1 #!/usr/bin/env python 2 3 """ 4 Common functions. 5 6 Copyright (C) 2007, 2008, 2009, 2010, 2011, 2012, 2013, 7 2014, 2015, 2016, 2017 Paul Boddie <paul@boddie.org.uk> 8 9 This program is free software; you can redistribute it and/or modify it under 10 the terms of the GNU General Public License as published by the Free Software 11 Foundation; either version 3 of the License, or (at your option) any later 12 version. 13 14 This program is distributed in the hope that it will be useful, but WITHOUT 15 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS 16 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more 17 details. 18 19 You should have received a copy of the GNU General Public License along with 20 this program. If not, see <http://www.gnu.org/licenses/>. 21 """ 22 23 from compiler.transformer import Transformer 24 from errors import InspectError 25 from os import listdir, makedirs, remove 26 from os.path import exists, isdir, join, split 27 from results import ConstantValueRef, LiteralSequenceRef, NameRef 28 import compiler.ast 29 30 class CommonOutput: 31 32 "Common output functionality." 33 34 def check_output(self): 35 36 "Check the existing output and remove it if irrelevant." 37 38 if not exists(self.output): 39 makedirs(self.output) 40 41 details = self.importer.get_cache_details() 42 recorded_details = self.get_output_details() 43 44 if recorded_details != details: 45 self.remove_output() 46 47 writefile(self.get_output_details_filename(), details) 48 49 def get_output_details_filename(self): 50 51 "Return the output details filename." 52 53 return join(self.output, "$details") 54 55 def get_output_details(self): 56 57 "Return details of the existing output." 58 59 details_filename = self.get_output_details_filename() 60 61 if not exists(details_filename): 62 return None 63 else: 64 return readfile(details_filename) 65 66 def remove_output(self, dirname=None): 67 68 "Remove the output." 69 70 dirname = dirname or self.output 71 72 for filename in listdir(dirname): 73 path = join(dirname, filename) 74 if isdir(path): 75 self.remove_output(path) 76 else: 77 remove(path) 78 79 class CommonModule: 80 81 "A common module representation." 82 83 def __init__(self, name, importer): 84 85 """ 86 Initialise this module with the given 'name' and an 'importer' which is 87 used to provide access to other modules when required. 88 """ 89 90 self.name = name 91 self.importer = importer 92 self.filename = None 93 94 # Inspection-related attributes. 95 96 self.astnode = None 97 self.encoding = None 98 self.iterators = {} 99 self.temp = {} 100 self.lambdas = {} 101 102 # Constants, literals and values. 103 104 self.constants = {} 105 self.constant_values = {} 106 self.literals = {} 107 self.literal_types = {} 108 109 # Nested namespaces. 110 111 self.namespace_path = [] 112 self.in_function = False 113 114 # Retain the assignment value expression and track invocations. 115 116 self.in_assignment = None 117 self.in_invocation = None 118 119 # Attribute chain state management. 120 121 self.attrs = [] 122 self.chain_assignment = [] 123 self.chain_invocation = [] 124 125 def __repr__(self): 126 return "CommonModule(%r, %r)" % (self.name, self.importer) 127 128 def parse_file(self, filename): 129 130 "Parse the file with the given 'filename', initialising attributes." 131 132 self.filename = filename 133 134 # Use the Transformer directly to obtain encoding information. 135 136 t = Transformer() 137 f = open(filename) 138 139 try: 140 self.astnode = t.parsesuite(f.read() + "\n") 141 self.encoding = t.encoding 142 finally: 143 f.close() 144 145 # Module-relative naming. 146 147 def get_global_path(self, name): 148 return "%s.%s" % (self.name, name) 149 150 def get_namespace_path(self): 151 return ".".join([self.name] + self.namespace_path) 152 153 def get_object_path(self, name): 154 return ".".join([self.name] + self.namespace_path + [name]) 155 156 def get_parent_path(self): 157 return ".".join([self.name] + self.namespace_path[:-1]) 158 159 # Namespace management. 160 161 def enter_namespace(self, name): 162 163 "Enter the namespace having the given 'name'." 164 165 self.namespace_path.append(name) 166 167 def exit_namespace(self): 168 169 "Exit the current namespace." 170 171 self.namespace_path.pop() 172 173 # Constant reference naming. 174 175 def get_constant_name(self, value, value_type, encoding=None): 176 177 """ 178 Add a new constant to the current namespace for 'value' with 179 'value_type'. 180 """ 181 182 path = self.get_namespace_path() 183 init_item(self.constants, path, dict) 184 return "$c%d" % add_counter_item(self.constants[path], (value, value_type, encoding)) 185 186 # Literal reference naming. 187 188 def get_literal_name(self): 189 190 "Add a new literal to the current namespace." 191 192 path = self.get_namespace_path() 193 init_item(self.literals, path, lambda: 0) 194 return "$C%d" % self.literals[path] 195 196 def next_literal(self): 197 self.literals[self.get_namespace_path()] += 1 198 199 # Temporary iterator naming. 200 201 def get_iterator_path(self): 202 return self.in_function and self.get_namespace_path() or self.name 203 204 def get_iterator_name(self): 205 path = self.get_iterator_path() 206 init_item(self.iterators, path, lambda: 0) 207 return "$i%d" % self.iterators[path] 208 209 def next_iterator(self): 210 self.iterators[self.get_iterator_path()] += 1 211 212 # Temporary variable naming. 213 214 def get_temporary_name(self): 215 path = self.get_namespace_path() 216 init_item(self.temp, path, lambda: 0) 217 return "$t%d" % self.temp[path] 218 219 def next_temporary(self): 220 self.temp[self.get_namespace_path()] += 1 221 222 # Arbitrary function naming. 223 224 def get_lambda_name(self): 225 path = self.get_namespace_path() 226 init_item(self.lambdas, path, lambda: 0) 227 name = "$l%d" % self.lambdas[path] 228 self.lambdas[path] += 1 229 return name 230 231 def reset_lambdas(self): 232 self.lambdas = {} 233 234 # Constant and literal recording. 235 236 def get_constant_value(self, value, literals=None): 237 238 """ 239 Encode the 'value' if appropriate, returning a value, a typename and any 240 encoding. 241 """ 242 243 if isinstance(value, unicode): 244 return value.encode("utf-8"), "unicode", self.encoding 245 246 # Attempt to convert plain strings to text. 247 248 elif isinstance(value, str) and self.encoding: 249 try: 250 return get_string_details(literals, self.encoding) 251 except UnicodeDecodeError: 252 pass 253 254 return value, value.__class__.__name__, None 255 256 def get_constant_reference(self, ref, value, encoding=None): 257 258 """ 259 Return a constant reference for the given 'ref' type and 'value', with 260 the optional 'encoding' applying to text values. 261 """ 262 263 constant_name = self.get_constant_name(value, ref.get_origin(), encoding) 264 265 # Return a reference for the constant. 266 267 objpath = self.get_object_path(constant_name) 268 name_ref = ConstantValueRef(constant_name, ref.instance_of(objpath), value) 269 270 # Record the value and type for the constant. 271 272 self._reserve_constant(objpath, name_ref.value, name_ref.get_origin(), encoding) 273 return name_ref 274 275 def reserve_constant(self, objpath, value, origin, encoding=None): 276 277 """ 278 Reserve a constant within 'objpath' with the given 'value' and having a 279 type with the given 'origin', with the optional 'encoding' applying to 280 text values. 281 """ 282 283 constant_name = self.get_constant_name(value, origin) 284 objpath = self.get_object_path(constant_name) 285 self._reserve_constant(objpath, value, origin, encoding) 286 287 def _reserve_constant(self, objpath, value, origin, encoding): 288 289 """ 290 Store a constant for 'objpath' with the given 'value' and 'origin', with 291 the optional 'encoding' applying to text values. 292 """ 293 294 self.constant_values[objpath] = value, origin, encoding 295 296 def get_literal_reference(self, name, ref, items, cls): 297 298 """ 299 Return a literal reference for the given type 'name', literal 'ref', 300 node 'items' and employing the given 'cls' as the class of the returned 301 reference object. 302 """ 303 304 # Construct an invocation using the items as arguments. 305 306 typename = "$L%s" % name 307 308 invocation = compiler.ast.CallFunc( 309 compiler.ast.Name(typename), 310 items 311 ) 312 313 # Get a name for the actual literal. 314 315 instname = self.get_literal_name() 316 self.next_literal() 317 318 # Record the type for the literal. 319 320 objpath = self.get_object_path(instname) 321 self.literal_types[objpath] = ref.get_origin() 322 323 # Return a wrapper for the invocation exposing the items. 324 325 return cls( 326 instname, 327 ref.instance_of(), 328 self.process_structure_node(invocation), 329 invocation.args 330 ) 331 332 # Node handling. 333 334 def process_structure(self, node): 335 336 """ 337 Within the given 'node', process the program structure. 338 339 During inspection, this will process global declarations, adjusting the 340 module namespace, and import statements, building a module dependency 341 hierarchy. 342 343 During translation, this will consult deduced program information and 344 output translated code. 345 """ 346 347 l = [] 348 for n in node.getChildNodes(): 349 l.append(self.process_structure_node(n)) 350 return l 351 352 def process_augassign_node(self, n): 353 354 "Process the given augmented assignment node 'n'." 355 356 op = operator_functions[n.op] 357 358 if isinstance(n.node, compiler.ast.Getattr): 359 target = compiler.ast.AssAttr(n.node.expr, n.node.attrname, "OP_ASSIGN") 360 elif isinstance(n.node, compiler.ast.Name): 361 target = compiler.ast.AssName(n.node.name, "OP_ASSIGN") 362 else: 363 target = n.node 364 365 assignment = compiler.ast.Assign( 366 [target], 367 compiler.ast.CallFunc( 368 compiler.ast.Name("$op%s" % op), 369 [n.node, n.expr])) 370 371 return self.process_structure_node(assignment) 372 373 def process_assignment_for_object(self, original_name, source): 374 375 """ 376 Return an assignment operation making 'original_name' refer to the given 377 'source'. 378 """ 379 380 assignment = compiler.ast.Assign( 381 [compiler.ast.AssName(original_name, "OP_ASSIGN")], 382 source 383 ) 384 385 return self.process_structure_node(assignment) 386 387 def process_assignment_node_items(self, n, expr): 388 389 """ 390 Process the given assignment node 'n' whose children are to be assigned 391 items of 'expr'. 392 """ 393 394 name_ref = self.process_structure_node(expr) 395 396 # Either unpack the items and present them directly to each assignment 397 # node. 398 399 if isinstance(name_ref, LiteralSequenceRef) and \ 400 self.process_literal_sequence_items(n, name_ref): 401 402 pass 403 404 # Or have the assignment nodes access each item via the sequence API. 405 406 else: 407 self.process_assignment_node_items_by_position(n, expr, name_ref) 408 409 def process_assignment_node_items_by_position(self, n, expr, name_ref): 410 411 """ 412 Process the given sequence assignment node 'n', converting the node to 413 the separate assignment of each target using positional access on a 414 temporary variable representing the sequence. Use 'expr' as the assigned 415 value and 'name_ref' as the reference providing any existing temporary 416 variable. 417 """ 418 419 assignments = [] 420 421 # Employ existing names to access the sequence. 422 # Literal sequences do not provide names of accessible objects. 423 424 if isinstance(name_ref, NameRef) and not isinstance(name_ref, LiteralSequenceRef): 425 temp = name_ref.name 426 427 # For other expressions, create a temporary name to reference the items. 428 429 else: 430 temp = self.get_temporary_name() 431 self.next_temporary() 432 433 assignments.append( 434 compiler.ast.Assign([compiler.ast.AssName(temp, "OP_ASSIGN")], expr) 435 ) 436 437 # Assign the items to the target nodes. 438 439 for i, node in enumerate(n.nodes): 440 assignments.append( 441 compiler.ast.Assign([node], compiler.ast.Subscript( 442 compiler.ast.Name(temp), "OP_APPLY", [compiler.ast.Const(i, str(i))])) 443 ) 444 445 return self.process_structure_node(compiler.ast.Stmt(assignments)) 446 447 def process_literal_sequence_items(self, n, name_ref): 448 449 """ 450 Process the given assignment node 'n', obtaining from the given 451 'name_ref' the items to be assigned to the assignment targets. 452 453 Return whether this method was able to process the assignment node as 454 a sequence of direct assignments. 455 """ 456 457 if len(n.nodes) == len(name_ref.items): 458 assigned_names, count = get_names_from_nodes(n.nodes) 459 accessed_names, _count = get_names_from_nodes(name_ref.items) 460 461 # Only assign directly between items if all assigned names are 462 # plain names (not attribute assignments), and if the assigned names 463 # do not appear in the accessed names. 464 465 if len(assigned_names) == count and \ 466 not assigned_names.intersection(accessed_names): 467 468 for node, item in zip(n.nodes, name_ref.items): 469 self.process_assignment_node(node, item) 470 471 return True 472 473 # Otherwise, use the position-based mechanism to obtain values. 474 475 else: 476 return False 477 else: 478 raise InspectError("In %s, item assignment needing %d items is given %d items." % ( 479 self.get_namespace_path(), len(n.nodes), len(name_ref.items))) 480 481 def process_compare_node(self, n): 482 483 """ 484 Process the given comparison node 'n', converting an operator sequence 485 from... 486 487 <expr1> <op1> <expr2> <op2> <expr3> 488 489 ...to... 490 491 <op1>(<expr1>, <expr2>) and <op2>(<expr2>, <expr3>) 492 """ 493 494 invocations = [] 495 last = n.expr 496 497 for op, op_node in n.ops: 498 op = operator_functions.get(op) 499 500 invocations.append(compiler.ast.CallFunc( 501 compiler.ast.Name("$op%s" % op), 502 [last, op_node])) 503 504 last = op_node 505 506 if len(invocations) > 1: 507 result = compiler.ast.And(invocations) 508 else: 509 result = invocations[0] 510 511 return self.process_structure_node(result) 512 513 def process_dict_node(self, node): 514 515 """ 516 Process the given dictionary 'node', returning a list of (key, value) 517 tuples. 518 """ 519 520 l = [] 521 for key, value in node.items: 522 l.append(( 523 self.process_structure_node(key), 524 self.process_structure_node(value))) 525 return l 526 527 def process_for_node(self, n): 528 529 """ 530 Generate attribute accesses for {n.list}.__iter__ and the next method on 531 the iterator, producing a replacement node for the original. 532 """ 533 534 node = compiler.ast.Stmt([ 535 536 # <next> = {n.list}.__iter__().next 537 538 compiler.ast.Assign( 539 [compiler.ast.AssName(self.get_iterator_name(), "OP_ASSIGN")], 540 compiler.ast.Getattr( 541 compiler.ast.CallFunc( 542 compiler.ast.Getattr(n.list, "__iter__"), 543 [] 544 ), "next")), 545 546 # try: 547 # while True: 548 # <var>... = <next>() 549 # ... 550 # except StopIteration: 551 # pass 552 553 compiler.ast.TryExcept( 554 compiler.ast.While( 555 compiler.ast.Name("True"), 556 compiler.ast.Stmt([ 557 compiler.ast.Assign( 558 [n.assign], 559 compiler.ast.CallFunc( 560 compiler.ast.Name(self.get_iterator_name()), 561 [] 562 )), 563 n.body]), 564 None), 565 [(compiler.ast.Name("StopIteration"), None, compiler.ast.Stmt([compiler.ast.Pass()]))], 566 None) 567 ]) 568 569 self.next_iterator() 570 self.process_structure_node(node) 571 572 def process_literal_sequence_node(self, n, name, ref, cls): 573 574 """ 575 Process the given literal sequence node 'n' as a function invocation, 576 with 'name' indicating the type of the sequence, and 'ref' being a 577 reference to the type. The 'cls' is used to instantiate a suitable name 578 reference. 579 """ 580 581 if name == "dict": 582 items = [] 583 for key, value in n.items: 584 items.append(compiler.ast.Tuple([key, value])) 585 else: # name in ("list", "tuple"): 586 items = n.nodes 587 588 return self.get_literal_reference(name, ref, items, cls) 589 590 def process_operator_node(self, n): 591 592 """ 593 Process the given operator node 'n' as an operator function invocation. 594 """ 595 596 op = operator_functions[n.__class__.__name__] 597 invocation = compiler.ast.CallFunc( 598 compiler.ast.Name("$op%s" % op), 599 list(n.getChildNodes()) 600 ) 601 return self.process_structure_node(invocation) 602 603 def process_print_node(self, n): 604 605 """ 606 Process the given print node 'n' as an invocation on a stream of the 607 form... 608 609 $print(dest, args, nl) 610 611 The special function name will be translated elsewhere. 612 """ 613 614 nl = isinstance(n, compiler.ast.Printnl) 615 invocation = compiler.ast.CallFunc( 616 compiler.ast.Name("$print"), 617 [n.dest or compiler.ast.Name("None"), 618 compiler.ast.List(list(n.nodes)), 619 nl and compiler.ast.Name("True") or compiler.ast.Name("False")] 620 ) 621 return self.process_structure_node(invocation) 622 623 def process_slice_node(self, n, expr=None): 624 625 """ 626 Process the given slice node 'n' as an operator function invocation. 627 """ 628 629 if n.flags == "OP_ASSIGN": op = "setslice" 630 elif n.flags == "OP_DELETE": op = "delslice" 631 else: op = "getslice" 632 633 invocation = compiler.ast.CallFunc( 634 compiler.ast.Name("$op%s" % op), 635 [n.expr, n.lower or compiler.ast.Name("None"), n.upper or compiler.ast.Name("None")] + 636 (expr and [expr] or []) 637 ) 638 639 # Fix parse tree structure. 640 641 if op == "delslice": 642 invocation = compiler.ast.Discard(invocation) 643 644 return self.process_structure_node(invocation) 645 646 def process_sliceobj_node(self, n): 647 648 """ 649 Process the given slice object node 'n' as a slice constructor. 650 """ 651 652 op = "slice" 653 invocation = compiler.ast.CallFunc( 654 compiler.ast.Name("$op%s" % op), 655 n.nodes 656 ) 657 return self.process_structure_node(invocation) 658 659 def process_subscript_node(self, n, expr=None): 660 661 """ 662 Process the given subscript node 'n' as an operator function invocation. 663 """ 664 665 if n.flags == "OP_ASSIGN": op = "setitem" 666 elif n.flags == "OP_DELETE": op = "delitem" 667 else: op = "getitem" 668 669 invocation = compiler.ast.CallFunc( 670 compiler.ast.Name("$op%s" % op), 671 [n.expr] + list(n.subs) + (expr and [expr] or []) 672 ) 673 674 # Fix parse tree structure. 675 676 if op == "delitem": 677 invocation = compiler.ast.Discard(invocation) 678 679 return self.process_structure_node(invocation) 680 681 def process_attribute_chain(self, n): 682 683 """ 684 Process the given attribute access node 'n'. Return a reference 685 describing the expression. 686 """ 687 688 # AssAttr/Getattr are nested with the outermost access being the last 689 # access in any chain. 690 691 self.attrs.insert(0, n.attrname) 692 attrs = self.attrs 693 694 # Break attribute chains where non-access nodes are found. 695 696 if not self.have_access_expression(n): 697 self.reset_attribute_chain() 698 699 # Descend into the expression, extending backwards any existing chain, 700 # or building another for the expression. 701 702 name_ref = self.process_structure_node(n.expr) 703 704 # Restore chain information applying to this node. 705 706 if not self.have_access_expression(n): 707 self.restore_attribute_chain(attrs) 708 709 # Return immediately if the expression was another access and thus a 710 # continuation backwards along the chain. The above processing will 711 # have followed the chain all the way to its conclusion. 712 713 if self.have_access_expression(n): 714 del self.attrs[0] 715 716 return name_ref 717 718 # Attribute chain handling. 719 720 def reset_attribute_chain(self): 721 722 "Reset the attribute chain for a subexpression of an attribute access." 723 724 self.attrs = [] 725 self.chain_assignment.append(self.in_assignment) 726 self.chain_invocation.append(self.in_invocation) 727 self.in_assignment = None 728 self.in_invocation = None 729 730 def restore_attribute_chain(self, attrs): 731 732 "Restore the attribute chain for an attribute access." 733 734 self.attrs = attrs 735 self.in_assignment = self.chain_assignment.pop() 736 self.in_invocation = self.chain_invocation.pop() 737 738 def have_access_expression(self, node): 739 740 "Return whether the expression associated with 'node' is Getattr." 741 742 return isinstance(node.expr, compiler.ast.Getattr) 743 744 def get_name_for_tracking(self, name, ref=None): 745 746 """ 747 Return the name to be used for attribute usage observations involving 748 the given 'name' in the current namespace. If 'ref' is indicated and 749 the name is being used outside a function, return the origin information 750 from 'ref'; otherwise, return a path computed using the current 751 namespace and the given name. 752 753 The intention of this method is to provide a suitably-qualified name 754 that can be tracked across namespaces. Where globals are being 755 referenced in class namespaces, they should be referenced using their 756 path within the module, not using a path within each class. 757 758 It may not be possible to identify a global within a function at the 759 time of inspection (since a global may appear later in a file). 760 Consequently, globals are identified by their local name rather than 761 their module-qualified path. 762 """ 763 764 # For functions, use the appropriate local names. 765 766 if self.in_function: 767 return name 768 769 # For static namespaces, use the given qualified name. 770 771 elif ref and ref.static(): 772 return ref.get_origin() 773 774 # For non-static objects in static namespaces, use any alias. 775 776 elif ref and ref.get_name(): 777 return ref.get_name() 778 779 # Otherwise, establish a name in the current namespace. 780 781 else: 782 return self.get_object_path(name) 783 784 def get_path_for_access(self): 785 786 "Outside functions, register accesses at the module level." 787 788 if not self.in_function: 789 return self.name 790 else: 791 return self.get_namespace_path() 792 793 def get_module_name(self, node): 794 795 """ 796 Using the given From 'node' in this module, calculate any relative import 797 information, returning a tuple containing a module to import along with any 798 names to import based on the node's name information. 799 800 Where the returned module is given as None, whole module imports should 801 be performed for the returned modules using the returned names. 802 """ 803 804 # Absolute import. 805 806 if node.level == 0: 807 return node.modname, node.names 808 809 # Relative to an ancestor of this module. 810 811 else: 812 path = self.name.split(".") 813 level = node.level 814 815 # Relative imports treat package roots as submodules. 816 817 if split(self.filename)[-1] == "__init__.py": 818 level -= 1 819 820 if level > len(path): 821 raise InspectError("Relative import %r involves too many levels up from module %r" % ( 822 ("%s%s" % ("." * node.level, node.modname or "")), self.name)) 823 824 basename = ".".join(path[:len(path)-level]) 825 826 # Name imports from a module. 827 828 if node.modname: 829 return "%s.%s" % (basename, node.modname), node.names 830 831 # Relative whole module imports. 832 833 else: 834 return basename, node.names 835 836 def get_argnames(args): 837 838 """ 839 Return a list of all names provided by 'args'. Since tuples may be 840 employed, the arguments are traversed depth-first. 841 """ 842 843 l = [] 844 for arg in args: 845 if isinstance(arg, tuple): 846 l += get_argnames(arg) 847 else: 848 l.append(arg) 849 return l 850 851 def get_names_from_nodes(nodes): 852 853 """ 854 Return the names employed in the given 'nodes' along with the number of 855 nodes excluding sequences. 856 """ 857 858 names = set() 859 count = 0 860 861 for node in nodes: 862 863 # Add names and count them. 864 865 if isinstance(node, (compiler.ast.AssName, compiler.ast.Name)): 866 names.add(node.name) 867 count += 1 868 869 # Add names from sequences and incorporate their counts. 870 871 elif isinstance(node, (compiler.ast.AssList, compiler.ast.AssTuple, 872 compiler.ast.List, compiler.ast.Set, 873 compiler.ast.Tuple)): 874 _names, _count = get_names_from_nodes(node.nodes) 875 names.update(_names) 876 count += _count 877 878 # Count non-name, non-sequence nodes. 879 880 else: 881 count += 1 882 883 return names, count 884 885 # Result classes. 886 887 class InstructionSequence: 888 889 "A generic sequence of instructions." 890 891 def __init__(self, instructions): 892 self.instructions = instructions 893 894 def get_value_instruction(self): 895 return self.instructions[-1] 896 897 def get_init_instructions(self): 898 return self.instructions[:-1] 899 900 # Dictionary utilities. 901 902 def init_item(d, key, fn): 903 904 """ 905 Add to 'd' an entry for 'key' using the callable 'fn' to make an initial 906 value where no entry already exists. 907 """ 908 909 if not d.has_key(key): 910 d[key] = fn() 911 return d[key] 912 913 def dict_for_keys(d, keys): 914 915 "Return a new dictionary containing entries from 'd' for the given 'keys'." 916 917 nd = {} 918 for key in keys: 919 if d.has_key(key): 920 nd[key] = d[key] 921 return nd 922 923 def make_key(s): 924 925 "Make sequence 's' into a tuple-based key, first sorting its contents." 926 927 l = list(s) 928 l.sort() 929 return tuple(l) 930 931 def add_counter_item(d, key): 932 933 """ 934 Make a mapping in 'd' for 'key' to the number of keys added before it, thus 935 maintaining a mapping of keys to their order of insertion. 936 """ 937 938 if not d.has_key(key): 939 d[key] = len(d.keys()) 940 return d[key] 941 942 def remove_items(d1, d2): 943 944 "Remove from 'd1' all items from 'd2'." 945 946 for key in d2.keys(): 947 if d1.has_key(key): 948 del d1[key] 949 950 # Set utilities. 951 952 def first(s): 953 return list(s)[0] 954 955 def same(s1, s2): 956 return set(s1) == set(s2) 957 958 # General input/output. 959 960 def readfile(filename): 961 962 "Return the contents of 'filename'." 963 964 f = open(filename) 965 try: 966 return f.read() 967 finally: 968 f.close() 969 970 def writefile(filename, s): 971 972 "Write to 'filename' the string 's'." 973 974 f = open(filename, "w") 975 try: 976 f.write(s) 977 finally: 978 f.close() 979 980 # General encoding. 981 982 def sorted_output(x): 983 984 "Sort sequence 'x' and return a string with commas separating the values." 985 986 x = map(str, x) 987 x.sort() 988 return ", ".join(x) 989 990 def get_string_details(literals, encoding): 991 992 """ 993 Determine whether 'literals' represent Unicode strings or byte strings, 994 using 'encoding' to reproduce byte sequences. 995 996 Each literal is the full program representation including prefix and quotes 997 recoded by the parser to UTF-8. Thus, any literal found to represent a byte 998 string needs to be translated back to its original encoding. 999 1000 Return a single encoded literal value, a type name, and the original 1001 encoding as a tuple. 1002 """ 1003 1004 typename = "unicode" 1005 1006 l = [] 1007 1008 for s in literals: 1009 out, _typename = get_literal_details(s) 1010 if _typename == "str": 1011 typename = "str" 1012 l.append(out) 1013 1014 out = "".join(l) 1015 1016 # For Unicode values, convert to the UTF-8 program representation. 1017 1018 if typename == "unicode": 1019 return out.encode("utf-8"), typename, encoding 1020 1021 # For byte string values, convert back to the original encoding. 1022 1023 else: 1024 return out.encode(encoding), typename, encoding 1025 1026 def get_literal_details(s): 1027 1028 """ 1029 Determine whether 's' represents a Unicode string or a byte string, where 1030 's' contains the full program representation of a literal including prefix 1031 and quotes, recoded by the parser to UTF-8. 1032 1033 Find and convert Unicode values starting with <backslash>u or <backslash>U, 1034 and byte or Unicode values starting with <backslash><octal digit> or 1035 <backslash>x. 1036 1037 Literals prefixed with "u" cause <backslash><octal digit> and <backslash>x 1038 to be considered as Unicode values. Otherwise, they produce byte values and 1039 cause unprefixed strings to be considered as byte strings. 1040 1041 Literals prefixed with "r" do not have their backslash-encoded values 1042 converted unless also prefixed with "u", in which case only the above value 1043 formats are converted, not any of the other special sequences for things 1044 like newlines. 1045 1046 Return the literal value as a Unicode object together with the appropriate 1047 type name in a tuple. 1048 """ 1049 1050 l = [] 1051 1052 # Identify the quote character and use it to identify the prefix. 1053 1054 quote_type = s[-1] 1055 prefix_end = s.find(quote_type) 1056 prefix = s[:prefix_end].lower() 1057 1058 if prefix not in ("", "b", "br", "r", "u", "ur"): 1059 raise ValueError, "String literal does not have a supported prefix: %s" % s 1060 1061 if "b" in prefix: 1062 typename = "str" 1063 else: 1064 typename = "unicode" 1065 1066 # Identify triple quotes or single quotes. 1067 1068 if len(s) >= 6 and s[-2] == quote_type and s[-3] == quote_type: 1069 quote = s[prefix_end:prefix_end+3] 1070 current = prefix_end + 3 1071 end = len(s) - 3 1072 else: 1073 quote = s[prefix_end] 1074 current = prefix_end + 1 1075 end = len(s) - 1 1076 1077 # Conversions of some quoted values. 1078 1079 searches = { 1080 "u" : (6, 16), 1081 "U" : (10, 16), 1082 "x" : (4, 16), 1083 } 1084 1085 octal_digits = map(str, range(0, 8)) 1086 1087 # Translations of some quoted values. 1088 1089 escaped = { 1090 "\\" : "\\", "'" : "'", '"' : '"', 1091 "a" : "\a", "b" : "\b", "f" : "\f", 1092 "n" : "\n", "r" : "\r", "t" : "\t", 1093 } 1094 1095 while current < end: 1096 1097 # Look for quoted values. 1098 1099 index = s.find("\\", current) 1100 if index == -1 or index + 1 == end: 1101 l.append(s[current:end]) 1102 break 1103 1104 # Add the preceding text. 1105 1106 l.append(s[current:index]) 1107 1108 # Handle quoted text. 1109 1110 term = s[index+1] 1111 1112 # Add Unicode values. Where a string is u-prefixed, even \o and \x 1113 # produce Unicode values. 1114 1115 if typename == "unicode" and ( 1116 term in ("u", "U") or 1117 "u" in prefix and (term == "x" or term in octal_digits)): 1118 1119 needed, base = searches.get(term, (4, 8)) 1120 value = convert_quoted_value(s, index, needed, end, base, unichr) 1121 l.append(value) 1122 current = index + needed 1123 1124 # Add raw byte values, changing the string type. 1125 1126 elif "r" not in prefix and ( 1127 term == "x" or term in octal_digits): 1128 1129 needed, base = searches.get(term, (4, 8)) 1130 value = convert_quoted_value(s, index, needed, end, base, chr) 1131 l.append(value) 1132 typename = "str" 1133 current = index + needed 1134 1135 # Add other escaped values. 1136 1137 elif "r" not in prefix and escaped.has_key(term): 1138 l.append(escaped[term]) 1139 current = index + 2 1140 1141 # Add other text as found. 1142 1143 else: 1144 l.append(s[index:index+2]) 1145 current = index + 2 1146 1147 # Collect the components into a single Unicode object. Since the literal 1148 # text was already in UTF-8 form, interpret plain strings as UTF-8 1149 # sequences. 1150 1151 out = [] 1152 1153 for value in l: 1154 if isinstance(value, unicode): 1155 out.append(value) 1156 else: 1157 out.append(unicode(value, "utf-8")) 1158 1159 return "".join(out), typename 1160 1161 def convert_quoted_value(s, index, needed, end, base, fn): 1162 1163 """ 1164 Interpret a quoted value in 's' at 'index' with the given 'needed' number of 1165 positions, and with the given 'end' indicating the first position after the 1166 end of the actual string content. 1167 1168 Use 'base' as the numerical base when interpreting the value, and use 'fn' 1169 to convert the value to an appropriate type. 1170 """ 1171 1172 s = s[index:min(index+needed, end)] 1173 1174 # Not a complete occurrence. 1175 1176 if len(s) < needed: 1177 return s 1178 1179 # Test for a well-formed value. 1180 1181 try: 1182 first = base == 8 and 1 or 2 1183 value = int(s[first:needed], base) 1184 except ValueError: 1185 return s 1186 else: 1187 return fn(value) 1188 1189 # Attribute chain decoding. 1190 1191 def get_attrnames(attrnames): 1192 1193 """ 1194 Split the qualified attribute chain 'attrnames' into its components, 1195 handling special attributes starting with "#" that indicate type 1196 conformance. 1197 """ 1198 1199 if attrnames.startswith("#"): 1200 return [attrnames] 1201 else: 1202 return attrnames.split(".") 1203 1204 def get_attrname_from_location(location): 1205 1206 """ 1207 Extract the first attribute from the attribute names employed in a 1208 'location'. 1209 """ 1210 1211 path, name, attrnames, access = location 1212 if not attrnames: 1213 return attrnames 1214 return get_attrnames(attrnames)[0] 1215 1216 def get_name_path(path, name): 1217 1218 "Return a suitable qualified name from the given 'path' and 'name'." 1219 1220 if "." in name: 1221 return name 1222 else: 1223 return "%s.%s" % (path, name) 1224 1225 # Usage-related functions. 1226 1227 def get_types_for_usage(attrnames, objects): 1228 1229 """ 1230 Identify the types that can support the given 'attrnames', using the 1231 given 'objects' as the catalogue of type details. 1232 """ 1233 1234 types = [] 1235 for name, _attrnames in objects.items(): 1236 if set(attrnames).issubset(_attrnames): 1237 types.append(name) 1238 return types 1239 1240 def get_invoked_attributes(usage): 1241 1242 "Obtain invoked attribute from the given 'usage'." 1243 1244 invoked = [] 1245 if usage: 1246 for attrname, invocation, assignment in usage: 1247 if invocation: 1248 invoked.append(attrname) 1249 return invoked 1250 1251 def get_assigned_attributes(usage): 1252 1253 "Obtain assigned attribute from the given 'usage'." 1254 1255 assigned = [] 1256 if usage: 1257 for attrname, invocation, assignment in usage: 1258 if assignment: 1259 assigned.append(attrname) 1260 return assigned 1261 1262 # Type and module functions. 1263 # NOTE: This makes assumptions about the __builtins__ structure. 1264 1265 def get_builtin_module(name): 1266 1267 "Return the module name containing the given type 'name'." 1268 1269 if name == "string": 1270 modname = "str" 1271 elif name == "utf8string": 1272 modname = "unicode" 1273 elif name == "NoneType": 1274 modname = "none" 1275 else: 1276 modname = name 1277 1278 return "__builtins__.%s" % modname 1279 1280 def get_builtin_type(name): 1281 1282 "Return the type name provided by the given Python value 'name'." 1283 1284 if name == "str": 1285 return "string" 1286 elif name == "unicode": 1287 return "utf8string" 1288 else: 1289 return name 1290 1291 def get_builtin_class(name): 1292 1293 "Return the full name of the built-in class having the given 'name'." 1294 1295 typename = get_builtin_type(name) 1296 module = get_builtin_module(typename) 1297 return "%s.%s" % (module, typename) 1298 1299 # Useful data. 1300 1301 predefined_constants = "False", "None", "NotImplemented", "True" 1302 1303 operator_functions = { 1304 1305 # Fundamental operations. 1306 1307 "is" : "is_", 1308 "is not" : "is_not", 1309 1310 # Binary operations. 1311 1312 "in" : "in_", 1313 "not in" : "not_in", 1314 "Add" : "add", 1315 "Bitand" : "and_", 1316 "Bitor" : "or_", 1317 "Bitxor" : "xor", 1318 "Div" : "div", 1319 "FloorDiv" : "floordiv", 1320 "LeftShift" : "lshift", 1321 "Mod" : "mod", 1322 "Mul" : "mul", 1323 "Power" : "pow", 1324 "RightShift" : "rshift", 1325 "Sub" : "sub", 1326 1327 # Unary operations. 1328 1329 "Invert" : "invert", 1330 "UnaryAdd" : "pos", 1331 "UnarySub" : "neg", 1332 1333 # Augmented assignment. 1334 1335 "+=" : "iadd", 1336 "-=" : "isub", 1337 "*=" : "imul", 1338 "/=" : "idiv", 1339 "//=" : "ifloordiv", 1340 "%=" : "imod", 1341 "**=" : "ipow", 1342 "<<=" : "ilshift", 1343 ">>=" : "irshift", 1344 "&=" : "iand", 1345 "^=" : "ixor", 1346 "|=" : "ior", 1347 1348 # Comparisons. 1349 1350 "==" : "eq", 1351 "!=" : "ne", 1352 "<" : "lt", 1353 "<=" : "le", 1354 ">=" : "ge", 1355 ">" : "gt", 1356 } 1357 1358 # vim: tabstop=4 expandtab shiftwidth=4