Parsing.py 120 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,
13
               Future=object, Options=object, error=object, warning=object,
14
               Builtin=object, ModuleNode=object, Utils=object,
15
               re=object, _unicode=object, _bytes=object,
16
               partial=object, reduce=object, _IS_PY3=cython.bint)
17

18
from io import StringIO
19
import re
20
import sys
21
from unicodedata import lookup as lookup_unicodechar
22
from functools import partial, reduce
Lisandro Dalcin's avatar
Lisandro Dalcin committed
23

24
from .Scanning import PyrexScanner, FileSourceDescriptor, StringSourceDescriptor
25 26 27 28
from . import Nodes
from . import ExprNodes
from . import Builtin
from . import StringEncoding
29
from .StringEncoding import EncodedString, bytes_literal, _unicode, _bytes
30 31 32 33 34 35
from .ModuleNode import ModuleNode
from .Errors import error, warning
from .. import Utils
from . import Future
from . import Options

36 37
_IS_PY3 = sys.version_info[0] >= 3

William Stein's avatar
William Stein committed
38

39 40 41 42 43 44 45 46 47
class Ctx(object):
    #  Parsing context
    level = 'other'
    visibility = 'private'
    cdef_flag = 0
    typedef_flag = 0
    api = 0
    overridable = 0
    nogil = 0
48
    namespace = None
Danilo Freitas's avatar
Danilo Freitas committed
49
    templates = None
50
    allow_struct_enum_decorator = False
51 52 53 54 55 56 57 58 59 60 61

    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

62

63
def p_ident(s, message="Expected an identifier"):
William Stein's avatar
William Stein committed
64 65 66 67 68 69 70 71 72 73 74 75
    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
76
        if s.sy != ',':
William Stein's avatar
William Stein committed
77 78 79 80 81 82 83 84 85 86
            break
        s.next()
    return names

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

87 88 89 90 91 92
def p_binop_operator(s):
    pos = s.position()
    op = s.sy
    s.next()
    return op, pos

William Stein's avatar
William Stein committed
93 94 95
def p_binop_expr(s, ops, p_sub_expr):
    n1 = p_sub_expr(s)
    while s.sy in ops:
96
        op, pos = p_binop_operator(s)
William Stein's avatar
William Stein committed
97 98
        n2 = p_sub_expr(s)
        n1 = ExprNodes.binop_node(pos, op, n1, n2)
99 100 101 102 103
        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
104 105
    return n1

Stefan Behnel's avatar
Stefan Behnel committed
106 107 108 109 110 111 112 113 114 115
#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:
116 117
        args, star_arg, starstar_arg = p_varargslist(
            s, terminator=':', annotated=False)
Stefan Behnel's avatar
Stefan Behnel committed
118 119
    s.expect(':')
    if allow_conditional:
120
        expr = p_test(s)
Stefan Behnel's avatar
Stefan Behnel committed
121 122 123 124 125 126 127 128 129 130 131 132
    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)

133
#test: or_test ['if' or_test 'else' test] | lambdef
William Stein's avatar
William Stein committed
134

Robert Bradshaw's avatar
Robert Bradshaw committed
135
def p_test(s):
136 137
    if s.sy == 'lambda':
        return p_lambdef(s)
Robert Bradshaw's avatar
Robert Bradshaw committed
138 139 140 141 142
    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
143 144 145
        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
146 147 148
    else:
        return expr

Stefan Behnel's avatar
Stefan Behnel committed
149 150 151 152 153 154 155
#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
156 157 158 159

#or_test: and_test ('or' and_test)*

def p_or_test(s):
William Stein's avatar
William Stein committed
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
    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):
192
    n1 = p_starred_expr(s)
William Stein's avatar
William Stein committed
193 194 195
    if s.sy in comparison_ops:
        pos = s.position()
        op = p_cmp_op(s)
196
        n2 = p_starred_expr(s)
197
        n1 = ExprNodes.PrimaryCmpNode(pos,
William Stein's avatar
William Stein committed
198 199 200 201 202
            operator = op, operand1 = n1, operand2 = n2)
        if s.sy in comparison_ops:
            n1.cascade = p_cascaded_cmp(s)
    return n1

203 204 205 206 207 208
def p_test_or_starred_expr(s):
    if s.sy == '*':
        return p_starred_expr(s)
    else:
        return p_test(s)

209
def p_starred_expr(s):
210
    pos = s.position()
211 212 213 214 215 216
    if s.sy == '*':
        starred = True
        s.next()
    else:
        starred = False
    expr = p_bit_expr(s)
217
    if starred:
218
        expr = ExprNodes.StarredUnpackingNode(pos, expr)
219 220
    return expr

William Stein's avatar
William Stein committed
221 222 223
def p_cascaded_cmp(s):
    pos = s.position()
    op = p_cmp_op(s)
224
    n2 = p_starred_expr(s)
225
    result = ExprNodes.CascadedCmpNode(pos,
William Stein's avatar
William Stein committed
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248
        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
249

250
comparison_ops = cython.declare(set, set([
251
    '<', '>', '==', '>=', '<=', '<>', '!=',
William Stein's avatar
William Stein committed
252
    'in', 'is', 'not'
253
]))
William Stein's avatar
William Stein committed
254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279

#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)

280
#term: factor (('*'|'@'|'/'|'%'|'//') factor)*
William Stein's avatar
William Stein committed
281 282

def p_term(s):
283
    return p_binop_expr(s, ('*', '@', '/', '%', '//'), p_factor)
William Stein's avatar
William Stein committed
284 285 286 287

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

def p_factor(s):
288 289 290 291
    # little indirection for C-ification purposes
    return _p_factor(s)

def _p_factor(s):
William Stein's avatar
William Stein committed
292 293 294 295 296 297
    sy = s.sy
    if sy in ('+', '-', '~'):
        op = s.sy
        pos = s.position()
        s.next()
        return ExprNodes.unop_node(pos, op, p_factor(s))
298 299 300 301 302 303 304 305 306 307 308
    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
309 310 311 312 313 314

def p_typecast(s):
    # s.sy == "<"
    pos = s.position()
    s.next()
    base_type = p_c_base_type(s)
315
    is_memslice = isinstance(base_type, Nodes.MemoryViewSliceTypeNode)
Robert Bradshaw's avatar
Robert Bradshaw committed
316 317 318 319
    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):
320
        s.error("Unknown type")
William Stein's avatar
William Stein committed
321
    declarator = p_c_declarator(s, empty = 1)
322 323 324 325 326
    if s.sy == '?':
        s.next()
        typecheck = 1
    else:
        typecheck = 0
William Stein's avatar
William Stein committed
327 328
    s.expect(">")
    operand = p_factor(s)
329 330 331 332
    if is_memslice:
        return ExprNodes.CythonArrayNode(pos, base_type_node=base_type,
                                         operand=operand)

333 334
    return ExprNodes.TypecastNode(pos,
        base_type = base_type,
William Stein's avatar
William Stein committed
335
        declarator = declarator,
336 337
        operand = operand,
        typecheck = typecheck)
William Stein's avatar
William Stein committed
338 339 340 341 342 343

def p_sizeof(s):
    # s.sy == ident "sizeof"
    pos = s.position()
    s.next()
    s.expect('(')
344
    # Here we decide if we are looking at an expression or type
345 346
    # If it is actually a type, but parsable as an expression,
    # we treat it as an expression here.
347
    if looking_at_expr(s):
348
        operand = p_test(s)
349 350
        node = ExprNodes.SizeofVarNode(pos, operand = operand)
    else:
William Stein's avatar
William Stein committed
351 352
        base_type = p_c_base_type(s)
        declarator = p_c_declarator(s, empty = 1)
353
        node = ExprNodes.SizeofTypeNode(pos,
William Stein's avatar
William Stein committed
354 355 356 357
            base_type = base_type, declarator = declarator)
    s.expect(')')
    return node

358

359 360 361 362
def p_yield_expression(s):
    # s.sy == "yield"
    pos = s.position()
    s.next()
363 364 365 366
    is_yield_from = False
    if s.sy == 'from':
        is_yield_from = True
        s.next()
367
    if s.sy != ')' and s.sy not in statement_terminators:
368 369
        # "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)
370
    else:
371
        if is_yield_from:
372 373
            s.error("'yield from' requires a source argument",
                    pos=pos, fatal=False)
374
        arg = None
375 376 377 378
    if is_yield_from:
        return ExprNodes.YieldFromExprNode(pos, arg=arg)
    else:
        return ExprNodes.YieldExprNode(pos, arg=arg)
379

380

381 382 383 384
def p_yield_statement(s):
    # s.sy == "yield"
    yield_expr = p_yield_expression(s)
    return Nodes.ExprStatNode(yield_expr.pos, expr=yield_expr)
385

386 387 388 389 390 391 392 393 394 395 396 397

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':
398
        return p_for_statement(s, is_async=True)
399
    elif s.sy == 'with':
400 401
        s.next()
        return p_with_items(s, is_async=True)
402 403 404 405 406 407
    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
408 409

def p_power(s):
410
    if s.systring == 'new' and s.peek()[0] == 'IDENT':
Danilo Freitas's avatar
Danilo Freitas committed
411
        return p_new_expr(s)
412 413 414 415
    await_pos = None
    if s.sy == 'await':
        await_pos = s.position()
        s.next()
William Stein's avatar
William Stein committed
416 417 418
    n1 = p_atom(s)
    while s.sy in ('(', '[', '.'):
        n1 = p_trailer(s, n1)
419 420
    if await_pos:
        n1 = ExprNodes.AwaitExprNode(await_pos, arg=n1)
William Stein's avatar
William Stein committed
421 422 423 424 425 426 427
    if s.sy == '**':
        pos = s.position()
        s.next()
        n2 = p_factor(s)
        n1 = ExprNodes.binop_node(pos, '**', n1, n2)
    return n1

428

Danilo Freitas's avatar
Danilo Freitas committed
429
def p_new_expr(s):
Danilo Freitas's avatar
Danilo Freitas committed
430
    # s.systring == 'new'.
Danilo Freitas's avatar
Danilo Freitas committed
431 432
    pos = s.position()
    s.next()
Robert Bradshaw's avatar
Robert Bradshaw committed
433 434
    cppclass = p_c_base_type(s)
    return p_call(s, ExprNodes.NewExprNode(pos, cppclass = cppclass))
Danilo Freitas's avatar
Danilo Freitas committed
435

William Stein's avatar
William Stein committed
436 437 438 439 440 441 442 443 444 445
#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()
446
        name = p_ident(s)
447
        return ExprNodes.AttributeNode(pos,
448
            obj=node1, attribute=name)
William Stein's avatar
William Stein committed
449

450

William Stein's avatar
William Stein committed
451 452 453
# arglist:  argument (',' argument)* [',']
# argument: [test '='] test       # Really [keyword '='] test

454 455 456 457 458 459 460
# 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
461 462 463 464 465
    # s.sy == '('
    pos = s.position()
    s.next()
    positional_args = []
    keyword_args = []
466 467 468
    starstar_seen = False
    last_was_tuple_unpack = False
    while s.sy != ')':
469
        if s.sy == '*':
470 471 472 473 474 475
            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
476
            s.next()
477 478
            keyword_args.append(p_test(s))
            starstar_seen = True
William Stein's avatar
William Stein committed
479
        else:
480
            arg = p_test(s)
481 482 483 484
            if s.sy == '=':
                s.next()
                if not arg.is_name:
                    s.error("Expected an identifier before '='",
485
                            pos=arg.pos)
486
                encoded_name = s.context.intern_ustring(arg.name)
487 488
                keyword = ExprNodes.IdentifierStringNode(
                    arg.pos, value=encoded_name)
489
                arg = p_test(s)
490 491 492
                keyword_args.append((keyword, arg))
            else:
                if keyword_args:
493 494 495 496 497 498
                    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
499
        if s.sy != ',':
William Stein's avatar
William Stein committed
500 501
            break
        s.next()
502

503
    if s.sy == 'for':
504 505
        if not keyword_args and not last_was_tuple_unpack:
            if len(positional_args) == 1 and len(positional_args[0]) == 1:
506
                positional_args = [[p_genexp(s, positional_args[0][0])]]
William Stein's avatar
William Stein committed
507
    s.expect(')')
508
    return positional_args or [[]], keyword_args
509

510 511

def p_call_build_packed_args(pos, positional_args, keyword_args):
512
    keyword_dict = None
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547

    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
548
                keyword_dict = ExprNodes.MergedDictNode(pos, keyword_args=kwargs)
549

550 551
    return arg_tuple, keyword_dict

552

553 554 555
def p_call(s, function):
    # s.sy == '('
    pos = s.position()
556
    positional_args, keyword_args = p_call_parse_args(s)
557

558 559
    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
560
    else:
561 562 563 564
        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
565 566 567 568 569 570 571 572 573

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

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

def p_index(s, base):
    # s.sy == '['
    pos = s.position()
    s.next()
574 575
    subscripts, is_single_value = p_subscript_list(s)
    if is_single_value and len(subscripts[0]) == 2:
William Stein's avatar
William Stein committed
576
        start, stop = subscripts[0]
577
        result = ExprNodes.SliceIndexNode(pos,
William Stein's avatar
William Stein committed
578 579 580
            base = base, start = start, stop = stop)
    else:
        indexes = make_slice_nodes(pos, subscripts)
581
        if is_single_value:
William Stein's avatar
William Stein committed
582 583 584 585 586 587 588 589 590
            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):
591
    is_single_value = True
William Stein's avatar
William Stein committed
592 593
    items = [p_subscript(s)]
    while s.sy == ',':
594
        is_single_value = False
William Stein's avatar
William Stein committed
595 596 597 598
        s.next()
        if s.sy == ']':
            break
        items.append(p_subscript(s))
599
    return items, is_single_value
William Stein's avatar
William Stein committed
600 601 602 603 604 605 606 607

#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()
608 609 610 611 612 613 614 615 616 617
    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
618 619 620 621 622

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:
623
        return p_test(s)
William Stein's avatar
William Stein committed
624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
    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)

655
#atom: '(' [yield_expr|testlist_comp] ')' | '[' [listmaker] ']' | '{' [dict_or_set_maker] '}' | '`' testlist '`' | NAME | NUMBER | STRING+
William Stein's avatar
William Stein committed
656 657 658 659 660 661 662 663

def p_atom(s):
    pos = s.position()
    sy = s.sy
    if sy == '(':
        s.next()
        if s.sy == ')':
            result = ExprNodes.TupleNode(pos, args = [])
664 665
        elif s.sy == 'yield':
            result = p_yield_expression(s)
William Stein's avatar
William Stein committed
666
        else:
667
            result = p_testlist_comp(s)
William Stein's avatar
William Stein committed
668 669 670 671 672
        s.expect(')')
        return result
    elif sy == '[':
        return p_list_maker(s)
    elif sy == '{':
673
        return p_dict_or_set_maker(s)
William Stein's avatar
William Stein committed
674 675
    elif sy == '`':
        return p_backquote_expr(s)
676 677 678
    elif sy == '.':
        expect_ellipsis(s)
        return ExprNodes.EllipsisNode(pos)
William Stein's avatar
William Stein committed
679
    elif sy == 'INT':
680
        return p_int_literal(s)
William Stein's avatar
William Stein committed
681 682 683 684 685 686 687 688
    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)
689
    elif sy == 'BEGIN_STRING':
690
        kind, bytes_value, unicode_value = p_cat_string_literal(s)
William Stein's avatar
William Stein committed
691
        if kind == 'c':
692
            return ExprNodes.CharNode(pos, value = bytes_value)
693
        elif kind == 'u':
694
            return ExprNodes.UnicodeNode(pos, value = unicode_value, bytes_value = bytes_value)
695
        elif kind == 'b':
696
            return ExprNodes.BytesNode(pos, value = bytes_value)
697 698 699
        elif kind == 'f':
            return ExprNodes.JoinedStrNode(pos, values = unicode_value)
        elif kind == '':
700
            return ExprNodes.StringNode(pos, value = bytes_value, unicode_value = unicode_value)
701 702
        else:
            s.error("invalid string kind '%s'" % kind)
William Stein's avatar
William Stein committed
703
    elif sy == 'IDENT':
704
        name = s.systring
William Stein's avatar
William Stein committed
705 706 707
        s.next()
        if name == "None":
            return ExprNodes.NoneNode(pos)
708
        elif name == "True":
709
            return ExprNodes.BoolNode(pos, value=True)
710
        elif name == "False":
711
            return ExprNodes.BoolNode(pos, value=False)
712
        elif name == "NULL" and not s.in_python_file:
713
            return ExprNodes.NullNode(pos)
William Stein's avatar
William Stein committed
714
        else:
715
            return p_name(s, name)
William Stein's avatar
William Stein committed
716 717 718
    else:
        s.error("Expected an identifier or literal")

719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749
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)

750

751 752
def p_name(s, name):
    pos = s.position()
753 754
    if not s.compile_time_expr and name in s.compile_time_env:
        value = s.compile_time_env.lookup_here(name)
755 756 757 758 759 760 761 762
        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)
763 764
    if value is None:
        return ExprNodes.NoneNode(pos)
765 766
    elif value is Ellipsis:
        return ExprNodes.EllipsisNode(pos)
767
    elif isinstance(value, bool):
768 769
        return ExprNodes.BoolNode(pos, value=value)
    elif isinstance(value, int):
770
        return ExprNodes.IntNode(pos, value=rep, constant_result=value)
771
    elif isinstance(value, float):
772
        return ExprNodes.FloatNode(pos, value=rep, constant_result=value)
773
    elif isinstance(value, _unicode):
774
        return ExprNodes.UnicodeNode(pos, value=EncodedString(value))
775
    elif isinstance(value, _bytes):
776 777
        bvalue = bytes_literal(value, 'ascii')  # actually: unknown encoding, but BytesLiteral requires one
        return ExprNodes.BytesNode(pos, value=bvalue, constant_result=value)
778 779 780 781 782
    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)
783
        else:
784 785
            # error already reported
            return None
786
    elif not _IS_PY3 and isinstance(value, long):
787
        return ExprNodes.IntNode(pos, value=rep.rstrip('L'), constant_result=value)
788 789
    error(pos, "Invalid type for compile-time constant: %r (type %s)"
               % (value, value.__class__.__name__))
790 791
    return None

792

William Stein's avatar
William Stein committed
793 794
def p_cat_string_literal(s):
    # A sequence of one or more adjacent string literals.
795
    # Returns (kind, bytes_value, unicode_value)
796 797
    # where kind in ('b', 'c', 'u', 'f', '')
    pos = s.position()
798 799 800
    kind, bytes_value, unicode_value = p_string_literal(s)
    if kind == 'c' or s.sy != 'BEGIN_STRING':
        return kind, bytes_value, unicode_value
801
    bstrings, ustrings, positions = [bytes_value], [unicode_value], [pos]
802 803 804 805 806 807
    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")
808
            continue
809
        elif next_kind != kind:
810
            # concatenating f strings and normal strings is allowed and leads to an f string
Stefan Behnel's avatar
Stefan Behnel committed
811
            if set([kind, next_kind]) in (set(['f', 'u']), set(['f', ''])):
812 813 814 815 816 817 818 819
                kind = 'f'
            else:
                error(pos, "Cannot mix string literals of different types, expected %s'', got %s''" %
                      (kind, next_kind))
                continue
        bstrings.append(next_bytes_value)
        ustrings.append(next_unicode_value)
        positions.append(pos)
820
    # join and rewrap the partial literals
821
    if kind in ('b', 'c', '') or kind == 'u' and None not in bstrings:
822
        # Py3 enforced unicode literals are parsed as bytes/unicode combination
823
        bytes_value = bytes_literal(StringEncoding.join_bytes(bstrings), s.source_encoding)
824
    if kind in ('u', ''):
Stefan Behnel's avatar
Stefan Behnel committed
825
        unicode_value = EncodedString(u''.join([u for u in ustrings if u is not None]))
826 827 828 829 830 831 832
    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
833
                unicode_value.append(ExprNodes.UnicodeNode(pos, value=EncodedString(u)))
834 835
    return kind, bytes_value, unicode_value

Stefan Behnel's avatar
Stefan Behnel committed
836

837
def p_opt_string_literal(s, required_type='u'):
838
    if s.sy == 'BEGIN_STRING':
839 840 841 842 843 844 845
        kind, bytes_value, unicode_value = p_string_literal(s, required_type)
        if required_type == 'u':
            return unicode_value
        elif required_type == 'b':
            return bytes_value
        else:
            s.error("internal parser configuration error")
William Stein's avatar
William Stein committed
846 847 848
    else:
        return None

Stefan Behnel's avatar
Stefan Behnel committed
849

850 851 852 853 854 855
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
856

857
def p_string_literal(s, kind_override=None):
858
    # A single string or char literal.  Returns (kind, bvalue, uvalue)
859
    # where kind in ('b', 'c', 'u', 'f', '').  The 'bvalue' is the source
860 861 862
    # 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
863 864
    # representations. In f-strings, the uvalue is a list of the Unicode
    # strings and f-string expressions that make up the f-string.
865

William Stein's avatar
William Stein committed
866 867
    # s.sy == 'BEGIN_STRING'
    pos = s.position()
868
    is_raw = False
869
    is_python3_source = s.context.language_level >= 3
Stefan Behnel's avatar
Stefan Behnel committed
870
    has_non_ascii_literal_characters = False
871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895
    kind_string = s.systring.rstrip('"\'').lower()
    if len(set(kind_string)) != len(kind_string):
        s.error('Duplicate string prefix character')
    if 'b' in kind_string and 'u' in kind_string:
        s.error('String prefixes b and u cannot be combined')
    if 'b' in kind_string and 'f' in kind_string:
        s.error('String prefixes b and f cannot be combined')
    if 'u' in kind_string and 'f' in kind_string:
        s.error('String prefixes u and f cannot be combined')

    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:
            s.error('Invalid string prefix for character literal')
        kind = 'c'
    elif 'f' in kind_string:
        kind = 'f'  # u is ignored
    elif 'b' in kind_string:
        kind = 'b'
    elif 'u' in kind_string:
        kind = 'u'
    else:
William Stein's avatar
William Stein committed
896
        kind = ''
897

898 899 900 901 902 903
    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
904
        if kind in ('u', 'f'):  # f-strings are scanned exactly like Unicode literals, but are parsed further later
905 906 907 908 909
            chars = StringEncoding.UnicodeLiteralBuilder()
        elif kind == '':
            chars = StringEncoding.StrLiteralBuilder(s.source_encoding)
        else:
            chars = StringEncoding.BytesLiteralBuilder(s.source_encoding)
910

William Stein's avatar
William Stein committed
911 912 913
    while 1:
        s.next()
        sy = s.sy
914
        systr = s.systring
915
        # print "p_string_literal: sy =", sy, repr(s.systring) ###
William Stein's avatar
William Stein committed
916
        if sy == 'CHARS':
917
            chars.append(systr)
Stefan Behnel's avatar
Stefan Behnel committed
918 919
            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
920
        elif sy == 'ESCAPE':
921
            if is_raw:
922
                chars.append(systr)
Stefan Behnel's avatar
Stefan Behnel committed
923
                if is_python3_source and not has_non_ascii_literal_characters \
924
                       and check_for_non_ascii_characters(systr):
Stefan Behnel's avatar
Stefan Behnel committed
925
                    has_non_ascii_literal_characters = True
William Stein's avatar
William Stein committed
926 927
            else:
                c = systr[1]
928 929 930
                if c in u"01234567":
                    chars.append_charval( int(systr[1:], 8) )
                elif c in u"'\"\\":
931
                    chars.append(c)
932 933 934 935
                elif c in u"abfnrtv":
                    chars.append(
                        StringEncoding.char_from_escape_sequence(systr))
                elif c == u'\n':
William Stein's avatar
William Stein committed
936
                    pass
937
                elif c == u'x':   # \xXX
938 939 940
                    if len(systr) == 4:
                        chars.append_charval( int(systr[2:], 16) )
                    else:
941 942
                        s.error("Invalid hex escape '%s'" % systr,
                                fatal=False)
Jelle Zijlstra's avatar
Jelle Zijlstra committed
943
                elif c in u'NUu' and kind in ('u', 'f', ''):   # \uxxxx, \Uxxxxxxxx, \N{...}
944 945 946
                    chrval = -1
                    if c == u'N':
                        try:
947
                            chrval = ord(lookup_unicodechar(systr[3:-1]))
948
                        except KeyError:
949 950
                            s.error("Unknown Unicode character name %s" %
                                    repr(systr[3:-1]).lstrip('u'))
951 952 953
                    elif len(systr) in (6,10):
                        chrval = int(systr[2:], 16)
                        if chrval > 1114111: # sys.maxunicode:
954
                            s.error("Invalid unicode escape '%s'" % systr)
955
                            chrval = -1
956
                    else:
957 958
                        s.error("Invalid unicode escape '%s'" % systr,
                                fatal=False)
959 960
                    if chrval >= 0:
                        chars.append_uescape(chrval, systr)
William Stein's avatar
William Stein committed
961
                else:
962
                    chars.append(u'\\' + systr[1:])
Stefan Behnel's avatar
Stefan Behnel committed
963
                    if is_python3_source and not has_non_ascii_literal_characters \
964
                           and check_for_non_ascii_characters(systr):
Stefan Behnel's avatar
Stefan Behnel committed
965
                        has_non_ascii_literal_characters = True
William Stein's avatar
William Stein committed
966
        elif sy == 'NEWLINE':
967
            chars.append(u'\n')
William Stein's avatar
William Stein committed
968 969 970
        elif sy == 'END_STRING':
            break
        elif sy == 'EOF':
971
            s.error("Unclosed string literal", pos=pos)
William Stein's avatar
William Stein committed
972
        else:
973
            s.error("Unexpected token %r:%r in string literal" %
William Stein's avatar
William Stein committed
974
                    (sy, s.systring))
975

976
    if kind == 'c':
977 978 979 980
        unicode_value = None
        bytes_value = chars.getchar()
        if len(bytes_value) != 1:
            error(pos, u"invalid character literal: %r" % bytes_value)
981
    else:
982
        bytes_value, unicode_value = chars.getstrings()
Stefan Behnel's avatar
Stefan Behnel committed
983
        if is_python3_source and has_non_ascii_literal_characters:
984
            # Python 3 forbids literal non-ASCII characters in byte strings
985
            if kind not in ('u', 'f'):
986 987
                s.error("bytes can only contain ASCII literal characters.",
                        pos=pos, fatal=False)
988
            bytes_value = None
989 990
    if kind == 'f':
        unicode_value = p_f_string(s, unicode_value, pos)
William Stein's avatar
William Stein committed
991
    s.next()
992
    return (kind, bytes_value, unicode_value)
William Stein's avatar
William Stein committed
993

994

995 996 997 998 999 1000 1001 1002 1003
def p_f_string(s, unicode_value, pos):
    # Parses a PEP 498 f-string literal into a list of nodes. Nodes are either UnicodeNodes
    # or FormattedValueNodes.
    values = []
    i = 0
    size = len(unicode_value)
    current_literal_start = 0
    while i < size:
        c = unicode_value[i]
1004
        if c in '{}':
Jelle Zijlstra's avatar
Jelle Zijlstra committed
1005 1006
            if i + 1 < size and unicode_value[i + 1] == c:
                encoded_str = EncodedString(unicode_value[current_literal_start:i + 1])
1007
                values.append(ExprNodes.UnicodeNode(pos, value=encoded_str))
1008
                i += 2
Jelle Zijlstra's avatar
Jelle Zijlstra committed
1009 1010 1011
                current_literal_start = i
            elif c == '}':
                s.error("single '}' encountered in format string")
1012 1013
            else:
                encoded_str = EncodedString(unicode_value[current_literal_start:i])
1014
                values.append(ExprNodes.UnicodeNode(pos, value=encoded_str))
1015 1016 1017 1018 1019 1020 1021
                i, expr_node = p_f_string_expr(s, unicode_value, pos, i + 1)
                current_literal_start = i
                values.append(expr_node)
        else:
            i += 1

    encoded_str = EncodedString(unicode_value[current_literal_start:])
1022
    values.append(ExprNodes.UnicodeNode(pos, value=encoded_str))
1023 1024 1025 1026 1027 1028 1029 1030 1031
    return values


def p_f_string_expr(s, unicode_value, pos, starting_index):
    # 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)
    conversion_char = None
1032
    format_spec_str = u''
1033 1034

    nested_depth = 0
1035 1036
    quote_char = None
    in_triple_quotes = False
1037 1038 1039 1040 1041 1042

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

1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
        if quote_char is not None:
            if c == '\\':
                i += 1
            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
                        quote_char = None
                        i += 2
                else:
                    quote_char = None
        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 '{[(':
1060 1061 1062 1063 1064 1065 1066 1067
            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
1068
                i += 1
1069 1070 1071 1072 1073 1074
                continue

            terminal_char = c
            break
        i += 1

1075 1076 1077 1078 1079 1080
    # 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():
        s.error("empty expression not allowed in f-string")
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094

    if terminal_char == '!':
        i += 1
        if i >= size:
            s.error("invalid conversion char at end of string")

        conversion_char = unicode_value[i]

        i += 1
        if i >= size:
            s.error("invalid conversion char at end of string")
        terminal_char = unicode_value[i]

    if terminal_char == ':':
1095 1096
        in_triple_quotes = False
        in_string = False
1097 1098 1099 1100 1101 1102
        nested_depth = 0
        start_format_spec = i + 1
        while True:
            if i >= size:
                s.error("missing '}' in format specifier")
            c = unicode_value[i]
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
            if not in_triple_quotes and not in_string:
                if c == '{':
                    if nested_depth >= 1:
                        s.error("nesting of '{' in format specifier is not allowed")
                    nested_depth += 1
                elif c == '}' and nested_depth == 0:
                    terminal_char = c
                    break
                elif c == '}':
                    nested_depth -= 1
            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
1119 1120 1121 1122 1123 1124 1125
            i += 1

        format_spec_str = unicode_value[start_format_spec:i]

    if terminal_char != '}':
        s.error("missing '}' in format string expression'")

1126 1127 1128
    # 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)
1129 1130 1131
    expr = p_testlist(scanner)  # TODO is testlist right here?

    # validate the conversion char
1132
    if conversion_char is not None and not ExprNodes.FormattedValueNode.find_conversion_func(conversion_char):
1133 1134 1135 1136
        s.error("invalid conversion character '%s'" % conversion_char)

    # the format spec is itself treated like an f-string
    if format_spec_str is not None:
1137
        format_spec = ExprNodes.JoinedStrNode(pos, values=p_f_string(s, format_spec_str, pos))
1138 1139 1140 1141
    else:
        format_spec = None

    return i + 1, ExprNodes.FormattedValueNode(
1142 1143
        s.position(), value=expr, conversion_char=conversion_char,
        format_spec=format_spec)
1144 1145


1146 1147 1148
# since PEP 448:
# list_display  ::=     "[" [listmaker] "]"
# listmaker     ::=     (test|star_expr) ( comp_for | (',' (test|star_expr))* [','] )
Stefan Behnel's avatar
Stefan Behnel committed
1149
# comp_iter     ::=     comp_for | comp_if
1150 1151
# comp_for      ::=     "for" expression_list "in" testlist [comp_iter]
# comp_if       ::=     "if" test [comp_iter]
1152

William Stein's avatar
William Stein committed
1153 1154 1155 1156
def p_list_maker(s):
    # s.sy == '['
    pos = s.position()
    s.next()
Robert Bradshaw's avatar
Robert Bradshaw committed
1157 1158
    if s.sy == ']':
        s.expect(']')
1159 1160 1161
        return ExprNodes.ListNode(pos, args=[])

    expr = p_test_or_starred_expr(s)
Robert Bradshaw's avatar
Robert Bradshaw committed
1162
    if s.sy == 'for':
1163 1164
        if expr.is_starred:
            s.error("iterable unpacking cannot be used in comprehension")
1165
        append = ExprNodes.ComprehensionAppendNode(pos, expr=expr)
1166
        loop = p_comp_for(s, append)
Robert Bradshaw's avatar
Robert Bradshaw committed
1167
        s.expect(']')
1168
        return ExprNodes.ComprehensionNode(
1169
            pos, loop=loop, append=append, type=Builtin.list_type,
1170
            # list comprehensions leak their loop variable in Py2
1171 1172 1173 1174 1175 1176
            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
1177
    else:
1178 1179 1180 1181
        exprs = [expr]
    s.expect(']')
    return ExprNodes.ListNode(pos, args=exprs)

1182

Stefan Behnel's avatar
Stefan Behnel committed
1183
def p_comp_iter(s, body):
Robert Bradshaw's avatar
Robert Bradshaw committed
1184
    if s.sy == 'for':
1185
        return p_comp_for(s, body)
Robert Bradshaw's avatar
Robert Bradshaw committed
1186
    elif s.sy == 'if':
Stefan Behnel's avatar
Stefan Behnel committed
1187
        return p_comp_if(s, body)
Robert Bradshaw's avatar
Robert Bradshaw committed
1188
    else:
1189 1190
        # insert the 'append' operation into the loop
        return body
William Stein's avatar
William Stein committed
1191

1192
def p_comp_for(s, body):
Robert Bradshaw's avatar
Robert Bradshaw committed
1193 1194 1195
    # s.sy == 'for'
    pos = s.position()
    s.next()
1196
    kw = p_for_bounds(s, allow_testlist=False)
1197
    kw.update(else_clause = None, body = p_comp_iter(s, body))
Robert Bradshaw's avatar
Robert Bradshaw committed
1198
    return Nodes.ForStatNode(pos, **kw)
1199

Stefan Behnel's avatar
Stefan Behnel committed
1200
def p_comp_if(s, body):
Robert Bradshaw's avatar
Robert Bradshaw committed
1201 1202 1203
    # s.sy == 'if'
    pos = s.position()
    s.next()
Stefan Behnel's avatar
Stefan Behnel committed
1204
    test = p_test_nocond(s)
1205
    return Nodes.IfStatNode(pos,
1206
        if_clauses = [Nodes.IfClauseNode(pos, condition = test,
Stefan Behnel's avatar
Stefan Behnel committed
1207
                                         body = p_comp_iter(s, body))],
Robert Bradshaw's avatar
Robert Bradshaw committed
1208
        else_clause = None )
1209

1210 1211 1212 1213 1214 1215

# 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
1216

1217
def p_dict_or_set_maker(s):
William Stein's avatar
William Stein committed
1218 1219 1220
    # s.sy == '{'
    pos = s.position()
    s.next()
1221
    if s.sy == '}':
William Stein's avatar
William Stein committed
1222
        s.next()
1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236
        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()
1237 1238 1239
            if s.sy == '*':
                s.error("expected expression, found '*'")
            item = p_starred_expr(s)
1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258
            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 == ',':
1259
            s.next()
1260 1261
            if s.sy == '}':
                break
1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276
        else:
            break

    if s.sy == 'for':
        # 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)
1277
            loop = p_comp_for(s, append)
1278
            s.expect('}')
1279
            return ExprNodes.ComprehensionNode(pos, loop=loop, append=append, type=comprehension_type)
1280
        else:
1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305
            # 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]
1306
        return ExprNodes.MergedSequenceNode(pos, args=items, type=Builtin.set_type)
1307
    else:
1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324
        # (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
1325

1326
# NOTE: no longer in Py3 :)
William Stein's avatar
William Stein committed
1327 1328 1329 1330
def p_backquote_expr(s):
    # s.sy == '`'
    pos = s.position()
    s.next()
1331 1332 1333 1334
    args = [p_test(s)]
    while s.sy == ',':
        s.next()
        args.append(p_test(s))
William Stein's avatar
William Stein committed
1335
    s.expect('`')
1336 1337 1338 1339
    if len(args) == 1:
        arg = args[0]
    else:
        arg = ExprNodes.TupleNode(pos, args = args)
William Stein's avatar
William Stein committed
1340 1341
    return ExprNodes.BackquoteNode(pos, arg = arg)

1342 1343
def p_simple_expr_list(s, expr=None):
    exprs = expr is not None and [expr] or []
William Stein's avatar
William Stein committed
1344
    while s.sy not in expr_terminators:
1345
        exprs.append( p_test(s) )
Stefan Behnel's avatar
Stefan Behnel committed
1346
        if s.sy != ',':
William Stein's avatar
William Stein committed
1347 1348 1349 1350
            break
        s.next()
    return exprs

1351

1352 1353 1354
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:
1355
        exprs.append(p_test_or_starred_expr(s))
1356 1357 1358 1359
        if s.sy != ',':
            break
        s.next()
    return exprs
1360

1361 1362 1363 1364

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

def p_testlist(s):
William Stein's avatar
William Stein committed
1365
    pos = s.position()
1366
    expr = p_test(s)
William Stein's avatar
William Stein committed
1367 1368
    if s.sy == ',':
        s.next()
1369
        exprs = p_simple_expr_list(s, expr)
William Stein's avatar
William Stein committed
1370 1371 1372 1373
        return ExprNodes.TupleNode(pos, args = exprs)
    else:
        return expr

1374
# testlist_star_expr: (test|star_expr) ( comp_for | (',' (test|star_expr))* [','] )
Robert Bradshaw's avatar
Robert Bradshaw committed
1375

1376
def p_testlist_star_expr(s):
Robert Bradshaw's avatar
Robert Bradshaw committed
1377
    pos = s.position()
1378
    expr = p_test_or_starred_expr(s)
Robert Bradshaw's avatar
Robert Bradshaw committed
1379
    if s.sy == ',':
1380
        s.next()
1381 1382
        exprs = p_test_or_starred_expr_list(s, expr)
        return ExprNodes.TupleNode(pos, args = exprs)
Robert Bradshaw's avatar
Robert Bradshaw committed
1383 1384 1385
    else:
        return expr

1386 1387 1388
# testlist_comp: (test|star_expr) ( comp_for | (',' (test|star_expr))* [','] )

def p_testlist_comp(s):
Robert Bradshaw's avatar
Robert Bradshaw committed
1389
    pos = s.position()
1390
    expr = p_test_or_starred_expr(s)
Robert Bradshaw's avatar
Robert Bradshaw committed
1391
    if s.sy == ',':
1392
        s.next()
1393
        exprs = p_test_or_starred_expr_list(s, expr)
Robert Bradshaw's avatar
Robert Bradshaw committed
1394
        return ExprNodes.TupleNode(pos, args = exprs)
1395 1396
    elif s.sy == 'for':
        return p_genexp(s, expr)
Robert Bradshaw's avatar
Robert Bradshaw committed
1397 1398
    else:
        return expr
1399 1400 1401

def p_genexp(s, expr):
    # s.sy == 'for'
1402 1403
    loop = p_comp_for(s, Nodes.ExprStatNode(
        expr.pos, expr = ExprNodes.YieldExprNode(expr.pos, arg=expr)))
1404 1405
    return ExprNodes.GeneratorExpressionNode(expr.pos, loop=loop)

1406 1407
expr_terminators = cython.declare(set, set([
    ')', ']', '}', ':', '=', 'NEWLINE']))
William Stein's avatar
William Stein committed
1408

1409

William Stein's avatar
William Stein committed
1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422
#-------------------------------------------------------
#
#   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)

1423

1424 1425 1426 1427 1428 1429
def p_nonlocal_statement(s):
    pos = s.position()
    s.next()
    names = p_ident_list(s)
    return Nodes.NonlocalNode(pos, names = names)

1430

William Stein's avatar
William Stein committed
1431
def p_expression_or_assignment(s):
1432
    expr_list = [p_testlist_star_expr(s)]
1433 1434 1435 1436 1437 1438
    if s.sy == '=' and expr_list[0].is_starred:
        # 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] = ...",
                pos=expr_list[0].pos)
William Stein's avatar
William Stein committed
1439 1440
    while s.sy == '=':
        s.next()
1441 1442 1443 1444 1445
        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
1446
    if len(expr_list) == 1:
1447
        if re.match(r"([-+*/%^&|]|<<|>>|\*\*|//|@)=", s.sy):
1448
            lhs = expr_list[0]
1449 1450 1451 1452 1453 1454
            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))
1455
            elif not isinstance(lhs, (ExprNodes.AttributeNode, ExprNodes.IndexNode, ExprNodes.NameNode)):
1456
                error(lhs.pos, "Illegal operand for inplace operation.")
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
1457
            operator = s.sy[:-1]
1458
            s.next()
1459 1460 1461 1462
            if s.sy == 'yield':
                rhs = p_yield_expression(s)
            else:
                rhs = p_testlist(s)
1463
            return Nodes.InPlaceAssignmentNode(lhs.pos, operator=operator, lhs=lhs, rhs=rhs)
1464
        expr = expr_list[0]
1465
        return Nodes.ExprStatNode(expr.pos, expr=expr)
1466

1467 1468
    rhs = expr_list[-1]
    if len(expr_list) == 2:
1469
        return Nodes.SingleAssignmentNode(rhs.pos, lhs=expr_list[0], rhs=rhs)
William Stein's avatar
William Stein committed
1470
    else:
1471 1472
        return Nodes.CascadedAssignmentNode(rhs.pos, lhs_list=expr_list[:-1], rhs=rhs)

William Stein's avatar
William Stein committed
1473 1474 1475 1476

def p_print_statement(s):
    # s.sy == 'print'
    pos = s.position()
1477
    ends_with_comma = 0
William Stein's avatar
William Stein committed
1478 1479
    s.next()
    if s.sy == '>>':
1480
        s.next()
1481
        stream = p_test(s)
1482 1483 1484 1485 1486
        if s.sy == ',':
            s.next()
            ends_with_comma = s.sy in ('NEWLINE', 'EOF')
    else:
        stream = None
William Stein's avatar
William Stein committed
1487 1488
    args = []
    if s.sy not in ('NEWLINE', 'EOF'):
1489
        args.append(p_test(s))
William Stein's avatar
William Stein committed
1490 1491 1492
        while s.sy == ',':
            s.next()
            if s.sy in ('NEWLINE', 'EOF'):
1493
                ends_with_comma = 1
William Stein's avatar
William Stein committed
1494
                break
1495
            args.append(p_test(s))
1496
    arg_tuple = ExprNodes.TupleNode(pos, args=args)
1497
    return Nodes.PrintStatNode(pos,
1498 1499 1500
        arg_tuple=arg_tuple, stream=stream,
        append_newline=not ends_with_comma)

William Stein's avatar
William Stein committed
1501

1502 1503 1504 1505
def p_exec_statement(s):
    # s.sy == 'exec'
    pos = s.position()
    s.next()
1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517
    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]
1518
    if s.sy == 'in':
1519 1520 1521
        if tuple_variant:
            s.error("tuple variant of exec does not support additional 'in' arguments",
                    fatal=False)
1522
        s.next()
1523
        args.append(p_test(s))
1524 1525
        if s.sy == ',':
            s.next()
1526
            args.append(p_test(s))
1527
    return Nodes.ExecStatNode(pos, args=args)
1528

William Stein's avatar
William Stein committed
1529 1530 1531 1532
def p_del_statement(s):
    # s.sy == 'del'
    pos = s.position()
    s.next()
1533
    # FIXME: 'exprlist' in Python
William Stein's avatar
William Stein committed
1534 1535 1536 1537 1538 1539 1540
    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:
1541
        s.expect_newline("Expected a newline", ignore_semicolon=True)
William Stein's avatar
William Stein committed
1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560
    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:
1561
        value = p_testlist(s)
William Stein's avatar
William Stein committed
1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572
    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
1573
    cause = None
William Stein's avatar
William Stein committed
1574
    if s.sy not in statement_terminators:
1575
        exc_type = p_test(s)
William Stein's avatar
William Stein committed
1576 1577
        if s.sy == ',':
            s.next()
1578
            exc_value = p_test(s)
William Stein's avatar
William Stein committed
1579 1580
            if s.sy == ',':
                s.next()
1581
                exc_tb = p_test(s)
Haoyu Bai's avatar
Haoyu Bai committed
1582 1583 1584
        elif s.sy == 'from':
            s.next()
            cause = p_test(s)
1585
    if exc_type or exc_value or exc_tb:
1586
        return Nodes.RaiseStatNode(pos,
1587 1588
            exc_type = exc_type,
            exc_value = exc_value,
Haoyu Bai's avatar
Haoyu Bai committed
1589 1590
            exc_tb = exc_tb,
            cause = cause)
1591 1592
    else:
        return Nodes.ReraiseStatNode(pos)
William Stein's avatar
William Stein committed
1593

1594

William Stein's avatar
William Stein committed
1595 1596 1597 1598 1599
def p_import_statement(s):
    # s.sy in ('import', 'cimport')
    pos = s.position()
    kind = s.sy
    s.next()
1600
    items = [p_dotted_name(s, as_allowed=1)]
William Stein's avatar
William Stein committed
1601 1602
    while s.sy == ',':
        s.next()
1603
        items.append(p_dotted_name(s, as_allowed=1))
William Stein's avatar
William Stein committed
1604
    stats = []
1605
    is_absolute = Future.absolute_import in s.context.future_directives
William Stein's avatar
William Stein committed
1606 1607
    for pos, target_name, dotted_name, as_name in items:
        if kind == 'cimport':
1608 1609 1610 1611 1612
            stat = Nodes.CImportStatNode(
                pos,
                module_name=dotted_name,
                as_name=as_name,
                is_absolute=is_absolute)
William Stein's avatar
William Stein committed
1613
        else:
1614
            if as_name and "." in dotted_name:
1615
                name_list = ExprNodes.ListNode(pos, args=[
1616
                    ExprNodes.IdentifierStringNode(pos, value=s.context.intern_ustring("*"))])
1617 1618
            else:
                name_list = None
1619 1620 1621 1622 1623 1624 1625 1626
            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
1627
        stats.append(stat)
1628 1629
    return Nodes.StatListNode(pos, stats=stats)

William Stein's avatar
William Stein committed
1630

Stefan Behnel's avatar
Stefan Behnel committed
1631
def p_from_import_statement(s, first_statement = 0):
William Stein's avatar
William Stein committed
1632 1633 1634
    # s.sy == 'from'
    pos = s.position()
    s.next()
Haoyu Bai's avatar
Haoyu Bai committed
1635 1636 1637 1638 1639 1640 1641
    if s.sy == '.':
        # count relative import level
        level = 0
        while s.sy == '.':
            level += 1
            s.next()
    else:
1642
        level = None
1643
    if level is not None and s.sy in ('import', 'cimport'):
Haoyu Bai's avatar
Haoyu Bai committed
1644
        # we are dealing with "from .. import foo, bar"
1645
        dotted_name_pos, dotted_name = s.position(), s.context.intern_ustring('')
William Stein's avatar
William Stein committed
1646
    else:
1647 1648 1649 1650
        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
1651
        s.error("Expected 'import' or 'cimport'")
1652 1653
    kind = s.sy
    s.next()
Haoyu Bai's avatar
Haoyu Bai committed
1654

1655
    is_cimport = kind == 'cimport'
1656
    is_parenthesized = False
William Stein's avatar
William Stein committed
1657
    if s.sy == '*':
1658
        imported_names = [(s.position(), s.context.intern_ustring("*"), None, None)]
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
1659 1660
        s.next()
    else:
1661 1662 1663
        if s.sy == '(':
            is_parenthesized = True
            s.next()
1664
        imported_names = [p_imported_name(s, is_cimport)]
William Stein's avatar
William Stein committed
1665 1666
    while s.sy == ',':
        s.next()
1667 1668
        if is_parenthesized and s.sy == ')':
            break
1669
        imported_names.append(p_imported_name(s, is_cimport))
1670 1671
    if is_parenthesized:
        s.expect(')')
Stefan Behnel's avatar
Stefan Behnel committed
1672 1673 1674
    if dotted_name == '__future__':
        if not first_statement:
            s.error("from __future__ imports must occur at the beginning of the file")
1675
        elif level:
Haoyu Bai's avatar
Haoyu Bai committed
1676
            s.error("invalid syntax")
Stefan Behnel's avatar
Stefan Behnel committed
1677
        else:
1678
            for (name_pos, name, as_name, kind) in imported_names:
1679 1680 1681
                if name == "braces":
                    s.error("not a chance", name_pos)
                    break
Stefan Behnel's avatar
Stefan Behnel committed
1682 1683 1684
                try:
                    directive = getattr(Future, name)
                except AttributeError:
1685
                    s.error("future feature %s is not defined" % name, name_pos)
Stefan Behnel's avatar
Stefan Behnel committed
1686 1687 1688 1689
                    break
                s.context.future_directives.add(directive)
        return Nodes.PassStatNode(pos)
    elif kind == 'cimport':
1690 1691 1692 1693
        return Nodes.FromCImportStatNode(
            pos, module_name=dotted_name,
            relative_level=level,
            imported_names=imported_names)
William Stein's avatar
William Stein committed
1694 1695 1696
    else:
        imported_name_strings = []
        items = []
1697
        for (name_pos, name, as_name, kind) in imported_names:
William Stein's avatar
William Stein committed
1698
            imported_name_strings.append(
1699
                ExprNodes.IdentifierStringNode(name_pos, value=name))
William Stein's avatar
William Stein committed
1700
            items.append(
1701
                (name, ExprNodes.NameNode(name_pos, name=as_name or name)))
William Stein's avatar
William Stein committed
1702
        import_list = ExprNodes.ListNode(
1703
            imported_names[0][0], args=imported_name_strings)
William Stein's avatar
William Stein committed
1704 1705
        return Nodes.FromImportStatNode(pos,
            module = ExprNodes.ImportNode(dotted_name_pos,
1706
                module_name = ExprNodes.IdentifierStringNode(pos, value = dotted_name),
Haoyu Bai's avatar
Haoyu Bai committed
1707
                level = level,
William Stein's avatar
William Stein committed
1708 1709 1710
                name_list = import_list),
            items = items)

1711 1712

imported_name_kinds = cython.declare(set, set(['class', 'struct', 'union']))
1713 1714

def p_imported_name(s, is_cimport):
William Stein's avatar
William Stein committed
1715
    pos = s.position()
1716 1717 1718 1719
    kind = None
    if is_cimport and s.systring in imported_name_kinds:
        kind = s.systring
        s.next()
William Stein's avatar
William Stein committed
1720 1721
    name = p_ident(s)
    as_name = p_as_name(s)
1722
    return (pos, name, as_name, kind)
William Stein's avatar
William Stein committed
1723

1724

William Stein's avatar
William Stein committed
1725 1726 1727 1728 1729 1730 1731 1732 1733 1734
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)
1735 1736
    return (pos, target_name, s.context.intern_ustring(u'.'.join(names)), as_name)

William Stein's avatar
William Stein committed
1737 1738 1739 1740 1741 1742 1743 1744

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

1745

William Stein's avatar
William Stein committed
1746 1747 1748 1749
def p_assert_statement(s):
    # s.sy == 'assert'
    pos = s.position()
    s.next()
1750
    cond = p_test(s)
William Stein's avatar
William Stein committed
1751 1752
    if s.sy == ',':
        s.next()
1753
        value = p_test(s)
William Stein's avatar
William Stein committed
1754 1755 1756 1757
    else:
        value = None
    return Nodes.AssertStatNode(pos, cond = cond, value = value)

1758

1759
statement_terminators = cython.declare(set, set([';', 'NEWLINE', 'EOF']))
William Stein's avatar
William Stein committed
1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774

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()
1775
    test = p_test(s)
William Stein's avatar
William Stein committed
1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790
    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()
1791
    test = p_test(s)
William Stein's avatar
William Stein committed
1792 1793
    body = p_suite(s)
    else_clause = p_else_clause(s)
1794 1795
    return Nodes.WhileStatNode(pos,
        condition = test, body = body,
William Stein's avatar
William Stein committed
1796 1797
        else_clause = else_clause)

1798 1799

def p_for_statement(s, is_async=False):
William Stein's avatar
William Stein committed
1800 1801 1802
    # s.sy == 'for'
    pos = s.position()
    s.next()
1803
    kw = p_for_bounds(s, allow_testlist=True, is_async=is_async)
1804 1805
    body = p_suite(s)
    else_clause = p_else_clause(s)
1806
    kw.update(body=body, else_clause=else_clause, is_async=is_async)
Robert Bradshaw's avatar
Robert Bradshaw committed
1807
    return Nodes.ForStatNode(pos, **kw)
1808

1809 1810

def p_for_bounds(s, allow_testlist=True, is_async=False):
William Stein's avatar
William Stein committed
1811 1812 1813
    target = p_for_target(s)
    if s.sy == 'in':
        s.next()
1814 1815 1816
        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
1817 1818 1819 1820 1821 1822
        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
1823 1824 1825 1826 1827 1828
        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)
1829
        step = p_for_from_step(s)
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
1830 1831 1832 1833
        if target is None:
            target = ExprNodes.NameNode(name2_pos, name = name2)
        else:
            if not target.is_name:
1834
                error(target.pos,
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
1835 1836 1837 1838
                    "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
1839
        if rel1[0] != rel2[0]:
William Stein's avatar
William Stein committed
1840 1841
            error(rel2_pos,
                "Relation directions in for-from do not match")
1842 1843 1844
        return dict(target = target,
                    bound1 = bound1,
                    relation1 = rel1,
1845 1846 1847 1848
                    relation2 = rel2,
                    bound2 = bound2,
                    step = step,
                    )
1849 1850 1851
    else:
        s.expect('in')
        return {}
William Stein's avatar
William Stein committed
1852 1853 1854 1855 1856 1857 1858 1859

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 '<', '<=', '>' '>='")
1860

1861
def p_for_from_step(s):
1862
    if s.sy == 'IDENT' and s.systring == 'by':
1863 1864 1865 1866 1867
        s.next()
        step = p_bit_expr(s)
        return step
    else:
        return None
William Stein's avatar
William Stein committed
1868

1869
inequality_relations = cython.declare(set, set(['<', '<=', '>', '>=']))
William Stein's avatar
William Stein committed
1870

1871
def p_target(s, terminator):
William Stein's avatar
William Stein committed
1872
    pos = s.position()
1873
    expr = p_starred_expr(s)
William Stein's avatar
William Stein committed
1874 1875 1876
    if s.sy == ',':
        s.next()
        exprs = [expr]
1877
        while s.sy != terminator:
1878
            exprs.append(p_starred_expr(s))
Stefan Behnel's avatar
Stefan Behnel committed
1879
            if s.sy != ',':
William Stein's avatar
William Stein committed
1880 1881 1882 1883 1884 1885
                break
            s.next()
        return ExprNodes.TupleNode(pos, args = exprs)
    else:
        return expr

1886

1887 1888 1889
def p_for_target(s):
    return p_target(s, 'in')

1890 1891

def p_for_iterator(s, allow_testlist=True, is_async=False):
William Stein's avatar
William Stein committed
1892
    pos = s.position()
1893 1894 1895 1896
    if allow_testlist:
        expr = p_testlist(s)
    else:
        expr = p_or_test(s)
1897 1898
    return (ExprNodes.AsyncIteratorNode if is_async else ExprNodes.IteratorNode)(pos, sequence=expr)

William Stein's avatar
William Stein committed
1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912

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)
1913
        body = Nodes.TryExceptStatNode(pos,
William Stein's avatar
William Stein committed
1914 1915
            body = body, except_clauses = except_clauses,
            else_clause = else_clause)
1916 1917 1918 1919
        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
1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932
        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
1933
    is_except_as = False
Stefan Behnel's avatar
Stefan Behnel committed
1934
    if s.sy != ':':
1935
        exc_type = p_test(s)
1936 1937 1938 1939 1940
        # normalise into list of single exception tests
        if isinstance(exc_type, ExprNodes.TupleNode):
            exc_type = exc_type.args
        else:
            exc_type = [exc_type]
1941 1942
        if s.sy == ',' or (s.sy == 'IDENT' and s.systring == 'as'
                           and s.context.language_level == 2):
William Stein's avatar
William Stein committed
1943
            s.next()
1944
            exc_value = p_test(s)
1945
        elif s.sy == 'IDENT' and s.systring == 'as':
1946
            # Py3 syntax requires a name here
1947
            s.next()
1948 1949 1950
            pos2 = s.position()
            name = p_ident(s)
            exc_value = ExprNodes.NameNode(pos2, name = name)
1951
            is_except_as = True
William Stein's avatar
William Stein committed
1952 1953
    body = p_suite(s)
    return Nodes.ExceptClauseNode(pos,
1954 1955
        pattern = exc_type, target = exc_value,
        body = body, is_except_as=is_except_as)
William Stein's avatar
William Stein committed
1956

1957
def p_include_statement(s, ctx):
William Stein's avatar
William Stein committed
1958 1959
    pos = s.position()
    s.next() # 'include'
1960
    unicode_include_file_name = p_string_literal(s, 'u')[2]
William Stein's avatar
William Stein committed
1961
    s.expect_newline("Syntax error in include statement")
1962
    if s.compile_time_eval:
1963
        include_file_name = unicode_include_file_name
1964 1965
        include_file_path = s.context.find_include_file(include_file_name, pos)
        if include_file_path:
1966
            s.included_files.append(include_file_name)
1967 1968 1969
            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)
1970
                tree = p_statement_list(s2, ctx)
1971 1972 1973
            return tree
        else:
            return None
William Stein's avatar
William Stein committed
1974
    else:
1975 1976
        return Nodes.PassStatNode(pos)

1977

1978
def p_with_statement(s):
1979
    s.next()  # 'with'
1980
    if s.systring == 'template' and not s.in_python_file:
1981 1982 1983 1984 1985
        node = p_with_template(s)
    else:
        node = p_with_items(s)
    return node

1986 1987

def p_with_items(s, is_async=False):
1988
    pos = s.position()
1989
    if not s.in_python_file and s.sy == 'IDENT' and s.systring in ('nogil', 'gil'):
1990 1991
        if is_async:
            s.error("with gil/nogil cannot be async")
1992 1993
        state = s.systring
        s.next()
1994
        if s.sy == ',':
Danilo Freitas's avatar
Danilo Freitas committed
1995
            s.next()
1996
            body = p_with_items(s)
Danilo Freitas's avatar
Danilo Freitas committed
1997
        else:
1998
            body = p_suite(s)
1999
        return Nodes.GILStatNode(pos, state=state, body=body)
2000
    else:
2001
        manager = p_test(s)
2002 2003 2004
        target = None
        if s.sy == 'IDENT' and s.systring == 'as':
            s.next()
2005 2006 2007
            target = p_starred_expr(s)
        if s.sy == ',':
            s.next()
2008
            body = p_with_items(s, is_async=is_async)
2009 2010
        else:
            body = p_suite(s)
2011 2012
    return Nodes.WithStatNode(pos, manager=manager, target=target, body=body, is_async=is_async)

2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037

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
2038
def p_simple_statement(s, first_statement = 0):
William Stein's avatar
William Stein committed
2039 2040 2041
    #print "p_simple_statement:", s.sy, s.systring ###
    if s.sy == 'global':
        node = p_global_statement(s)
2042 2043
    elif s.sy == 'nonlocal':
        node = p_nonlocal_statement(s)
William Stein's avatar
William Stein committed
2044 2045
    elif s.sy == 'print':
        node = p_print_statement(s)
2046 2047
    elif s.sy == 'exec':
        node = p_exec_statement(s)
William Stein's avatar
William Stein committed
2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060
    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
2061
        node = p_from_import_statement(s, first_statement = first_statement)
2062
    elif s.sy == 'yield':
2063
        node = p_yield_statement(s)
William Stein's avatar
William Stein committed
2064 2065 2066 2067 2068 2069 2070 2071
    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

2072
def p_simple_statement_list(s, ctx, first_statement = 0):
William Stein's avatar
William Stein committed
2073 2074
    # Parse a series of simple statements on one line
    # separated by semicolons.
Stefan Behnel's avatar
Stefan Behnel committed
2075
    stat = p_simple_statement(s, first_statement = first_statement)
2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096
    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)
William Stein's avatar
William Stein committed
2097 2098 2099
    s.expect_newline("Syntax error in simple statement list")
    return stat

2100 2101 2102
def p_compile_time_expr(s):
    old = s.compile_time_expr
    s.compile_time_expr = 1
2103
    expr = p_testlist(s)
2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116
    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)
    value = expr.compile_time_value(denv)
    #print "p_DEF_statement: %s = %r" % (name, value) ###
    denv.declare(name, value)
2117
    s.expect_newline("Expected a newline", ignore_semicolon=True)
2118 2119
    return Nodes.PassStatNode(pos)

2120
def p_IF_statement(s, ctx):
Dag Sverre Seljebotn's avatar
Merge  
Dag Sverre Seljebotn committed
2121
    pos = s.position()
2122 2123 2124 2125 2126 2127 2128 2129
    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))
2130
        body = p_suite(s, ctx)
2131 2132 2133
        if s.compile_time_eval:
            result = body
            current_eval = 0
Stefan Behnel's avatar
Stefan Behnel committed
2134
        if s.sy != 'ELIF':
2135 2136 2137 2138
            break
    if s.sy == 'ELSE':
        s.next()
        s.compile_time_eval = current_eval
2139
        body = p_suite(s, ctx)
2140 2141 2142
        if current_eval:
            result = body
    if not result:
Stefan Behnel's avatar
Stefan Behnel committed
2143
        result = Nodes.PassStatNode(pos)
2144 2145 2146
    s.compile_time_eval = saved_eval
    return result

2147 2148
def p_statement(s, ctx, first_statement = 0):
    cdef_flag = ctx.cdef_flag
Robert Bradshaw's avatar
Robert Bradshaw committed
2149
    decorators = None
William Stein's avatar
William Stein committed
2150
    if s.sy == 'ctypedef':
2151
        if ctx.level not in ('module', 'module_pxd'):
William Stein's avatar
William Stein committed
2152
            s.error("ctypedef statement not allowed here")
2153 2154
        #if ctx.api:
        #    error(s.position(), "'api' not allowed with 'ctypedef'")
2155
        return p_ctypedef_statement(s, ctx)
2156 2157 2158
    elif s.sy == 'DEF':
        return p_DEF_statement(s)
    elif s.sy == 'IF':
2159
        return p_IF_statement(s, ctx)
2160
    elif s.sy == '@':
Haoyu Bai's avatar
Haoyu Bai committed
2161
        if ctx.level not in ('module', 'class', 'c_class', 'function', 'property', 'module_pxd', 'c_class_pxd', 'other'):
2162 2163 2164
            s.error('decorator not allowed here')
        s.level = ctx.level
        decorators = p_decorators(s)
2165 2166 2167 2168 2169
        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")
2170 2171
    elif s.sy == 'pass' and cdef_flag:
        # empty cdef block
2172
        return p_pass_statement(s, with_newline=1)
2173 2174 2175 2176 2177

    overridable = 0
    if s.sy == 'cdef':
        cdef_flag = 1
        s.next()
2178
    elif s.sy == 'cpdef':
2179 2180 2181 2182 2183 2184 2185
        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
2186
        node = p_cdef_statement(s, ctx(overridable=overridable))
2187
        if decorators is not None:
2188
            tup = (Nodes.CFuncDefNode, Nodes.CVarDefNode, Nodes.CClassDefNode)
2189
            if ctx.allow_struct_enum_decorator:
2190
                tup += (Nodes.CStructOrUnionDefNode, Nodes.CEnumDefNode)
2191
            if not isinstance(node, tup):
2192
                s.error("Decorators can only be followed by functions or classes")
2193 2194
            node.decorators = decorators
        return node
William Stein's avatar
William Stein committed
2195
    else:
2196
        if ctx.api:
2197
            s.error("'api' not allowed with this statement", fatal=False)
2198
        elif s.sy == 'def':
2199 2200 2201
            # 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'):
2202
                s.error('def statement not allowed here')
2203
            s.level = ctx.level
2204 2205
            return p_def_statement(s, decorators)
        elif s.sy == 'class':
2206
            if ctx.level not in ('module', 'function', 'class', 'other'):
2207
                s.error("class definition not allowed here")
2208
            return p_class_statement(s, decorators)
2209 2210 2211 2212 2213 2214 2215
        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':
2216
            return p_pass_statement(s, with_newline=True)
William Stein's avatar
William Stein committed
2217
        else:
2218
            if ctx.level in ('c_class_pxd', 'property'):
2219 2220 2221
                node = p_ignorable_statement(s)
                if node is not None:
                    return node
2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232
                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)
2233 2234 2235
            elif s.sy == 'async':
                s.next()
                return p_async_statement(s, ctx, decorators)
2236
            else:
2237
                if s.sy == 'IDENT' and s.systring == 'async':
2238
                    ident_name = s.systring
2239 2240 2241
                    # PEP 492 enables the async/await keywords when it spots "async def ..."
                    s.next()
                    if s.sy == 'def':
2242
                        return p_async_statement(s, ctx, decorators)
2243 2244
                    elif decorators:
                        s.error("Decorators can only be followed by functions or classes")
2245
                    s.put_back('IDENT', ident_name)  # re-insert original token
2246 2247
                return p_simple_statement_list(s, ctx, first_statement=first_statement)

William Stein's avatar
William Stein committed
2248

2249
def p_statement_list(s, ctx, first_statement = 0):
William Stein's avatar
William Stein committed
2250 2251 2252 2253
    # Parse a series of statements separated by newlines.
    pos = s.position()
    stats = []
    while s.sy not in ('DEDENT', 'EOF'):
2254 2255 2256 2257 2258 2259 2260 2261
        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:
2262 2263 2264
        return stats[0]
    else:
        return Nodes.StatListNode(pos, stats = stats)
William Stein's avatar
William Stein committed
2265

2266 2267 2268 2269 2270 2271

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
2272 2273 2274 2275 2276
    s.expect(':')
    doc = None
    if s.sy == 'NEWLINE':
        s.next()
        s.expect_indent()
2277
        if with_doc_only:
2278
            doc = p_doc_string(s)
2279
        body = p_statement_list(s, ctx)
William Stein's avatar
William Stein committed
2280 2281
        s.expect_dedent()
    else:
2282
        if ctx.api:
2283
            s.error("'api' not allowed with this statement", fatal=False)
2284 2285
        if ctx.level in ('module', 'class', 'function', 'other'):
            body = p_simple_statement_list(s, ctx)
William Stein's avatar
William Stein committed
2286 2287
        else:
            body = p_pass_statement(s)
2288
            s.expect_newline("Syntax error in declarations", ignore_semicolon=True)
2289 2290 2291 2292
    if not with_doc_only:
        doc, body = _extract_docstring(body)
    return doc, body

William Stein's avatar
William Stein committed
2293

2294
def p_positional_and_keyword_args(s, end_sy_set, templates = None):
2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307
    """
    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 == '**':
2308
            s.error('Argument expansion not allowed here.', fatal=False)
2309 2310

        parsed_type = False
2311
        if s.sy == 'IDENT' and s.peek()[0] == '=':
2312
            ident = s.systring
2313
            s.next() # s.sy is '='
2314
            s.next()
2315
            if looking_at_expr(s):
2316
                arg = p_test(s)
2317 2318
            else:
                base_type = p_c_base_type(s, templates = templates)
2319
                declarator = p_c_declarator(s, empty = 1)
2320
                arg = Nodes.CComplexBaseTypeNode(base_type.pos,
2321 2322
                    base_type = base_type, declarator = declarator)
                parsed_type = True
2323
            keyword_node = ExprNodes.IdentifierStringNode(arg.pos, value=ident)
2324 2325
            keyword_args.append((keyword_node, arg))
            was_keyword = True
2326

2327
        else:
2328
            if looking_at_expr(s):
2329
                arg = p_test(s)
2330 2331
            else:
                base_type = p_c_base_type(s, templates = templates)
2332
                declarator = p_c_declarator(s, empty = 1)
2333
                arg = Nodes.CComplexBaseTypeNode(base_type.pos,
2334
                    base_type = base_type, declarator = declarator)
2335 2336 2337 2338 2339
                parsed_type = True
            positional_args.append(arg)
            pos_idx += 1
            if len(keyword_args) > 0:
                s.error("Non-keyword arg following keyword arg",
2340
                        pos=arg.pos)
2341 2342 2343 2344

        if s.sy != ',':
            if s.sy not in end_sy_set:
                if parsed_type:
2345
                    s.error("Unmatched %s" % " or ".join(end_sy_set))
2346 2347 2348 2349
            break
        s.next()
    return positional_args, keyword_args

Danilo Freitas's avatar
Danilo Freitas committed
2350
def p_c_base_type(s, self_flag = 0, nonempty = 0, templates = None):
William Stein's avatar
William Stein committed
2351 2352 2353
    # 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 == '(':
2354
        return p_c_complex_base_type(s, templates = templates)
William Stein's avatar
William Stein committed
2355
    else:
Danilo Freitas's avatar
Danilo Freitas committed
2356
        return p_c_simple_base_type(s, self_flag, nonempty = nonempty, templates = templates)
William Stein's avatar
William Stein committed
2357

2358 2359 2360 2361 2362 2363 2364 2365
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
2366

2367 2368
calling_convention_words = cython.declare(
    set, set(["__stdcall", "__cdecl", "__fastcall"]))
2369

Stefan Behnel's avatar
Stefan Behnel committed
2370

2371
def p_c_complex_base_type(s, templates = None):
William Stein's avatar
William Stein committed
2372 2373 2374
    # s.sy == '('
    pos = s.position()
    s.next()
Stefan Behnel's avatar
Stefan Behnel committed
2375 2376 2377 2378
    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)
2379 2380 2381 2382 2383 2384
    if s.sy == ',':
        components = [type_node]
        while s.sy == ',':
            s.next()
            if s.sy == ')':
                break
Stefan Behnel's avatar
Stefan Behnel committed
2385 2386 2387 2388
            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))
2389 2390 2391
        type_node = Nodes.CTupleBaseTypeNode(pos, components = components)

    s.expect(')')
2392 2393 2394 2395 2396 2397
    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
2398

William Stein's avatar
William Stein committed
2399

Danilo Freitas's avatar
Danilo Freitas committed
2400
def p_c_simple_base_type(s, self_flag, nonempty, templates = None):
2401
    #print "p_c_simple_base_type: self_flag =", self_flag, nonempty
William Stein's avatar
William Stein committed
2402 2403 2404
    is_basic = 0
    signed = 1
    longness = 0
2405
    complex = 0
William Stein's avatar
William Stein committed
2406
    module_path = []
2407
    pos = s.position()
2408 2409
    if not s.sy == 'IDENT':
        error(pos, "Expected an identifier, found '%s'" % s.sy)
Robert Bradshaw's avatar
Robert Bradshaw committed
2410 2411 2412 2413 2414
    if s.systring == 'const':
        s.next()
        base_type = p_c_base_type(s,
            self_flag = self_flag, nonempty = nonempty, templates = templates)
        return Nodes.CConstTypeNode(pos, base_type = base_type)
William Stein's avatar
William Stein committed
2415 2416 2417
    if looking_at_base_type(s):
        #print "p_c_simple_base_type: looking_at_base_type at", s.position()
        is_basic = 1
2418 2419
        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
2420 2421 2422
            name = s.systring
            s.next()
        else:
2423 2424 2425 2426 2427
            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
2428
                name = 'int'  # long [int], short [int], long [int] complex, etc.
2429 2430 2431
        if s.sy == 'IDENT' and s.systring == 'complex':
            complex = 1
            s.next()
2432 2433
    elif looking_at_dotted_name(s):
        #print "p_c_simple_base_type: looking_at_type_name at", s.position()
2434
        name = s.systring
2435 2436 2437 2438 2439
        s.next()
        while s.sy == '.':
            module_path.append(name)
            s.next()
            name = p_ident(s)
2440
    else:
2441 2442 2443
        name = s.systring
        s.next()
        if nonempty and s.sy != 'IDENT':
2444
            # Make sure this is not a declaration of a variable or function.
2445 2446
            if s.sy == '(':
                s.next()
2447 2448
                if (s.sy == '*' or s.sy == '**' or s.sy == '&'
                        or (s.sy == 'IDENT' and s.systring in calling_convention_words)):
2449 2450 2451 2452 2453
                    s.put_back('(', '(')
                else:
                    s.put_back('(', '(')
                    s.put_back('IDENT', name)
                    name = None
Robert Bradshaw's avatar
Robert Bradshaw committed
2454
            elif s.sy not in ('*', '**', '[', '&'):
2455 2456
                s.put_back('IDENT', name)
                name = None
Danilo Freitas's avatar
Danilo Freitas committed
2457

2458
    type_node = Nodes.CSimpleBaseTypeNode(pos,
William Stein's avatar
William Stein committed
2459 2460
        name = name, module_path = module_path,
        is_basic_c_type = is_basic, signed = signed,
2461
        complex = complex, longness = longness,
Danilo Freitas's avatar
Danilo Freitas committed
2462
        is_self_arg = self_flag, templates = templates)
William Stein's avatar
William Stein committed
2463

2464
    #    declarations here.
2465
    if s.sy == '[':
2466
        if is_memoryviewslice_access(s):
Dag Sverre Seljebotn's avatar
Dag Sverre Seljebotn committed
2467
            type_node = p_memoryviewslice_access(s, type_node)
2468 2469
        else:
            type_node = p_buffer_or_template(s, type_node, templates)
2470

Robert Bradshaw's avatar
Robert Bradshaw committed
2471 2472 2473 2474
    if s.sy == '.':
        s.next()
        name = p_ident(s)
        type_node = Nodes.CNestedBaseTypeNode(pos, base_type = type_node, name = name)
2475

Robert Bradshaw's avatar
Robert Bradshaw committed
2476
    return type_node
2477

2478
def p_buffer_or_template(s, base_type_node, templates):
2479 2480 2481
    # s.sy == '['
    pos = s.position()
    s.next()
2482 2483
    # Note that buffer_positional_options_count=1, so the only positional argument is dtype.
    # For templated types, all parameters are types.
2484
    positional_args, keyword_args = (
2485
        p_positional_and_keyword_args(s, (']',), templates)
2486 2487
    )
    s.expect(']')
2488

2489 2490
    if s.sy == '[':
        base_type_node = p_buffer_or_template(s, base_type_node, templates)
2491 2492 2493 2494 2495 2496

    keyword_dict = ExprNodes.DictNode(pos,
        key_value_pairs = [
            ExprNodes.DictItemNode(pos=key.pos, key=key, value=value)
            for key, value in keyword_args
        ])
2497
    result = Nodes.TemplatedTypeNode(pos,
2498 2499
        positional_args = positional_args,
        keyword_args = keyword_dict,
2500
        base_type_node = base_type_node)
2501
    return result
2502

2503 2504 2505 2506 2507 2508
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:
2509 2510 2511
        # declaration of either memoryview slice or buffer.
        if is_memoryviewslice_access(s):
            return p_memoryviewslice_access(s, base_type_node)
2512
        else:
Mark Florisson's avatar
Mark Florisson committed
2513 2514
            return p_buffer_or_template(s, base_type_node, None)
            # return p_buffer_access(s, base_type_node)
2515
    elif not empty and not nonempty:
2516 2517 2518 2519 2520
        # 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)
2521 2522 2523
        else:
            return base_type_node

2524
def is_memoryviewslice_access(s):
2525
    # s.sy == '['
2526
    # a memoryview slice declaration is distinguishable from a buffer access
2527
    # declaration by the first entry in the bracketed list.  The buffer will
2528
    # not have an unnested colon in the first entry; the memoryview slice will.
2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539
    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

2540
    for sv in saved[::-1]:
2541 2542 2543 2544
        s.put_back(*sv)

    return retval

2545
def p_memoryviewslice_access(s, base_type_node):
2546 2547 2548
    # s.sy == '['
    pos = s.position()
    s.next()
2549
    subscripts, _ = p_subscript_list(s)
2550 2551 2552 2553 2554 2555
    # 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)
2556
    result = Nodes.MemoryViewSliceTypeNode(pos,
2557 2558 2559
            base_type_node = base_type_node,
            axes = indexes)
    return result
2560

2561
def looking_at_name(s):
2562 2563
    return s.sy == 'IDENT' and not s.systring in calling_convention_words

2564
def looking_at_expr(s):
2565
    if s.systring in base_type_start_words:
2566
        return False
2567 2568 2569 2570 2571
    elif s.sy == 'IDENT':
        is_type = False
        name = s.systring
        dotted_path = []
        s.next()
2572

2573 2574 2575 2576
        while s.sy == '.':
            s.next()
            dotted_path.append(s.systring)
            s.expect('IDENT')
2577

2578
        saved = s.sy, s.systring
2579 2580 2581
        if s.sy == 'IDENT':
            is_type = True
        elif s.sy == '*' or s.sy == '**':
2582
            s.next()
2583
            is_type = s.sy in (')', ']')
2584 2585 2586 2587 2588 2589 2590 2591 2592
            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)
2593

2594 2595 2596 2597
        dotted_path.reverse()
        for p in dotted_path:
            s.put_back('IDENT', p)
            s.put_back('.', '.')
2598

2599
        s.put_back('IDENT', name)
2600
        return not is_type and saved[0]
2601
    else:
2602
        return True
William Stein's avatar
William Stein committed
2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616

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
2617

2618 2619 2620 2621 2622 2623
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
2624
    result = looking_at_expr(s) == u'('
2625 2626 2627 2628
    if not result:
        s.start_line, s.start_col = position
    return result

2629 2630
basic_c_type_names = cython.declare(
    set, set(["void", "char", "int", "float", "double", "bint"]))
2631

2632
special_basic_c_types = cython.declare(dict, {
2633
    # name : (signed, longness)
2634
    "Py_UNICODE" : (0, 0),
Stefan Behnel's avatar
Stefan Behnel committed
2635
    "Py_UCS4"    : (0, 0),
Stefan Behnel's avatar
Stefan Behnel committed
2636
    "Py_hash_t"  : (2, 0),
2637
    "Py_ssize_t" : (2, 0),
2638
    "ssize_t"    : (2, 0),
2639
    "size_t"     : (0, 0),
2640
    "ptrdiff_t"  : (2, 0),
2641
})
William Stein's avatar
William Stein committed
2642

2643 2644
sign_and_longness_words = cython.declare(
    set, set(["short", "long", "signed", "unsigned"]))
William Stein's avatar
William Stein committed
2645

2646 2647 2648 2649 2650
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
2651

2652 2653
struct_enum_union = cython.declare(
    set, set(["struct", "union", "enum", "packed"]))
Mark Florisson's avatar
Mark Florisson committed
2654

William Stein's avatar
William Stein committed
2655 2656 2657 2658 2659 2660
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
2661 2662
        elif s.systring == 'signed':
            signed = 2
William Stein's avatar
William Stein committed
2663 2664 2665 2666 2667 2668 2669 2670
        elif s.systring == 'short':
            longness = -1
        elif s.systring == 'long':
            longness += 1
        s.next()
    return signed, longness

def p_opt_cname(s):
2671 2672 2673
    literal = p_opt_string_literal(s, 'u')
    if literal is not None:
        cname = EncodedString(literal)
Stefan Behnel's avatar
Stefan Behnel committed
2674
        cname.encoding = s.source_encoding
William Stein's avatar
William Stein committed
2675 2676 2677 2678
    else:
        cname = None
    return cname

2679 2680 2681
def p_c_declarator(s, ctx = Ctx(), empty = 0, is_type = 0, cmethod_flag = 0,
                   assignable = 0, nonempty = 0,
                   calling_convention_allowed = 0):
2682 2683
    # 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
2684 2685 2686
    # If cmethod_flag is true, then if this declarator declares
    # a function, it's a C method of an extension type.
    pos = s.position()
2687 2688
    if s.sy == '(':
        s.next()
2689
        if s.sy == ')' or looking_at_name(s):
2690
            base = Nodes.CNameDeclaratorNode(pos, name=s.context.intern_ustring(u""), cname=None)
2691
            result = p_c_func_declarator(s, pos, ctx, base, cmethod_flag)
2692
        else:
2693 2694 2695 2696
            result = p_c_declarator(s, ctx, empty = empty, is_type = is_type,
                                    cmethod_flag = cmethod_flag,
                                    nonempty = nonempty,
                                    calling_convention_allowed = 1)
2697 2698
            s.expect(')')
    else:
2699 2700
        result = p_c_simple_declarator(s, ctx, empty, is_type, cmethod_flag,
                                       assignable, nonempty)
Stefan Behnel's avatar
Stefan Behnel committed
2701
    if not calling_convention_allowed and result.calling_convention and s.sy != '(':
2702 2703 2704 2705 2706 2707 2708 2709
        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()
2710
            result = p_c_func_declarator(s, pos, ctx, result, cmethod_flag)
2711 2712 2713 2714 2715 2716
        cmethod_flag = 0
    return result

def p_c_array_declarator(s, base):
    pos = s.position()
    s.next() # '['
Stefan Behnel's avatar
Stefan Behnel committed
2717
    if s.sy != ']':
2718
        dim = p_testlist(s)
2719 2720 2721 2722 2723
    else:
        dim = None
    s.expect(']')
    return Nodes.CArrayDeclaratorNode(pos, base = base, dimension = dim)

root's avatar
root committed
2724
def p_c_func_declarator(s, pos, ctx, base, cmethod_flag):
2725
    #  Opening paren has already been skipped
2726 2727
    args = p_c_arg_list(s, ctx, cmethod_flag = cmethod_flag,
                        nonempty_declarators = 0)
2728 2729 2730 2731 2732
    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)
2733
    return Nodes.CFuncDeclaratorNode(pos,
2734 2735
        base = base, args = args, has_varargs = ellipsis,
        exception_value = exc_val, exception_check = exc_check,
2736
        nogil = nogil or ctx.nogil or with_gil, with_gil = with_gil)
2737

2738
supported_overloaded_operators = cython.declare(set, set([
2739
    '+', '-', '*', '/', '%',
Robert Bradshaw's avatar
Robert Bradshaw committed
2740
    '++', '--', '~', '|', '&', '^', '<<', '>>', ',',
Robert Bradshaw's avatar
Robert Bradshaw committed
2741
    '==', '!=', '>=', '>', '<=', '<',
2742 2743
    '[]', '()', '!', '=',
    'bool',
2744
]))
2745

2746 2747
def p_c_simple_declarator(s, ctx, empty, is_type, cmethod_flag,
                          assignable, nonempty):
2748 2749
    pos = s.position()
    calling_convention = p_calling_convention(s)
William Stein's avatar
William Stein committed
2750 2751
    if s.sy == '*':
        s.next()
2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764
        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)
2765
        result = Nodes.CPtrDeclaratorNode(pos,
William Stein's avatar
William Stein committed
2766 2767 2768
            base = base)
    elif s.sy == '**': # scanner returns this as a single token
        s.next()
2769 2770 2771
        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
2772 2773 2774
        result = Nodes.CPtrDeclaratorNode(pos,
            base = Nodes.CPtrDeclaratorNode(pos,
                base = base))
Robert Bradshaw's avatar
merge  
Robert Bradshaw committed
2775
    elif s.sy == '&':
2776 2777 2778 2779 2780
        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
2781
    else:
2782 2783
        rhs = None
        if s.sy == 'IDENT':
2784
            name = s.systring
2785 2786
            if empty:
                error(s.position(), "Declarator should be empty")
William Stein's avatar
William Stein committed
2787
            s.next()
2788
            cname = p_opt_cname(s)
2789
            if name != 'operator' and s.sy == '=' and assignable:
2790
                s.next()
2791
                rhs = p_test(s)
William Stein's avatar
William Stein committed
2792
        else:
2793 2794 2795 2796
            if nonempty:
                error(s.position(), "Empty declarator")
            name = ""
            cname = None
2797
        if cname is None and ctx.namespace is not None and nonempty:
2798
            cname = ctx.namespace + "::" + name
2799
        if name == 'operator' and ctx.visibility == 'extern' and nonempty:
2800
            op = s.sy
2801
            if [1 for c in op if c in '+-*/<=>!%&|([^~,']:
2802
                s.next()
2803 2804 2805 2806 2807 2808 2809
                # Handle diphthong operators.
                if op == '(':
                    s.expect(')')
                    op = '()'
                elif op == '[':
                    s.expect(']')
                    op = '[]'
Stefan Behnel's avatar
Stefan Behnel committed
2810 2811
                elif op in ('-', '+', '|', '&') and s.sy == op:
                    op *= 2       # ++, --, ...
2812
                    s.next()
Stefan Behnel's avatar
Stefan Behnel committed
2813 2814
                elif s.sy == '=':
                    op += s.sy    # +=, -=, ...
2815 2816
                    s.next()
                if op not in supported_overloaded_operators:
2817 2818 2819
                    s.error("Overloading operator '%s' not yet supported." % op,
                            fatal=False)
                name += op
2820 2821 2822 2823 2824 2825 2826
            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()
2827
        result = Nodes.CNameDeclaratorNode(pos,
Robert Bradshaw's avatar
Robert Bradshaw committed
2828
            name = name, cname = cname, default = rhs)
2829
    result.calling_convention = calling_convention
William Stein's avatar
William Stein committed
2830 2831
    return result

2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846
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
2847 2848 2849 2850 2851 2852 2853 2854
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
2855 2856 2857
        elif s.sy == '+':
            exc_check = '+'
            s.next()
Robert Bradshaw's avatar
Robert Bradshaw committed
2858 2859 2860 2861
            if s.sy == 'IDENT':
                name = s.systring
                s.next()
                exc_val = p_name(s, name)
William Stein's avatar
William Stein committed
2862 2863 2864 2865
        else:
            if s.sy == '?':
                exc_check = 1
                s.next()
2866
            exc_val = p_test(s)
William Stein's avatar
William Stein committed
2867 2868
    return exc_val, exc_check

2869
c_arg_list_terminators = cython.declare(set, set(['*', '**', '.', ')', ':']))
William Stein's avatar
William Stein committed
2870

2871
def p_c_arg_list(s, ctx = Ctx(), in_pyfunc = 0, cmethod_flag = 0,
2872
                 nonempty_declarators = 0, kw_only = 0, annotated = 1):
2873 2874
    #  Comma-separated list of C argument declarations, possibly empty.
    #  May have a trailing comma.
William Stein's avatar
William Stein committed
2875
    args = []
2876 2877
    is_self_arg = cmethod_flag
    while s.sy not in c_arg_list_terminators:
2878
        args.append(p_c_arg_decl(s, ctx, in_pyfunc, is_self_arg,
2879 2880
            nonempty = nonempty_declarators, kw_only = kw_only,
            annotated = annotated))
2881 2882 2883 2884
        if s.sy != ',':
            break
        s.next()
        is_self_arg = 0
William Stein's avatar
William Stein committed
2885 2886 2887 2888 2889 2890 2891 2892 2893
    return args

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

2894 2895
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
2896
    pos = s.position()
2897
    not_none = or_none = 0
William Stein's avatar
William Stein committed
2898
    default = None
2899
    annotation = None
2900 2901 2902 2903 2904 2905 2906 2907 2908
    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)
2909
    declarator = p_c_declarator(s, ctx, nonempty = nonempty)
2910 2911
    if s.sy in ('not', 'or') and not s.in_python_file:
        kind = s.sy
William Stein's avatar
William Stein committed
2912 2913 2914 2915 2916 2917
        s.next()
        if s.sy == 'IDENT' and s.systring == 'None':
            s.next()
        else:
            s.error("Expected 'None'")
        if not in_pyfunc:
2918 2919 2920
            error(pos, "'%s None' only allowed in Python functions" % kind)
        or_none = kind == 'or'
        not_none = kind == 'not'
2921 2922
    if annotated and s.sy == ':':
        s.next()
2923
        annotation = p_test(s)
William Stein's avatar
William Stein committed
2924 2925
    if s.sy == '=':
        s.next()
Stefan Behnel's avatar
Stefan Behnel committed
2926
        if 'pxd' in ctx.level:
2927 2928
            if s.sy not in ['*', '?']:
                error(pos, "default values cannot be specified in pxd files, use ? or *")
Robert Bradshaw's avatar
Robert Bradshaw committed
2929
            default = ExprNodes.BoolNode(1)
2930 2931
            s.next()
        else:
2932
            default = p_test(s)
William Stein's avatar
William Stein committed
2933 2934 2935 2936
    return Nodes.CArgDeclNode(pos,
        base_type = base_type,
        declarator = declarator,
        not_none = not_none,
2937
        or_none = or_none,
2938
        default = default,
2939
        annotation = annotation,
2940
        kw_only = kw_only)
William Stein's avatar
William Stein committed
2941

2942 2943 2944 2945 2946 2947 2948
def p_api(s):
    if s.sy == 'IDENT' and s.systring == 'api':
        s.next()
        return 1
    else:
        return 0

2949
def p_cdef_statement(s, ctx):
William Stein's avatar
William Stein committed
2950
    pos = s.position()
2951 2952 2953 2954 2955 2956 2957
    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
2958 2959
    elif s.sy == 'import':
        s.next()
2960
        return p_cdef_extern_block(s, pos, ctx)
2961
    elif p_nogil(s):
2962
        ctx.nogil = 1
2963 2964 2965 2966 2967 2968
        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")
2969
        return p_cdef_block(s, ctx)
William Stein's avatar
William Stein committed
2970
    elif s.sy == 'class':
2971
        if ctx.level not in ('module', 'module_pxd'):
William Stein's avatar
William Stein committed
2972
            error(pos, "Extension type definition not allowed here")
2973 2974
        if ctx.overridable:
            error(pos, "Extension types cannot be declared cpdef")
2975
        return p_c_class_definition(s, pos, ctx)
Robert Bradshaw's avatar
Robert Bradshaw committed
2976 2977
    elif s.sy == 'IDENT' and s.systring == 'cppclass':
        return p_cpp_class_definition(s, pos, ctx)
Mark Florisson's avatar
Mark Florisson committed
2978
    elif s.sy == 'IDENT' and s.systring in struct_enum_union:
2979
        if ctx.level not in ('module', 'module_pxd'):
William Stein's avatar
William Stein committed
2980
            error(pos, "C struct/union/enum definition not allowed here")
2981
        if ctx.overridable:
2982 2983
            if s.systring != 'enum':
                error(pos, "C struct/union cannot be declared cpdef")
Mark Florisson's avatar
Mark Florisson committed
2984 2985 2986
        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
2987
    else:
2988
        return p_c_func_or_var_declaration(s, pos, ctx)
2989

2990 2991
def p_cdef_block(s, ctx):
    return p_suite(s, ctx(cdef_flag = 1))
William Stein's avatar
William Stein committed
2992

2993
def p_cdef_extern_block(s, pos, ctx):
2994 2995
    if ctx.overridable:
        error(pos, "cdef extern blocks cannot be declared cpdef")
William Stein's avatar
William Stein committed
2996 2997 2998 2999 3000
    include_file = None
    s.expect('from')
    if s.sy == '*':
        s.next()
    else:
3001
        include_file = p_string_literal(s, 'u')[2]
3002
    ctx = ctx(cdef_flag = 1, visibility = 'extern')
3003 3004
    if s.systring == "namespace":
        s.next()
3005
        ctx.namespace = p_string_literal(s, 'u')[2]
3006 3007 3008
    if p_nogil(s):
        ctx.nogil = 1
    body = p_suite(s, ctx)
William Stein's avatar
William Stein committed
3009 3010
    return Nodes.CDefExternNode(pos,
        include_file = include_file,
3011
        body = body,
Robert Bradshaw's avatar
Robert Bradshaw committed
3012
        namespace = ctx.namespace)
William Stein's avatar
William Stein committed
3013

3014
def p_c_enum_definition(s, pos, ctx):
William Stein's avatar
William Stein committed
3015 3016 3017 3018 3019 3020
    # s.sy == ident 'enum'
    s.next()
    if s.sy == 'IDENT':
        name = s.systring
        s.next()
        cname = p_opt_cname(s)
3021 3022
        if cname is None and ctx.namespace is not None:
            cname = ctx.namespace + "::" + name
William Stein's avatar
William Stein committed
3023 3024 3025 3026 3027 3028
    else:
        name = None
        cname = None
    items = None
    s.expect(':')
    items = []
Stefan Behnel's avatar
Stefan Behnel committed
3029
    if s.sy != 'NEWLINE':
3030
        p_c_enum_line(s, ctx, items)
William Stein's avatar
William Stein committed
3031 3032 3033 3034
    else:
        s.next() # 'NEWLINE'
        s.expect_indent()
        while s.sy not in ('DEDENT', 'EOF'):
3035
            p_c_enum_line(s, ctx, items)
William Stein's avatar
William Stein committed
3036
        s.expect_dedent()
3037 3038 3039
    return Nodes.CEnumDefNode(
        pos, name = name, cname = cname, items = items,
        typedef_flag = ctx.typedef_flag, visibility = ctx.visibility,
3040
        create_wrapper = ctx.overridable,
3041
        api = ctx.api, in_pxd = ctx.level == 'module_pxd')
William Stein's avatar
William Stein committed
3042

3043
def p_c_enum_line(s, ctx, items):
Stefan Behnel's avatar
Stefan Behnel committed
3044
    if s.sy != 'pass':
3045
        p_c_enum_item(s, ctx, items)
William Stein's avatar
William Stein committed
3046 3047 3048 3049
        while s.sy == ',':
            s.next()
            if s.sy in ('NEWLINE', 'EOF'):
                break
3050
            p_c_enum_item(s, ctx, items)
William Stein's avatar
William Stein committed
3051 3052 3053 3054
    else:
        s.next()
    s.expect_newline("Syntax error in enum item list")

3055
def p_c_enum_item(s, ctx, items):
William Stein's avatar
William Stein committed
3056 3057 3058
    pos = s.position()
    name = p_ident(s)
    cname = p_opt_cname(s)
3059 3060
    if cname is None and ctx.namespace is not None:
        cname = ctx.namespace + "::" + name
William Stein's avatar
William Stein committed
3061 3062 3063
    value = None
    if s.sy == '=':
        s.next()
3064
        value = p_test(s)
3065
    items.append(Nodes.CEnumDefItemNode(pos,
William Stein's avatar
William Stein committed
3066 3067
        name = name, cname = cname, value = value))

3068
def p_c_struct_or_union_definition(s, pos, ctx):
3069 3070 3071 3072
    packed = False
    if s.systring == 'packed':
        packed = True
        s.next()
Dag Sverre Seljebotn's avatar
Dag Sverre Seljebotn committed
3073
        if s.sy != 'IDENT' or s.systring != 'struct':
3074
            s.expected('struct')
William Stein's avatar
William Stein committed
3075 3076 3077 3078 3079
    # s.sy == ident 'struct' or 'union'
    kind = s.systring
    s.next()
    name = p_ident(s)
    cname = p_opt_cname(s)
3080 3081
    if cname is None and ctx.namespace is not None:
        cname = ctx.namespace + "::" + name
William Stein's avatar
William Stein committed
3082 3083 3084 3085 3086 3087
    attributes = None
    if s.sy == ':':
        s.next()
        s.expect('NEWLINE')
        s.expect_indent()
        attributes = []
3088
        body_ctx = Ctx()
Stefan Behnel's avatar
Stefan Behnel committed
3089 3090
        while s.sy != 'DEDENT':
            if s.sy != 'pass':
William Stein's avatar
William Stein committed
3091
                attributes.append(
3092
                    p_c_func_or_var_declaration(s, s.position(), body_ctx))
William Stein's avatar
William Stein committed
3093 3094 3095 3096 3097 3098
            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
3099
    return Nodes.CStructOrUnionDefNode(pos,
William Stein's avatar
William Stein committed
3100
        name = name, cname = cname, kind = kind, attributes = attributes,
3101
        typedef_flag = ctx.typedef_flag, visibility = ctx.visibility,
3102
        api = ctx.api, in_pxd = ctx.level == 'module_pxd', packed = packed)
William Stein's avatar
William Stein committed
3103

Mark Florisson's avatar
Mark Florisson committed
3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143
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
3144 3145 3146 3147 3148
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
3149
        if prev_visibility != 'private' and visibility != prev_visibility:
William Stein's avatar
William Stein committed
3150
            s.error("Conflicting visibility options '%s' and '%s'"
3151
                % (prev_visibility, visibility), fatal=False)
William Stein's avatar
William Stein committed
3152 3153
        s.next()
    return visibility
3154

3155
def p_c_modifiers(s):
3156
    if s.sy == 'IDENT' and s.systring in ('inline',):
3157
        modifier = s.systring
3158
        s.next()
3159 3160
        return [modifier] + p_c_modifiers(s)
    return []
William Stein's avatar
William Stein committed
3161

3162 3163
def p_c_func_or_var_declaration(s, pos, ctx):
    cmethod_flag = ctx.level in ('c_class', 'c_class_pxd')
3164
    modifiers = p_c_modifiers(s)
Danilo Freitas's avatar
Danilo Freitas committed
3165
    base_type = p_c_base_type(s, nonempty = 1, templates = ctx.templates)
3166 3167 3168
    declarator = p_c_declarator(s, ctx, cmethod_flag = cmethod_flag,
                                assignable = 1, nonempty = 1)
    declarator.overridable = ctx.overridable
Robert Bradshaw's avatar
Robert Bradshaw committed
3169 3170 3171 3172 3173
    if s.sy == 'IDENT' and s.systring == 'const' and ctx.level == 'cpp_class':
        s.next()
        is_const_method = 1
    else:
        is_const_method = 0
William Stein's avatar
William Stein committed
3174
    if s.sy == ':':
3175
        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
3176
            s.error("C function definition not allowed here")
3177
        doc, suite = p_suite_with_docstring(s, Ctx(level='function'))
William Stein's avatar
William Stein committed
3178
        result = Nodes.CFuncDefNode(pos,
3179
            visibility = ctx.visibility,
William Stein's avatar
William Stein committed
3180
            base_type = base_type,
3181
            declarator = declarator,
3182
            body = suite,
3183
            doc = doc,
3184
            modifiers = modifiers,
3185
            api = ctx.api,
Robert Bradshaw's avatar
Robert Bradshaw committed
3186 3187
            overridable = ctx.overridable,
            is_const_method = is_const_method)
William Stein's avatar
William Stein committed
3188
    else:
Stefan Behnel's avatar
Stefan Behnel committed
3189
        #if api:
3190
        #    s.error("'api' not allowed with variable declaration")
3191 3192
        if is_const_method:
            declarator.is_const_method = is_const_method
William Stein's avatar
William Stein committed
3193 3194 3195 3196 3197
        declarators = [declarator]
        while s.sy == ',':
            s.next()
            if s.sy == 'NEWLINE':
                break
3198 3199
            declarator = p_c_declarator(s, ctx, cmethod_flag = cmethod_flag,
                                        assignable = 1, nonempty = 1)
William Stein's avatar
William Stein committed
3200
            declarators.append(declarator)
3201
        doc_line = s.start_line + 1
3202
        s.expect_newline("Syntax error in C variable declaration", ignore_semicolon=True)
Stefan Behnel's avatar
Stefan Behnel committed
3203
        if ctx.level in ('c_class', 'c_class_pxd') and s.start_line == doc_line:
3204 3205 3206
            doc = p_doc_string(s)
        else:
            doc = None
3207
        result = Nodes.CVarDefNode(pos,
3208 3209
            visibility = ctx.visibility,
            base_type = base_type,
3210
            declarators = declarators,
3211
            in_pxd = ctx.level in ('module_pxd', 'c_class_pxd'),
3212
            doc = doc,
3213
            api = ctx.api,
3214
            modifiers = modifiers,
3215
            overridable = ctx.overridable)
William Stein's avatar
William Stein committed
3216 3217
    return result

3218
def p_ctypedef_statement(s, ctx):
William Stein's avatar
William Stein committed
3219 3220 3221
    # s.sy == 'ctypedef'
    pos = s.position()
    s.next()
3222
    visibility = p_visibility(s, ctx.visibility)
3223
    api = p_api(s)
3224
    ctx = ctx(typedef_flag = 1, visibility = visibility)
3225 3226
    if api:
        ctx.api = 1
William Stein's avatar
William Stein committed
3227
    if s.sy == 'class':
3228
        return p_c_class_definition(s, pos, ctx)
Mark Florisson's avatar
Mark Florisson committed
3229 3230 3231 3232
    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
3233
    else:
3234
        base_type = p_c_base_type(s, nonempty = 1)
3235
        declarator = p_c_declarator(s, ctx, is_type = 1, nonempty = 1)
3236
        s.expect_newline("Syntax error in ctypedef statement", ignore_semicolon=True)
3237 3238
        return Nodes.CTypeDefNode(
            pos, base_type = base_type,
Robert Bradshaw's avatar
Robert Bradshaw committed
3239
            declarator = declarator,
3240
            visibility = visibility, api = api,
3241
            in_pxd = ctx.level == 'module_pxd')
William Stein's avatar
William Stein committed
3242

3243 3244
def p_decorators(s):
    decorators = []
3245
    while s.sy == '@':
3246 3247
        pos = s.position()
        s.next()
3248 3249
        decstring = p_dotted_name(s, as_allowed=0)[2]
        names = decstring.split('.')
3250
        decorator = ExprNodes.NameNode(pos, name=s.context.intern_ustring(names[0]))
3251
        for name in names[1:]:
3252 3253
            decorator = ExprNodes.AttributeNode(
                pos, attribute=s.context.intern_ustring(name), obj=decorator)
3254 3255 3256 3257 3258 3259
        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

3260 3261

def p_def_statement(s, decorators=None, is_async_def=False):
William Stein's avatar
William Stein committed
3262 3263
    # s.sy == 'def'
    pos = s.position()
3264 3265 3266
    # 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
3267
    s.next()
3268
    name = p_ident(s)
3269
    s.expect('(')
Stefan Behnel's avatar
Stefan Behnel committed
3270 3271 3272
    args, star_arg, starstar_arg = p_varargslist(s, terminator=')')
    s.expect(')')
    if p_nogil(s):
3273
        error(pos, "Python function cannot be declared nogil")
3274 3275 3276
    return_type_annotation = None
    if s.sy == '->':
        s.next()
3277
        return_type_annotation = p_test(s)
3278

3279
    doc, body = p_suite_with_docstring(s, Ctx(level='function'))
3280
    if is_async_def:
3281
        s.exit_async()
3282

3283 3284 3285 3286 3287
    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
3288

3289 3290 3291
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
3292 3293 3294 3295
    star_arg = None
    starstar_arg = None
    if s.sy == '*':
        s.next()
3296
        if s.sy == 'IDENT':
3297
            star_arg = p_py_arg_decl(s, annotated=annotated)
William Stein's avatar
William Stein committed
3298 3299
        if s.sy == ',':
            s.next()
3300
            args.extend(p_c_arg_list(s, in_pyfunc = 1,
3301
                nonempty_declarators = 1, kw_only = 1, annotated = annotated))
Stefan Behnel's avatar
Stefan Behnel committed
3302
        elif s.sy != terminator:
3303 3304
            s.error("Syntax error in Python function argument list")
    if s.sy == '**':
William Stein's avatar
William Stein committed
3305
        s.next()
3306
        starstar_arg = p_py_arg_decl(s, annotated=annotated)
3307 3308
    if s.sy == ',':
        s.next()
Stefan Behnel's avatar
Stefan Behnel committed
3309
    return (args, star_arg, starstar_arg)
William Stein's avatar
William Stein committed
3310

3311
def p_py_arg_decl(s, annotated = 1):
William Stein's avatar
William Stein committed
3312 3313
    pos = s.position()
    name = p_ident(s)
3314
    annotation = None
3315
    if annotated and s.sy == ':':
3316
        s.next()
3317
        annotation = p_test(s)
3318
    return Nodes.PyArgDeclNode(pos, name = name, annotation = annotation)
William Stein's avatar
William Stein committed
3319

3320

3321
def p_class_statement(s, decorators):
William Stein's avatar
William Stein committed
3322 3323 3324
    # s.sy == 'class'
    pos = s.position()
    s.next()
3325 3326
    class_name = EncodedString(p_ident(s))
    class_name.encoding = s.source_encoding  # FIXME: why is this needed?
3327 3328
    arg_tuple = None
    keyword_dict = None
William Stein's avatar
William Stein committed
3329
    if s.sy == '(':
3330 3331
        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)
3332 3333
    if arg_tuple is None:
        # XXX: empty arg_tuple
3334 3335
        arg_tuple = ExprNodes.TupleNode(pos, args=[])
    doc, body = p_suite_with_docstring(s, Ctx(level='class'))
3336 3337 3338 3339 3340 3341
    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
3342

3343

3344
def p_c_class_definition(s, pos,  ctx):
William Stein's avatar
William Stein committed
3345 3346 3347 3348 3349 3350 3351 3352
    # 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)
3353
    if module_path and ctx.visibility != 'extern':
William Stein's avatar
William Stein committed
3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370
        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
    base_class_module = None
    base_class_name = None
    if s.sy == '(':
        s.next()
        base_class_path = [p_ident(s)]
        while s.sy == '.':
            s.next()
            base_class_path.append(p_ident(s))
        if s.sy == ',':
3371
            s.error("C class may only have one base class", fatal=False)
William Stein's avatar
William Stein committed
3372 3373 3374 3375
        s.expect(')')
        base_class_module = ".".join(base_class_path[:-1])
        base_class_name = base_class_path[-1]
    if s.sy == '[':
3376 3377
        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")
William Stein's avatar
William Stein committed
3378 3379
        objstruct_name, typeobj_name = p_c_class_options(s)
    if s.sy == ':':
3380
        if ctx.level == 'module_pxd':
William Stein's avatar
William Stein committed
3381 3382 3383
            body_level = 'c_class_pxd'
        else:
            body_level = 'c_class'
3384
        doc, body = p_suite_with_docstring(s, Ctx(level=body_level))
William Stein's avatar
William Stein committed
3385 3386 3387 3388
    else:
        s.expect_newline("Syntax error in C class definition")
        doc = None
        body = None
3389
    if ctx.visibility == 'extern':
William Stein's avatar
William Stein committed
3390 3391 3392 3393
        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")
3394
    elif ctx.visibility == 'public':
William Stein's avatar
William Stein committed
3395 3396 3397 3398
        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")
3399 3400
    elif ctx.visibility == 'private':
        if ctx.api:
3401 3402 3403 3404
            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")
3405
    else:
3406
        error(pos, "Invalid class visibility '%s'" % ctx.visibility)
William Stein's avatar
William Stein committed
3407
    return Nodes.CClassDefNode(pos,
3408 3409 3410
        visibility = ctx.visibility,
        typedef_flag = ctx.typedef_flag,
        api = ctx.api,
William Stein's avatar
William Stein committed
3411 3412 3413 3414 3415 3416 3417
        module_name = ".".join(module_path),
        class_name = class_name,
        as_name = as_name,
        base_class_module = base_class_module,
        base_class_name = base_class_name,
        objstruct_name = objstruct_name,
        typeobj_name = typeobj_name,
3418
        in_pxd = ctx.level == 'module_pxd',
William Stein's avatar
William Stein committed
3419 3420 3421 3422 3423 3424 3425 3426
        doc = doc,
        body = body)

def p_c_class_options(s):
    objstruct_name = None
    typeobj_name = None
    s.expect('[')
    while 1:
Stefan Behnel's avatar
Stefan Behnel committed
3427
        if s.sy != 'IDENT':
William Stein's avatar
William Stein committed
3428 3429 3430 3431 3432 3433 3434
            break
        if s.systring == 'object':
            s.next()
            objstruct_name = p_ident(s)
        elif s.systring == 'type':
            s.next()
            typeobj_name = p_ident(s)
Stefan Behnel's avatar
Stefan Behnel committed
3435
        if s.sy != ',':
William Stein's avatar
William Stein committed
3436 3437 3438 3439 3440
            break
        s.next()
    s.expect(']', "Expected 'object' or 'type'")
    return objstruct_name, typeobj_name

3441

William Stein's avatar
William Stein committed
3442 3443
def p_property_decl(s):
    pos = s.position()
3444
    s.next()  # 'property'
William Stein's avatar
William Stein committed
3445
    name = p_ident(s)
3446 3447 3448 3449
    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
3450

3451 3452 3453 3454 3455 3456 3457
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)
3458
        s.expect_newline("Syntax error in string", ignore_semicolon=True)
3459 3460 3461 3462
        return Nodes.ExprStatNode(pos, expr=string_node)
    return None


3463 3464 3465 3466
def p_doc_string(s):
    if s.sy == 'BEGIN_STRING':
        pos = s.position()
        kind, bytes_result, unicode_result = p_cat_string_literal(s)
3467
        s.expect_newline("Syntax error in doc string", ignore_semicolon=True)
3468 3469 3470 3471 3472 3473 3474
        if kind in ('u', ''):
            return unicode_result
        warning(pos, "Python 3 requires docstrings to be unicode strings")
        return bytes_result
    else:
        return None

3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510

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


3511 3512
def p_code(s, level=None, ctx=Ctx):
    body = p_statement_list(s, ctx(level = level), first_statement = 1)
3513 3514 3515 3516
    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
3517

3518 3519
_match_compiler_directive_comment = cython.declare(object, re.compile(
    r"^#\s*cython\s*:\s*((\w|[.])+\s*=.*)$").match)
3520 3521

def p_compiler_directive_comments(s):
3522
    result = {}
3523
    while s.sy == 'commentline':
3524
        m = _match_compiler_directive_comment(s.systring)
3525
        if m:
3526
            directives = m.group(1).strip()
3527
            try:
3528 3529
                result.update(Options.parse_directive_list(
                    directives, ignore_unknown=True))
3530
            except ValueError as e:
3531
                s.error(e.args[0], fatal=False)
3532 3533 3534
        s.next()
    return result

3535
def p_module(s, pxd, full_module_name, ctx=Ctx):
William Stein's avatar
William Stein committed
3536
    pos = s.position()
3537

3538
    directive_comments = p_compiler_directive_comments(s)
3539 3540
    s.parse_comments = False

3541
    if 'language_level' in directive_comments:
Stefan Behnel's avatar
Stefan Behnel committed
3542
        s.context.set_language_level(directive_comments['language_level'])
3543

3544
    doc = p_doc_string(s)
William Stein's avatar
William Stein committed
3545 3546 3547 3548
    if pxd:
        level = 'module_pxd'
    else:
        level = 'module'
3549

3550
    body = p_statement_list(s, ctx(level=level), first_statement = 1)
Stefan Behnel's avatar
Stefan Behnel committed
3551
    if s.sy != 'EOF':
William Stein's avatar
William Stein committed
3552 3553
        s.error("Syntax error in statement [%s,%s]" % (
            repr(s.sy), repr(s.systring)))
3554 3555
    return ModuleNode(pos, doc = doc, body = body,
                      full_module_name = full_module_name,
3556
                      directive_comments = directive_comments)
William Stein's avatar
William Stein committed
3557

3558 3559 3560 3561 3562 3563 3564 3565 3566 3567
def p_template_definition(s):
    name = p_ident(s)
    if s.sy == '=':
        s.expect('=')
        s.expect('*')
        required = False
    else:
        required = True
    return name, required

3568 3569 3570 3571 3572
def p_cpp_class_definition(s, pos,  ctx):
    # s.sy == 'cppclass'
    s.next()
    module_path = []
    class_name = p_ident(s)
3573 3574 3575
    cname = p_opt_cname(s)
    if cname is None and ctx.namespace is not None:
        cname = ctx.namespace + "::" + class_name
3576
    if s.sy == '.':
3577
        error(pos, "Qualified class name not allowed C++ class")
Danilo Freitas's avatar
Danilo Freitas committed
3578 3579
    if s.sy == '[':
        s.next()
3580
        templates = [p_template_definition(s)]
Danilo Freitas's avatar
Danilo Freitas committed
3581 3582
        while s.sy == ',':
            s.next()
3583
            templates.append(p_template_definition(s))
Danilo Freitas's avatar
Danilo Freitas committed
3584
        s.expect(']')
3585
        template_names = [name for name, required in templates]
3586 3587
    else:
        templates = None
3588
        template_names = None
3589
    if s.sy == '(':
3590
        s.next()
3591
        base_classes = [p_c_base_type(s, templates = template_names)]
3592
        while s.sy == ',':
3593
            s.next()
3594
            base_classes.append(p_c_base_type(s, templates = template_names))
3595
        s.expect(')')
3596 3597
    else:
        base_classes = []
3598
    if s.sy == '[':
3599
        error(s.position(), "Name options not allowed for C++ class")
3600
    nogil = p_nogil(s)
3601
    if s.sy == ':':
3602 3603 3604 3605
        s.next()
        s.expect('NEWLINE')
        s.expect_indent()
        attributes = []
3606
        body_ctx = Ctx(visibility = ctx.visibility, level='cpp_class', nogil=nogil or ctx.nogil)
3607
        body_ctx.templates = template_names
3608
        while s.sy != 'DEDENT':
3609 3610
            if s.sy != 'pass':
                attributes.append(p_cpp_class_attribute(s, body_ctx))
3611 3612 3613 3614
            else:
                s.next()
                s.expect_newline("Expected a newline")
        s.expect_dedent()
3615
    else:
3616
        attributes = None
3617 3618 3619
        s.expect_newline("Syntax error in C++ class definition")
    return Nodes.CppClassNode(pos,
        name = class_name,
3620
        cname = cname,
3621
        base_classes = base_classes,
3622 3623
        visibility = ctx.visibility,
        in_pxd = ctx.level == 'module_pxd',
Danilo Freitas's avatar
Danilo Freitas committed
3624 3625
        attributes = attributes,
        templates = templates)
3626

3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643
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)
    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

3644

William Stein's avatar
William Stein committed
3645 3646 3647 3648 3649 3650
#----------------------------------------------
#
#   Debugging
#
#----------------------------------------------

Stefan Behnel's avatar
Stefan Behnel committed
3651
def print_parse_tree(f, node, level, key = None):
William Stein's avatar
William Stein committed
3652 3653 3654 3655 3656 3657
    ind = "  " * level
    if node:
        f.write(ind)
        if key:
            f.write("%s: " % key)
        t = type(node)
Stefan Behnel's avatar
Stefan Behnel committed
3658
        if t is tuple:
William Stein's avatar
William Stein committed
3659
            f.write("(%s @ %s\n" % (node[0], node[1]))
3660
            for i in range(2, len(node)):
William Stein's avatar
William Stein committed
3661 3662 3663
                print_parse_tree(f, node[i], level+1)
            f.write("%s)\n" % ind)
            return
3664
        elif isinstance(node, Nodes.Node):
William Stein's avatar
William Stein committed
3665 3666 3667 3668 3669 3670
            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
3671
                if name != 'tag' and name != 'pos':
William Stein's avatar
William Stein committed
3672 3673
                    print_parse_tree(f, value, level+1, name)
            return
Stefan Behnel's avatar
Stefan Behnel committed
3674
        elif t is list:
William Stein's avatar
William Stein committed
3675
            f.write("[\n")
3676
            for i in range(len(node)):
William Stein's avatar
William Stein committed
3677 3678 3679 3680
                print_parse_tree(f, node[i], level+1)
            f.write("%s]\n" % ind)
            return
    f.write("%s%s\n" % (ind, node))