and nonempty:
2876
            cname = ctx.namespace + "::" + name
2877
        if name == 'operator' and ctx.visibility == 'extern' and nonempty:
2878
            op = s.sy
2879
            if [1 for c in op if c in '+-*/<=>!%&|([^~,']:
2880
                s.next()
2881 2882 2883 2884 2885 2886 2887
                # Handle diphthong operators.
                if op == '(':
                    s.expect(')')
                    op = '()'
                elif op == '[':
                    s.expect(']')
                    op = '[]'
Stefan Behnel's avatar
Stefan Behnel committed
2888 2889
                elif op in ('-', '+', '|', '&') and s.sy == op:
                    op *= 2       # ++, --, ...
2890
                    s.next()
Stefan Behnel's avatar
Stefan Behnel committed
2891 2892
                elif s.sy == '=':
                    op += s.sy    # +=, -=, ...
2893 2894
                    s.next()
                if op not in supported_overloaded_operators:
2895 2896 2897
                    s.error("Overloading operator '%s' not yet supported." % op,
                            fatal=False)
                name += op
2898 2899 2900 2901 2902 2903 2904
            elif op == 'IDENT':
                op = s.systring;
                if op not in supported_overloaded_operators:
                    s.error("Overloading operator '%s' not yet supported." % op,
                            fatal=False)
                name = name + ' ' + op
                s.next()
2905
        result = Nodes.CNameDeclaratorNode(pos,
Robert Bradshaw's avatar
Robert Bradshaw committed
2906
            name = name, cname = cname, default = rhs)
2907
    result.calling_convention = calling_convention
William Stein's avatar
William Stein committed
2908 2909
    return result

2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924
def p_nogil(s):
    if s.sy == 'IDENT' and s.systring == 'nogil':
        s.next()
        return 1
    else:
        return 0

def p_with_gil(s):
    if s.sy == 'with':
        s.next()
        s.expect_keyword('gil')
        return 1
    else:
        return 0

William Stein's avatar
William Stein committed
2925 2926 2927 2928 2929 2930 2931 2932
def p_exception_value_clause(s):
    exc_val = None
    exc_check = 0
    if s.sy == 'except':
        s.next()
        if s.sy == '*':
            exc_check = 1
            s.next()
Felix Wu's avatar
Felix Wu committed
2933 2934 2935
        elif s.sy == '+':
            exc_check = '+'
            s.next()
Robert Bradshaw's avatar
Robert Bradshaw committed
2936 2937 2938 2939
            if s.sy == 'IDENT':
                name = s.systring
                s.next()
                exc_val = p_name(s, name)
2940 2941 2942
            elif s.sy == '*':
                exc_val = ExprNodes.CharNode(s.position(), value=u'*')
                s.next()
William Stein's avatar
William Stein committed
2943 2944 2945 2946
        else:
            if s.sy == '?':
                exc_check = 1
                s.next()
2947
            exc_val = p_test(s)
William Stein's avatar
William Stein committed
2948 2949
    return exc_val, exc_check

2950
c_arg_list_terminators = cython.declare(set, set(['*', '**', '.', ')', ':']))
William Stein's avatar
William Stein committed
2951

2952
def p_c_arg_list(s, ctx = Ctx(), in_pyfunc = 0, cmethod_flag = 0,
2953
                 nonempty_declarators = 0, kw_only = 0, annotated = 1):
2954 2955
    #  Comma-separated list of C argument declarations, possibly empty.
    #  May have a trailing comma.
William Stein's avatar
William Stein committed
2956
    args = []
2957 2958
    is_self_arg = cmethod_flag
    while s.sy not in c_arg_list_terminators:
2959
        args.append(p_c_arg_decl(s, ctx, in_pyfunc, is_self_arg,
2960 2961
            nonempty = nonempty_declarators, kw_only = kw_only,
            annotated = annotated))
2962 2963 2964 2965
        if s.sy != ',':
            break
        s.next()
        is_self_arg = 0
William Stein's avatar
William Stein committed
2966 2967 2968 2969 2970 2971 2972 2973 2974
    return args

def p_optional_ellipsis(s):
    if s.sy == '.':
        expect_ellipsis(s)
        return 1
    else:
        return 0

2975 2976
def p_c_arg_decl(s, ctx, in_pyfunc, cmethod_flag = 0, nonempty = 0,
                 kw_only = 0, annotated = 1):
William Stein's avatar
William Stein committed
2977
    pos = s.position()
2978
    not_none = or_none = 0
William Stein's avatar
William Stein committed
2979
    default = None
2980
    annotation = None
2981 2982 2983 2984 2985 2986 2987 2988 2989
    if s.in_python_file:
        # empty type declaration
        base_type = Nodes.CSimpleBaseTypeNode(pos,
            name = None, module_path = [],
            is_basic_c_type = 0, signed = 0,
            complex = 0, longness = 0,
            is_self_arg = cmethod_flag, templates = None)
    else:
        base_type = p_c_base_type(s, cmethod_flag, nonempty = nonempty)
2990
    declarator = p_c_declarator(s, ctx, nonempty = nonempty)
2991 2992
    if s.sy in ('not', 'or') and not s.in_python_file:
        kind = s.sy
William Stein's avatar
William Stein committed
2993 2994 2995 2996 2997 2998
        s.next()
        if s.sy == 'IDENT' and s.systring == 'None':
            s.next()
        else:
            s.error("Expected 'None'")
        if not in_pyfunc:
2999 3000 3001
            error(pos, "'%s None' only allowed in Python functions" % kind)
        or_none = kind == 'or'
        not_none = kind == 'not'
3002 3003
    if annotated and s.sy == ':':
        s.next()
3004
        annotation = p_test(s)
William Stein's avatar
William Stein committed
3005 3006
    if s.sy == '=':
        s.next()
3007 3008 3009 3010 3011 3012 3013 3014
        if 'pxd' in ctx.level:
            if s.sy in ['*', '?']:
                # TODO(github/1736): Make this an error for inline declarations.
                default = ExprNodes.NoneNode(pos)
                s.next()
            elif 'inline' in ctx.modifiers:
                default = p_test(s)
            else:
3015 3016
                error(pos, "default values cannot be specified in pxd files, use ? or *")
        else:
3017
            default = p_test(s)
William Stein's avatar
William Stein committed
3018 3019 3020 3021
    return Nodes.CArgDeclNode(pos,
        base_type = base_type,
        declarator = declarator,
        not_none = not_none,
3022
        or_none = or_none,
3023
        default = default,
3024
        annotation = annotation,
3025
        kw_only = kw_only)
William Stein's avatar
William Stein committed
3026

3027 3028 3029 3030 3031 3032 3033
def p_api(s):
    if s.sy == 'IDENT' and s.systring == 'api':
        s.next()
        return 1
    else:
        return 0

3034
def p_cdef_statement(s, ctx):
William Stein's avatar
William Stein committed
3035
    pos = s.position()
3036 3037 3038 3039 3040 3041 3042
    ctx.visibility = p_visibility(s, ctx.visibility)
    ctx.api = ctx.api or p_api(s)
    if ctx.api:
        if ctx.visibility not in ('private', 'public'):
            error(pos, "Cannot combine 'api' with '%s'" % ctx.visibility)
    if (ctx.visibility == 'extern') and s.sy == 'from':
        return p_cdef_extern_block(s, pos, ctx)
Robert Bradshaw's avatar
Robert Bradshaw committed
3043 3044
    elif s.sy == 'import':
        s.next()
3045
        return p_cdef_extern_block(s, pos, ctx)
3046
    elif p_nogil(s):
3047
        ctx.nogil = 1
3048 3049 3050 3051 3052 3053
        if ctx.overridable:
            error(pos, "cdef blocks cannot be declared cpdef")
        return p_cdef_block(s, ctx)
    elif s.sy == ':':
        if ctx.overridable:
            error(pos, "cdef blocks cannot be declared cpdef")
3054
        return p_cdef_block(s, ctx)
William Stein's avatar
William Stein committed
3055
    elif s.sy == 'class':
3056
        if ctx.level not in ('module', 'module_pxd'):
William Stein's avatar
William Stein committed
3057
            error(pos, "Extension type definition not allowed here")
3058 3059
        if ctx.overridable:
            error(pos, "Extension types cannot be declared cpdef")
3060
        return p_c_class_definition(s, pos, ctx)
Robert Bradshaw's avatar
Robert Bradshaw committed
3061 3062
    elif s.sy == 'IDENT' and s.systring == 'cppclass':
        return p_cpp_class_definition(s, pos, ctx)
Mark Florisson's avatar
Mark Florisson committed
3063
    elif s.sy == 'IDENT' and s.systring in struct_enum_union:
3064
        if ctx.level not in ('module', 'module_pxd'):
William Stein's avatar
William Stein committed
3065
            error(pos, "C struct/union/enum definition not allowed here")
3066
        if ctx.overridable:
3067 3068
            if s.systring != 'enum':
                error(pos, "C struct/union cannot be declared cpdef")
Mark Florisson's avatar
Mark Florisson committed
3069 3070 3071
        return p_struct_enum(s, pos, ctx)
    elif s.sy == 'IDENT' and s.systring == 'fused':
        return p_fused_definition(s, pos, ctx)
William Stein's avatar
William Stein committed
3072
    else:
3073
        return p_c_func_or_var_declaration(s, pos, ctx)
3074

3075 3076
def p_cdef_block(s, ctx):
    return p_suite(s, ctx(cdef_flag = 1))
William Stein's avatar
William Stein committed
3077

3078
def p_cdef_extern_block(s, pos, ctx):
3079 3080
    if ctx.overridable:
        error(pos, "cdef extern blocks cannot be declared cpdef")
William Stein's avatar
William Stein committed
3081 3082 3083 3084 3085
    include_file = None
    s.expect('from')
    if s.sy == '*':
        s.next()
    else:
3086
        include_file = p_string_literal(s, 'u')[2]
3087
    ctx = ctx(cdef_flag = 1, visibility = 'extern')
3088 3089
    if s.systring == "namespace":
        s.next()
3090
        ctx.namespace = p_string_literal(s, 'u')[2]
3091 3092
    if p_nogil(s):
        ctx.nogil = 1
3093 3094 3095 3096

    # Use "docstring" as verbatim string to include
    verbatim_include, body = p_suite_with_docstring(s, ctx, True)

William Stein's avatar
William Stein committed
3097 3098
    return Nodes.CDefExternNode(pos,
        include_file = include_file,
3099
        verbatim_include = verbatim_include,
3100
        body = body,
Robert Bradshaw's avatar
Robert Bradshaw committed
3101
        namespace = ctx.namespace)
William Stein's avatar
William Stein committed
3102

3103
def p_c_enum_definition(s, pos, ctx):
William Stein's avatar
William Stein committed
3104 3105 3106 3107 3108 3109
    # s.sy == ident 'enum'
    s.next()
    if s.sy == 'IDENT':
        name = s.systring
        s.next()
        cname = p_opt_cname(s)
3110 3111
        if cname is None and ctx.namespace is not None:
            cname = ctx.namespace + "::" + name
William Stein's avatar
William Stein committed
3112 3113 3114 3115 3116 3117
    else:
        name = None
        cname = None
    items = None
    s.expect(':')
    items = []
Stefan Behnel's avatar
Stefan Behnel committed
3118
    if s.sy != 'NEWLINE':
3119
        p_c_enum_line(s, ctx, items)
William Stein's avatar
William Stein committed
3120 3121 3122 3123
    else:
        s.next() # 'NEWLINE'
        s.expect_indent()
        while s.sy not in ('DEDENT', 'EOF'):
3124
            p_c_enum_line(s, ctx, items)
William Stein's avatar
William Stein committed
3125
        s.expect_dedent()
3126 3127 3128
    return Nodes.CEnumDefNode(
        pos, name = name, cname = cname, items = items,
        typedef_flag = ctx.typedef_flag, visibility = ctx.visibility,
3129
        create_wrapper = ctx.overridable,
3130
        api = ctx.api, in_pxd = ctx.level == 'module_pxd')
William Stein's avatar
William Stein committed
3131

3132
def p_c_enum_line(s, ctx, items):
Stefan Behnel's avatar
Stefan Behnel committed
3133
    if s.sy != 'pass':
3134
        p_c_enum_item(s, ctx, items)
William Stein's avatar
William Stein committed
3135 3136 3137 3138
        while s.sy == ',':
            s.next()
            if s.sy in ('NEWLINE', 'EOF'):
                break
3139
            p_c_enum_item(s, ctx, items)
William Stein's avatar
William Stein committed
3140 3141 3142 3143
    else:
        s.next()
    s.expect_newline("Syntax error in enum item list")

3144
def p_c_enum_item(s, ctx, items):
William Stein's avatar
William Stein committed
3145 3146 3147
    pos = s.position()
    name = p_ident(s)
    cname = p_opt_cname(s)
3148 3149
    if cname is None and ctx.namespace is not None:
        cname = ctx.namespace + "::" + name
William Stein's avatar
William Stein committed
3150 3151 3152
    value = None
    if s.sy == '=':
        s.next()
3153
        value = p_test(s)
3154
    items.append(Nodes.CEnumDefItemNode(pos,
William Stein's avatar
William Stein committed
3155 3156
        name = name, cname = cname, value = value))

3157
def p_c_struct_or_union_definition(s, pos, ctx):
3158 3159 3160 3161
    packed = False
    if s.systring == 'packed':
        packed = True
        s.next()
Dag Sverre Seljebotn's avatar
Dag Sverre Seljebotn committed
3162
        if s.sy != 'IDENT' or s.systring != 'struct':
3163
            s.expected('struct')
William Stein's avatar
William Stein committed
3164 3165 3166 3167 3168
    # s.sy == ident 'struct' or 'union'
    kind = s.systring
    s.next()
    name = p_ident(s)
    cname = p_opt_cname(s)
3169 3170
    if cname is None and ctx.namespace is not None:
        cname = ctx.namespace + "::" + name
William Stein's avatar
William Stein committed
3171 3172 3173 3174 3175 3176
    attributes = None
    if s.sy == ':':
        s.next()
        s.expect('NEWLINE')
        s.expect_indent()
        attributes = []
3177
        body_ctx = Ctx()
Stefan Behnel's avatar
Stefan Behnel committed
3178 3179
        while s.sy != 'DEDENT':
            if s.sy != 'pass':
William Stein's avatar
William Stein committed
3180
                attributes.append(
3181
                    p_c_func_or_var_declaration(s, s.position(), body_ctx))
William Stein's avatar
William Stein committed
3182 3183 3184 3185 3186 3187
            else:
                s.next()
                s.expect_newline("Expected a newline")
        s.expect_dedent()
    else:
        s.expect_newline("Syntax error in struct or union definition")
Robert Bradshaw's avatar
Robert Bradshaw committed
3188
    return Nodes.CStructOrUnionDefNode(pos,
William Stein's avatar
William Stein committed
3189
        name = name, cname = cname, kind = kind, attributes = attributes,
3190
        typedef_flag = ctx.typedef_flag, visibility = ctx.visibility,
3191
        api = ctx.api, in_pxd = ctx.level == 'module_pxd', packed = packed)
William Stein's avatar
William Stein committed
3192

Mark Florisson's avatar
Mark Florisson committed
3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232
def p_fused_definition(s, pos, ctx):
    """
    c(type)def fused my_fused_type:
        ...
    """
    # s.systring == 'fused'

    if ctx.level not in ('module', 'module_pxd'):
        error(pos, "Fused type definition not allowed here")

    s.next()
    name = p_ident(s)

    s.expect(":")
    s.expect_newline()
    s.expect_indent()

    types = []
    while s.sy != 'DEDENT':
        if s.sy != 'pass':
            #types.append(p_c_declarator(s))
            types.append(p_c_base_type(s)) #, nonempty=1))
        else:
            s.next()

        s.expect_newline()

    s.expect_dedent()

    if not types:
        error(pos, "Need at least one type")

    return Nodes.FusedTypeNode(pos, name=name, types=types)

def p_struct_enum(s, pos, ctx):
    if s.systring == 'enum':
        return p_c_enum_definition(s, pos, ctx)
    else:
        return p_c_struct_or_union_definition(s, pos, ctx)

William Stein's avatar
William Stein committed
3233 3234 3235 3236 3237
def p_visibility(s, prev_visibility):
    pos = s.position()
    visibility = prev_visibility
    if s.sy == 'IDENT' and s.systring in ('extern', 'public', 'readonly'):
        visibility = s.systring
Stefan Behnel's avatar
Stefan Behnel committed
3238
        if prev_visibility != 'private' and visibility != prev_visibility:
William Stein's avatar
William Stein committed
3239
            s.error("Conflicting visibility options '%s' and '%s'"
3240
                % (prev_visibility, visibility), fatal=False)
William Stein's avatar
William Stein committed
3241 3242
        s.next()
    return visibility
3243

3244
def p_c_modifiers(s):
3245
    if s.sy == 'IDENT' and s.systring in ('inline',):
3246
        modifier = s.systring
3247
        s.next()
3248 3249
        return [modifier] + p_c_modifiers(s)
    return []
William Stein's avatar
William Stein committed
3250

3251 3252
def p_c_func_or_var_declaration(s, pos, ctx):
    cmethod_flag = ctx.level in ('c_class', 'c_class_pxd')
3253
    modifiers = p_c_modifiers(s)
Danilo Freitas's avatar
Danilo Freitas committed
3254
    base_type = p_c_base_type(s, nonempty = 1, templates = ctx.templates)
3255
    declarator = p_c_declarator(s, ctx(modifiers=modifiers), cmethod_flag = cmethod_flag,
3256 3257
                                assignable = 1, nonempty = 1)
    declarator.overridable = ctx.overridable
Robert Bradshaw's avatar
Robert Bradshaw committed
3258 3259 3260 3261 3262
    if s.sy == 'IDENT' and s.systring == 'const' and ctx.level == 'cpp_class':
        s.next()
        is_const_method = 1
    else:
        is_const_method = 0
3263 3264 3265 3266 3267 3268 3269 3270
    if s.sy == '->':
        # Special enough to give a better error message and keep going.
        s.error(
            "Return type annotation is not allowed in cdef/cpdef signatures. "
            "Please define it before the function name, as in C signatures.",
            fatal=False)
        s.next()
        p_test(s)  # Keep going, but ignore result.
William Stein's avatar
William Stein committed
3271
    if s.sy == ':':
3272
        if ctx.level not in ('module', 'c_class', 'module_pxd', 'c_class_pxd', 'cpp_class') and not ctx.templates:
William Stein's avatar
William Stein committed
3273
            s.error("C function definition not allowed here")
3274
        doc, suite = p_suite_with_docstring(s, Ctx(level='function'))
William Stein's avatar
William Stein committed
3275
        result = Nodes.CFuncDefNode(pos,
3276
            visibility = ctx.visibility,
William Stein's avatar
William Stein committed
3277
            base_type = base_type,
3278
            declarator = declarator,
3279
            body = suite,
3280
            doc = doc,
3281
            modifiers = modifiers,
3282
            api = ctx.api,
Robert Bradshaw's avatar
Robert Bradshaw committed
3283 3284
            overridable = ctx.overridable,
            is_const_method = is_const_method)
William Stein's avatar
William Stein committed
3285
    else:
Stefan Behnel's avatar
Stefan Behnel committed
3286
        #if api:
3287
        #    s.error("'api' not allowed with variable declaration")
3288 3289
        if is_const_method:
            declarator.is_const_method = is_const_method
William Stein's avatar
William Stein committed
3290 3291 3292 3293 3294
        declarators = [declarator]
        while s.sy == ',':
            s.next()
            if s.sy == 'NEWLINE':
                break
3295 3296
            declarator = p_c_declarator(s, ctx, cmethod_flag = cmethod_flag,
                                        assignable = 1, nonempty = 1)
William Stein's avatar
William Stein committed
3297
            declarators.append(declarator)
3298
        doc_line = s.start_line + 1
3299
        s.expect_newline("Syntax error in C variable declaration", ignore_semicolon=True)
Stefan Behnel's avatar
Stefan Behnel committed
3300
        if ctx.level in ('c_class', 'c_class_pxd') and s.start_line == doc_line:
3301 3302 3303
            doc = p_doc_string(s)
        else:
            doc = None
3304
        result = Nodes.CVarDefNode(pos,
3305 3306
            visibility = ctx.visibility,
            base_type = base_type,
3307
            declarators = declarators,
3308
            in_pxd = ctx.level in ('module_pxd', 'c_class_pxd'),
3309
            doc = doc,
3310
            api = ctx.api,
3311
            modifiers = modifiers,
3312
            overridable = ctx.overridable)
William Stein's avatar
William Stein committed
3313 3314
    return result

3315
def p_ctypedef_statement(s, ctx):
William Stein's avatar
William Stein committed
3316 3317 3318
    # s.sy == 'ctypedef'
    pos = s.position()
    s.next()
3319
    visibility = p_visibility(s, ctx.visibility)
3320
    api = p_api(s)
3321
    ctx = ctx(typedef_flag = 1, visibility = visibility)
3322 3323
    if api:
        ctx.api = 1
William Stein's avatar
William Stein committed
3324
    if s.sy == 'class':
3325
        return p_c_class_definition(s, pos, ctx)
Mark Florisson's avatar
Mark Florisson committed
3326 3327 3328 3329
    elif s.sy == 'IDENT' and s.systring in struct_enum_union:
        return p_struct_enum(s, pos, ctx)
    elif s.sy == 'IDENT' and s.systring == 'fused':
        return p_fused_definition(s, pos, ctx)
William Stein's avatar
William Stein committed
3330
    else:
3331
        base_type = p_c_base_type(s, nonempty = 1)
3332
        declarator = p_c_declarator(s, ctx, is_type = 1, nonempty = 1)
3333
        s.expect_newline("Syntax error in ctypedef statement", ignore_semicolon=True)
3334 3335
        return Nodes.CTypeDefNode(
            pos, base_type = base_type,
Robert Bradshaw's avatar
Robert Bradshaw committed
3336
            declarator = declarator,
3337
            visibility = visibility, api = api,
3338
            in_pxd = ctx.level == 'module_pxd')
William Stein's avatar
William Stein committed
3339

3340 3341
def p_decorators(s):
    decorators = []
3342
    while s.sy == '@':
3343 3344
        pos = s.position()
        s.next()
3345 3346
        decstring = p_dotted_name(s, as_allowed=0)[2]
        names = decstring.split('.')
3347
        decorator = ExprNodes.NameNode(pos, name=s.context.intern_ustring(names[0]))
3348
        for name in names[1:]:
3349 3350
            decorator = ExprNodes.AttributeNode(
                pos, attribute=s.context.intern_ustring(name), obj=decorator)
3351 3352 3353 3354 3355 3356
        if s.sy == '(':
            decorator = p_call(s, decorator)
        decorators.append(Nodes.DecoratorNode(pos, decorator=decorator))
        s.expect_newline("Expected a newline after decorator")
    return decorators

3357

3358 3359 3360 3361 3362 3363 3364 3365 3366 3367
def _reject_cdef_modifier_in_py(s, name):
    """Step over incorrectly placed cdef modifiers (@see _CDEF_MODIFIERS) to provide a good error message for them.
    """
    if s.sy == 'IDENT' and name in _CDEF_MODIFIERS:
        # Special enough to provide a good error message.
        s.error("Cannot use cdef modifier '%s' in Python function signature. Use a decorator instead." % name, fatal=False)
        return p_ident(s)  # Keep going, in case there are other errors.
    return name


3368
def p_def_statement(s, decorators=None, is_async_def=False):
William Stein's avatar
William Stein committed
3369 3370
    # s.sy == 'def'
    pos = s.position()
3371 3372 3373
    # PEP 492 switches the async/await keywords on in "async def" functions
    if is_async_def:
        s.enter_async()
William Stein's avatar
William Stein committed
3374
    s.next()
3375
    name = _reject_cdef_modifier_in_py(s, p_ident(s))
3376 3377 3378 3379 3380
    s.expect(
        '(',
        "Expected '(', found '%s'. Did you use cdef syntax in a Python declaration? "
        "Use decorators and Python type annotations instead." % (
            s.systring if s.sy == 'IDENT' else s.sy))
Stefan Behnel's avatar
Stefan Behnel committed
3381 3382
    args, star_arg, starstar_arg = p_varargslist(s, terminator=')')
    s.expect(')')
3383
    _reject_cdef_modifier_in_py(s, s.systring)
3384 3385 3386
    return_type_annotation = None
    if s.sy == '->':
        s.next()
3387
        return_type_annotation = p_test(s)
3388
        _reject_cdef_modifier_in_py(s, s.systring)
3389

3390
    doc, body = p_suite_with_docstring(s, Ctx(level='function'))
3391
    if is_async_def:
3392
        s.exit_async()
3393

3394 3395 3396 3397 3398
    return Nodes.DefNode(
        pos, name=name, args=args, star_arg=star_arg, starstar_arg=starstar_arg,
        doc=doc, body=body, decorators=decorators, is_async_def=is_async_def,
        return_type_annotation=return_type_annotation)

Stefan Behnel's avatar
Stefan Behnel committed
3399

3400 3401 3402
def p_varargslist(s, terminator=')', annotated=1):
    args = p_c_arg_list(s, in_pyfunc = 1, nonempty_declarators = 1,
                        annotated = annotated)
William Stein's avatar
William Stein committed
3403 3404 3405 3406
    star_arg = None
    starstar_arg = None
    if s.sy == '*':
        s.next()
3407
        if s.sy == 'IDENT':
3408
            star_arg = p_py_arg_decl(s, annotated=annotated)
William Stein's avatar
William Stein committed
3409 3410
        if s.sy == ',':
            s.next()
3411
            args.extend(p_c_arg_list(s, in_pyfunc = 1,
3412
                nonempty_declarators = 1, kw_only = 1, annotated = annotated))
Stefan Behnel's avatar
Stefan Behnel committed
3413
        elif s.sy != terminator:
3414 3415
            s.error("Syntax error in Python function argument list")
    if s.sy == '**':
William Stein's avatar
William Stein committed
3416
        s.next()
3417
        starstar_arg = p_py_arg_decl(s, annotated=annotated)
3418 3419
    if s.sy == ',':
        s.next()
Stefan Behnel's avatar
Stefan Behnel committed
3420
    return (args, star_arg, starstar_arg)
William Stein's avatar
William Stein committed
3421

3422
def p_py_arg_decl(s, annotated = 1):
William Stein's avatar
William Stein committed
3423 3424
    pos = s.position()
    name = p_ident(s)
3425
    annotation = None
3426
    if annotated and s.sy == ':':
3427
        s.next()
3428
        annotation = p_test(s)
3429
    return Nodes.PyArgDeclNode(pos, name = name, annotation = annotation)
William Stein's avatar
William Stein committed
3430

3431

3432
def p_class_statement(s, decorators):
William Stein's avatar
William Stein committed
3433 3434 3435
    # s.sy == 'class'
    pos = s.position()
    s.next()
3436 3437
    class_name = EncodedString(p_ident(s))
    class_name.encoding = s.source_encoding  # FIXME: why is this needed?
3438 3439
    arg_tuple = None
    keyword_dict = None
William Stein's avatar
William Stein committed
3440
    if s.sy == '(':
3441 3442
        positional_args, keyword_args = p_call_parse_args(s, allow_genexp=False)
        arg_tuple, keyword_dict = p_call_build_packed_args(pos, positional_args, keyword_args)
3443 3444
    if arg_tuple is None:
        # XXX: empty arg_tuple
3445 3446
        arg_tuple = ExprNodes.TupleNode(pos, args=[])
    doc, body = p_suite_with_docstring(s, Ctx(level='class'))
3447 3448 3449 3450 3451 3452
    return Nodes.PyClassDefNode(
        pos, name=class_name,
        bases=arg_tuple,
        keyword_args=keyword_dict,
        doc=doc, body=body, decorators=decorators,
        force_py3_semantics=s.context.language_level >= 3)
William Stein's avatar
William Stein committed
3453

3454

3455
def p_c_class_definition(s, pos,  ctx):
William Stein's avatar
William Stein committed
3456 3457 3458 3459 3460 3461 3462 3463
    # s.sy == 'class'
    s.next()
    module_path = []
    class_name = p_ident(s)
    while s.sy == '.':
        s.next()
        module_path.append(class_name)
        class_name = p_ident(s)
3464
    if module_path and ctx.visibility != 'extern':
William Stein's avatar
William Stein committed
3465 3466 3467 3468 3469 3470 3471 3472
        error(pos, "Qualified class name only allowed for 'extern' C class")
    if module_path and s.sy == 'IDENT' and s.systring == 'as':
        s.next()
        as_name = p_ident(s)
    else:
        as_name = class_name
    objstruct_name = None
    typeobj_name = None
3473
    bases = None
3474
    check_size = None
William Stein's avatar
William Stein committed
3475
    if s.sy == '(':
3476 3477
        positional_args, keyword_args = p_call_parse_args(s, allow_genexp=False)
        if keyword_args:
Robert Bradshaw's avatar
Robert Bradshaw committed
3478
            s.error("C classes cannot take keyword bases.")
3479 3480
        bases, _ = p_call_build_packed_args(pos, positional_args, keyword_args)
    if bases is None:
Robert Bradshaw's avatar
Robert Bradshaw committed
3481
        bases = ExprNodes.TupleNode(pos, args=[])
3482

William Stein's avatar
William Stein committed
3483
    if s.sy == '[':
3484 3485
        if ctx.visibility not in ('public', 'extern') and not ctx.api:
            error(s.position(), "Name options only allowed for 'public', 'api', or 'extern' C class")
3486
        objstruct_name, typeobj_name, check_size = p_c_class_options(s)
William Stein's avatar
William Stein committed
3487
    if s.sy == ':':
3488
        if ctx.level == 'module_pxd':
William Stein's avatar
William Stein committed
3489 3490 3491
            body_level = 'c_class_pxd'
        else:
            body_level = 'c_class'
3492
        doc, body = p_suite_with_docstring(s, Ctx(level=body_level))
William Stein's avatar
William Stein committed
3493 3494 3495 3496
    else:
        s.expect_newline("Syntax error in C class definition")
        doc = None
        body = None
3497
    if ctx.visibility == 'extern':
William Stein's avatar
William Stein committed
3498 3499 3500 3501
        if not module_path:
            error(pos, "Module name required for 'extern' C class")
        if typeobj_name:
            error(pos, "Type object name specification not allowed for 'extern' C class")
3502
    elif ctx.visibility == 'public':
William Stein's avatar
William Stein committed
3503 3504 3505 3506
        if not objstruct_name:
            error(pos, "Object struct name specification required for 'public' C class")
        if not typeobj_name:
            error(pos, "Type object name specification required for 'public' C class")
3507 3508
    elif ctx.visibility == 'private':
        if ctx.api:
3509 3510 3511 3512
            if not objstruct_name:
                error(pos, "Object struct name specification required for 'api' C class")
            if not typeobj_name:
                error(pos, "Type object name specification required for 'api' C class")
3513
    else:
3514
        error(pos, "Invalid class visibility '%s'" % ctx.visibility)
William Stein's avatar
William Stein committed
3515
    return Nodes.CClassDefNode(pos,
3516 3517 3518
        visibility = ctx.visibility,
        typedef_flag = ctx.typedef_flag,
        api = ctx.api,
William Stein's avatar
William Stein committed
3519 3520 3521
        module_name = ".".join(module_path),
        class_name = class_name,
        as_name = as_name,
3522
        bases = bases,
William Stein's avatar
William Stein committed
3523 3524
        objstruct_name = objstruct_name,
        typeobj_name = typeobj_name,
3525
        check_size = check_size,
3526
        in_pxd = ctx.level == 'module_pxd',
William Stein's avatar
William Stein committed
3527 3528 3529
        doc = doc,
        body = body)

3530

William Stein's avatar
William Stein committed
3531 3532 3533
def p_c_class_options(s):
    objstruct_name = None
    typeobj_name = None
3534
    check_size = None
William Stein's avatar
William Stein committed
3535 3536
    s.expect('[')
    while 1:
Stefan Behnel's avatar
Stefan Behnel committed
3537
        if s.sy != 'IDENT':
William Stein's avatar
William Stein committed
3538 3539 3540 3541 3542 3543 3544
            break
        if s.systring == 'object':
            s.next()
            objstruct_name = p_ident(s)
        elif s.systring == 'type':
            s.next()
            typeobj_name = p_ident(s)
3545 3546
        elif s.systring == 'check_size':
            s.next()
mattip's avatar
mattip committed
3547
            check_size = p_ident(s)
3548 3549
            if check_size not in ('ignore', 'warn', 'error'):
                s.error("Expected one of ignore, warn or error, found %r" % check_size)
Stefan Behnel's avatar
Stefan Behnel committed
3550
        if s.sy != ',':
William Stein's avatar
William Stein committed
3551 3552
            break
        s.next()
3553
    s.expect(']', "Expected 'object', 'type' or 'check_size'")
3554
    return objstruct_name, typeobj_name, check_size
William Stein's avatar
William Stein committed
3555

3556

William Stein's avatar
William Stein committed
3557 3558
def p_property_decl(s):
    pos = s.position()
3559
    s.next()  # 'property'
William Stein's avatar
William Stein committed
3560
    name = p_ident(s)
3561 3562 3563 3564
    doc, body = p_suite_with_docstring(
        s, Ctx(level='property'), with_doc_only=True)
    return Nodes.PropertyNode(pos, name=name, doc=doc, body=body)

William Stein's avatar
William Stein committed
3565

3566 3567 3568 3569 3570 3571 3572
def p_ignorable_statement(s):
    """
    Parses any kind of ignorable statement that is allowed in .pxd files.
    """
    if s.sy == 'BEGIN_STRING':
        pos = s.position()
        string_node = p_atom(s)
3573
        s.expect_newline("Syntax error in string", ignore_semicolon=True)
3574 3575 3576 3577
        return Nodes.ExprStatNode(pos, expr=string_node)
    return None


3578 3579 3580 3581
def p_doc_string(s):
    if s.sy == 'BEGIN_STRING':
        pos = s.position()
        kind, bytes_result, unicode_result = p_cat_string_literal(s)
3582
        s.expect_newline("Syntax error in doc string", ignore_semicolon=True)
3583 3584 3585 3586 3587 3588 3589
        if kind in ('u', ''):
            return unicode_result
        warning(pos, "Python 3 requires docstrings to be unicode strings")
        return bytes_result
    else:
        return None

3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625

def _extract_docstring(node):
    """
    Extract a docstring from a statement or from the first statement
    in a list.  Remove the statement if found.  Return a tuple
    (plain-docstring or None, node).
    """
    doc_node = None
    if node is None:
        pass
    elif isinstance(node, Nodes.ExprStatNode):
        if node.expr.is_string_literal:
            doc_node = node.expr
            node = Nodes.StatListNode(node.pos, stats=[])
    elif isinstance(node, Nodes.StatListNode) and node.stats:
        stats = node.stats
        if isinstance(stats[0], Nodes.ExprStatNode):
            if stats[0].expr.is_string_literal:
                doc_node = stats[0].expr
                del stats[0]

    if doc_node is None:
        doc = None
    elif isinstance(doc_node, ExprNodes.BytesNode):
        warning(node.pos,
                "Python 3 requires docstrings to be unicode strings")
        doc = doc_node.value
    elif isinstance(doc_node, ExprNodes.StringNode):
        doc = doc_node.unicode_value
        if doc is None:
            doc = doc_node.value
    else:
        doc = doc_node.value
    return doc, node


3626 3627
def p_code(s, level=None, ctx=Ctx):
    body = p_statement_list(s, ctx(level = level), first_statement = 1)
3628 3629 3630 3631
    if s.sy != 'EOF':
        s.error("Syntax error in statement [%s,%s]" % (
            repr(s.sy), repr(s.systring)))
    return body
William Stein's avatar
William Stein committed
3632

3633

3634 3635
_match_compiler_directive_comment = cython.declare(object, re.compile(
    r"^#\s*cython\s*:\s*((\w|[.])+\s*=.*)$").match)
3636

3637

3638
def p_compiler_directive_comments(s):
3639
    result = {}
3640
    while s.sy == 'commentline':
3641
        pos = s.position()
3642
        m = _match_compiler_directive_comment(s.systring)
3643
        if m:
3644
            directives_string = m.group(1).strip()
3645
            try:
3646
                new_directives = Options.parse_directive_list(directives_string, ignore_unknown=True)
3647
            except ValueError as e:
3648
                s.error(e.args[0], fatal=False)
3649
                s.next()
3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666
                continue

            for name in new_directives:
                if name not in result:
                    pass
                elif new_directives[name] == result[name]:
                    warning(pos, "Duplicate directive found: %s" % (name,))
                else:
                    s.error("Conflicting settings found for top-level directive %s: %r and %r" % (
                        name, result[name], new_directives[name]), pos=pos)

            if 'language_level' in new_directives:
                # Make sure we apply the language level already to the first token that follows the comments.
                s.context.set_language_level(new_directives['language_level'])

            result.update(new_directives)

3667 3668 3669
        s.next()
    return result

3670

3671
def p_module(s, pxd, full_module_name, ctx=Ctx):
William Stein's avatar
William Stein committed
3672
    pos = s.position()
3673

3674
    directive_comments = p_compiler_directive_comments(s)
3675 3676
    s.parse_comments = False

3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687
    if s.context.language_level is None:
        s.context.set_language_level(2)
        if pos[0].filename:
            import warnings
            warnings.warn(
                "Cython directive 'language_level' not set, using 2 for now (Py2). "
                "This will change in a later release! File: %s" % pos[0].filename,
                FutureWarning,
                stacklevel=1 if cython.compiled else 2,
            )

3688
    doc = p_doc_string(s)
William Stein's avatar
William Stein committed
3689 3690 3691 3692
    if pxd:
        level = 'module_pxd'
    else:
        level = 'module'
3693

3694
    body = p_statement_list(s, ctx(level=level), first_statement = 1)
Stefan Behnel's avatar
Stefan Behnel committed
3695
    if s.sy != 'EOF':
William Stein's avatar
William Stein committed
3696 3697
        s.error("Syntax error in statement [%s,%s]" % (
            repr(s.sy), repr(s.systring)))
3698 3699
    return ModuleNode(pos, doc = doc, body = body,
                      full_module_name = full_module_name,
3700
                      directive_comments = directive_comments)
William Stein's avatar
William Stein committed
3701

3702 3703 3704 3705 3706 3707 3708 3709 3710 3711
def p_template_definition(s):
    name = p_ident(s)
    if s.sy == '=':
        s.expect('=')
        s.expect('*')
        required = False
    else:
        required = True
    return name, required

3712 3713 3714 3715 3716
def p_cpp_class_definition(s, pos,  ctx):
    # s.sy == 'cppclass'
    s.next()
    module_path = []
    class_name = p_ident(s)
3717 3718 3719
    cname = p_opt_cname(s)
    if cname is None and ctx.namespace is not None:
        cname = ctx.namespace + "::" + class_name
3720
    if s.sy == '.':
3721
        error(pos, "Qualified class name not allowed C++ class")
Danilo Freitas's avatar
Danilo Freitas committed
3722 3723
    if s.sy == '[':
        s.next()
3724
        templates = [p_template_definition(s)]
Danilo Freitas's avatar
Danilo Freitas committed
3725 3726
        while s.sy == ',':
            s.next()
3727
            templates.append(p_template_definition(s))
Danilo Freitas's avatar
Danilo Freitas committed
3728
        s.expect(']')
3729
        template_names = [name for name, required in templates]
3730 3731
    else:
        templates = None
3732
        template_names = None
3733
    if s.sy == '(':
3734
        s.next()
3735
        base_classes = [p_c_base_type(s, templates = template_names)]
3736
        while s.sy == ',':
3737
            s.next()
3738
            base_classes.append(p_c_base_type(s, templates = template_names))
3739
        s.expect(')')
3740 3741
    else:
        base_classes = []
3742
    if s.sy == '[':
3743
        error(s.position(), "Name options not allowed for C++ class")
3744
    nogil = p_nogil(s)
3745
    if s.sy == ':':
3746 3747 3748 3749
        s.next()
        s.expect('NEWLINE')
        s.expect_indent()
        attributes = []
3750
        body_ctx = Ctx(visibility = ctx.visibility, level='cpp_class', nogil=nogil or ctx.nogil)
3751
        body_ctx.templates = template_names
3752
        while s.sy != 'DEDENT':
3753 3754
            if s.sy != 'pass':
                attributes.append(p_cpp_class_attribute(s, body_ctx))
3755 3756 3757 3758
            else:
                s.next()
                s.expect_newline("Expected a newline")
        s.expect_dedent()
3759
    else:
3760
        attributes = None
3761 3762 3763
        s.expect_newline("Syntax error in C++ class definition")
    return Nodes.CppClassNode(pos,
        name = class_name,
3764
        cname = cname,
3765
        base_classes = base_classes,
3766 3767
        visibility = ctx.visibility,
        in_pxd = ctx.level == 'module_pxd',
Danilo Freitas's avatar
Danilo Freitas committed
3768 3769
        attributes = attributes,
        templates = templates)
3770

3771 3772 3773 3774 3775 3776
def p_cpp_class_attribute(s, ctx):
    decorators = None
    if s.sy == '@':
        decorators = p_decorators(s)
    if s.systring == 'cppclass':
        return p_cpp_class_definition(s, s.position(), ctx)
3777
    elif s.systring == 'ctypedef':
3778
        return p_ctypedef_statement(s, ctx)
3779
    elif s.sy == 'IDENT' and s.systring in struct_enum_union:
3780 3781 3782 3783
        if s.systring != 'enum':
            return p_cpp_class_definition(s, s.position(), ctx)
        else:
            return p_struct_enum(s, s.position(), ctx)
3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794
    else:
        node = p_c_func_or_var_declaration(s, s.position(), ctx)
        if decorators is not None:
            tup = Nodes.CFuncDefNode, Nodes.CVarDefNode, Nodes.CClassDefNode
            if ctx.allow_struct_enum_decorator:
                tup += Nodes.CStructOrUnionDefNode, Nodes.CEnumDefNode
            if not isinstance(node, tup):
                s.error("Decorators can only be followed by functions or classes")
            node.decorators = decorators
        return node

3795

William Stein's avatar
William Stein committed
3796 3797 3798 3799 3800 3801
#----------------------------------------------
#
#   Debugging
#
#----------------------------------------------

Stefan Behnel's avatar
Stefan Behnel committed
3802
def print_parse_tree(f, node, level, key = None):
William Stein's avatar
William Stein committed
3803 3804 3805 3806 3807 3808
    ind = "  " * level
    if node:
        f.write(ind)
        if key:
            f.write("%s: " % key)
        t = type(node)
Stefan Behnel's avatar
Stefan Behnel committed
3809
        if t is tuple:
William Stein's avatar
William Stein committed
3810
            f.write("(%s @ %s\n" % (node[0], node[1]))
3811
            for i in range(2, len(node)):
William Stein's avatar
William Stein committed
3812 3813 3814
                print_parse_tree(f, node[i], level+1)
            f.write("%s)\n" % ind)
            return
3815
        elif isinstance(node, Nodes.Node):
William Stein's avatar
William Stein committed
3816 3817 3818 3819 3820 3821
            try:
                tag = node.tag
            except AttributeError:
                tag = node.__class__.__name__
            f.write("%s @ %s\n" % (tag, node.pos))
            for name, value in node.__dict__.items():
Stefan Behnel's avatar
Stefan Behnel committed
3822
                if name != 'tag' and name != 'pos':
William Stein's avatar
William Stein committed
3823 3824
                    print_parse_tree(f, value, level+1, name)
            return
Stefan Behnel's avatar
Stefan Behnel committed
3825
        elif t is list:
William Stein's avatar
William Stein committed
3826
            f.write("[\n")
3827
            for i in range(len(node)):
William Stein's avatar
William Stein committed
3828 3829 3830 3831
                print_parse_tree(f, node[i], level+1)
            f.write("%s]\n" % ind)
            return
    f.write("%s%s\n" % (ind, node))