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 = False 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 op = n.flags == "OP_ASSIGN" and "setslice" or "getslice" 630 invocation = compiler.ast.CallFunc( 631 compiler.ast.Name("$op%s" % op), 632 [n.expr, n.lower or compiler.ast.Name("None"), n.upper or compiler.ast.Name("None")] + 633 (expr and [expr] or []) 634 ) 635 return self.process_structure_node(invocation) 636 637 def process_sliceobj_node(self, n): 638 639 """ 640 Process the given slice object node 'n' as a slice constructor. 641 """ 642 643 op = "slice" 644 invocation = compiler.ast.CallFunc( 645 compiler.ast.Name("$op%s" % op), 646 n.nodes 647 ) 648 return self.process_structure_node(invocation) 649 650 def process_subscript_node(self, n, expr=None): 651 652 """ 653 Process the given subscript node 'n' as an operator function invocation. 654 """ 655 656 op = n.flags == "OP_ASSIGN" and "setitem" or "getitem" 657 invocation = compiler.ast.CallFunc( 658 compiler.ast.Name("$op%s" % op), 659 [n.expr] + list(n.subs) + (expr and [expr] or []) 660 ) 661 return self.process_structure_node(invocation) 662 663 def process_attribute_chain(self, n): 664 665 """ 666 Process the given attribute access node 'n'. Return a reference 667 describing the expression. 668 """ 669 670 # AssAttr/Getattr are nested with the outermost access being the last 671 # access in any chain. 672 673 self.attrs.insert(0, n.attrname) 674 attrs = self.attrs 675 676 # Break attribute chains where non-access nodes are found. 677 678 if not self.have_access_expression(n): 679 self.reset_attribute_chain() 680 681 # Descend into the expression, extending backwards any existing chain, 682 # or building another for the expression. 683 684 name_ref = self.process_structure_node(n.expr) 685 686 # Restore chain information applying to this node. 687 688 if not self.have_access_expression(n): 689 self.restore_attribute_chain(attrs) 690 691 # Return immediately if the expression was another access and thus a 692 # continuation backwards along the chain. The above processing will 693 # have followed the chain all the way to its conclusion. 694 695 if self.have_access_expression(n): 696 del self.attrs[0] 697 698 return name_ref 699 700 # Attribute chain handling. 701 702 def reset_attribute_chain(self): 703 704 "Reset the attribute chain for a subexpression of an attribute access." 705 706 self.attrs = [] 707 self.chain_assignment.append(self.in_assignment) 708 self.chain_invocation.append(self.in_invocation) 709 self.in_assignment = None 710 self.in_invocation = False 711 712 def restore_attribute_chain(self, attrs): 713 714 "Restore the attribute chain for an attribute access." 715 716 self.attrs = attrs 717 self.in_assignment = self.chain_assignment.pop() 718 self.in_invocation = self.chain_invocation.pop() 719 720 def have_access_expression(self, node): 721 722 "Return whether the expression associated with 'node' is Getattr." 723 724 return isinstance(node.expr, compiler.ast.Getattr) 725 726 def get_name_for_tracking(self, name, path=None): 727 728 """ 729 Return the name to be used for attribute usage observations involving 730 the given 'name' in the current namespace. If 'path' is indicated and 731 the name is being used outside a function, return the path value; 732 otherwise, return a path computed using the current namespace and the 733 given name. 734 735 The intention of this method is to provide a suitably-qualified name 736 that can be tracked across namespaces. Where globals are being 737 referenced in class namespaces, they should be referenced using their 738 path within the module, not using a path within each class. 739 740 It may not be possible to identify a global within a function at the 741 time of inspection (since a global may appear later in a file). 742 Consequently, globals are identified by their local name rather than 743 their module-qualified path. 744 """ 745 746 # For functions, use the appropriate local names. 747 748 if self.in_function: 749 return name 750 751 # For static namespaces, use the given qualified name. 752 753 elif path: 754 return path 755 756 # Otherwise, establish a name in the current namespace. 757 758 else: 759 return self.get_object_path(name) 760 761 def get_path_for_access(self): 762 763 "Outside functions, register accesses at the module level." 764 765 if not self.in_function: 766 return self.name 767 else: 768 return self.get_namespace_path() 769 770 def get_module_name(self, node): 771 772 """ 773 Using the given From 'node' in this module, calculate any relative import 774 information, returning a tuple containing a module to import along with any 775 names to import based on the node's name information. 776 777 Where the returned module is given as None, whole module imports should 778 be performed for the returned modules using the returned names. 779 """ 780 781 # Absolute import. 782 783 if node.level == 0: 784 return node.modname, node.names 785 786 # Relative to an ancestor of this module. 787 788 else: 789 path = self.name.split(".") 790 level = node.level 791 792 # Relative imports treat package roots as submodules. 793 794 if split(self.filename)[-1] == "__init__.py": 795 level -= 1 796 797 if level > len(path): 798 raise InspectError("Relative import %r involves too many levels up from module %r" % ( 799 ("%s%s" % ("." * node.level, node.modname or "")), self.name)) 800 801 basename = ".".join(path[:len(path)-level]) 802 803 # Name imports from a module. 804 805 if node.modname: 806 return "%s.%s" % (basename, node.modname), node.names 807 808 # Relative whole module imports. 809 810 else: 811 return basename, node.names 812 813 def get_argnames(args): 814 815 """ 816 Return a list of all names provided by 'args'. Since tuples may be 817 employed, the arguments are traversed depth-first. 818 """ 819 820 l = [] 821 for arg in args: 822 if isinstance(arg, tuple): 823 l += get_argnames(arg) 824 else: 825 l.append(arg) 826 return l 827 828 def get_names_from_nodes(nodes): 829 830 """ 831 Return the names employed in the given 'nodes' along with the number of 832 nodes excluding sequences. 833 """ 834 835 names = set() 836 count = 0 837 838 for node in nodes: 839 840 # Add names and count them. 841 842 if isinstance(node, (compiler.ast.AssName, compiler.ast.Name)): 843 names.add(node.name) 844 count += 1 845 846 # Add names from sequences and incorporate their counts. 847 848 elif isinstance(node, (compiler.ast.AssList, compiler.ast.AssTuple, 849 compiler.ast.List, compiler.ast.Set, 850 compiler.ast.Tuple)): 851 _names, _count = get_names_from_nodes(node.nodes) 852 names.update(_names) 853 count += _count 854 855 # Count non-name, non-sequence nodes. 856 857 else: 858 count += 1 859 860 return names, count 861 862 # Result classes. 863 864 class InstructionSequence: 865 866 "A generic sequence of instructions." 867 868 def __init__(self, instructions): 869 self.instructions = instructions 870 871 def get_value_instruction(self): 872 return self.instructions[-1] 873 874 def get_init_instructions(self): 875 return self.instructions[:-1] 876 877 # Dictionary utilities. 878 879 def init_item(d, key, fn): 880 881 """ 882 Add to 'd' an entry for 'key' using the callable 'fn' to make an initial 883 value where no entry already exists. 884 """ 885 886 if not d.has_key(key): 887 d[key] = fn() 888 return d[key] 889 890 def dict_for_keys(d, keys): 891 892 "Return a new dictionary containing entries from 'd' for the given 'keys'." 893 894 nd = {} 895 for key in keys: 896 if d.has_key(key): 897 nd[key] = d[key] 898 return nd 899 900 def make_key(s): 901 902 "Make sequence 's' into a tuple-based key, first sorting its contents." 903 904 l = list(s) 905 l.sort() 906 return tuple(l) 907 908 def add_counter_item(d, key): 909 910 """ 911 Make a mapping in 'd' for 'key' to the number of keys added before it, thus 912 maintaining a mapping of keys to their order of insertion. 913 """ 914 915 if not d.has_key(key): 916 d[key] = len(d.keys()) 917 return d[key] 918 919 def remove_items(d1, d2): 920 921 "Remove from 'd1' all items from 'd2'." 922 923 for key in d2.keys(): 924 if d1.has_key(key): 925 del d1[key] 926 927 # Set utilities. 928 929 def first(s): 930 return list(s)[0] 931 932 def same(s1, s2): 933 return set(s1) == set(s2) 934 935 # General input/output. 936 937 def readfile(filename): 938 939 "Return the contents of 'filename'." 940 941 f = open(filename) 942 try: 943 return f.read() 944 finally: 945 f.close() 946 947 def writefile(filename, s): 948 949 "Write to 'filename' the string 's'." 950 951 f = open(filename, "w") 952 try: 953 f.write(s) 954 finally: 955 f.close() 956 957 # General encoding. 958 959 def sorted_output(x): 960 961 "Sort sequence 'x' and return a string with commas separating the values." 962 963 x = map(str, x) 964 x.sort() 965 return ", ".join(x) 966 967 def get_string_details(literals, encoding): 968 969 """ 970 Determine whether 'literals' represent Unicode strings or byte strings, 971 using 'encoding' to reproduce byte sequences. 972 973 Each literal is the full program representation including prefix and quotes 974 recoded by the parser to UTF-8. Thus, any literal found to represent a byte 975 string needs to be translated back to its original encoding. 976 977 Return a single encoded literal value, a type name, and the original 978 encoding as a tuple. 979 """ 980 981 typename = "unicode" 982 983 l = [] 984 985 for s in literals: 986 out, _typename = get_literal_details(s) 987 if _typename == "str": 988 typename = "str" 989 l.append(out) 990 991 out = "".join(l) 992 993 # For Unicode values, convert to the UTF-8 program representation. 994 995 if typename == "unicode": 996 return out.encode("utf-8"), typename, encoding 997 998 # For byte string values, convert back to the original encoding. 999 1000 else: 1001 return out.encode(encoding), typename, encoding 1002 1003 def get_literal_details(s): 1004 1005 """ 1006 Determine whether 's' represents a Unicode string or a byte string, where 1007 's' contains the full program representation of a literal including prefix 1008 and quotes, recoded by the parser to UTF-8. 1009 1010 Find and convert Unicode values starting with <backslash>u or <backslash>U, 1011 and byte or Unicode values starting with <backslash><octal digit> or 1012 <backslash>x. 1013 1014 Literals prefixed with "u" cause <backslash><octal digit> and <backslash>x 1015 to be considered as Unicode values. Otherwise, they produce byte values and 1016 cause unprefixed strings to be considered as byte strings. 1017 1018 Literals prefixed with "r" do not have their backslash-encoded values 1019 converted unless also prefixed with "u", in which case only the above value 1020 formats are converted, not any of the other special sequences for things 1021 like newlines. 1022 1023 Return the literal value as a Unicode object together with the appropriate 1024 type name in a tuple. 1025 """ 1026 1027 l = [] 1028 1029 # Identify the quote character and use it to identify the prefix. 1030 1031 quote_type = s[-1] 1032 prefix_end = s.find(quote_type) 1033 prefix = s[:prefix_end].lower() 1034 1035 if prefix not in ("", "b", "br", "r", "u", "ur"): 1036 raise ValueError, "String literal does not have a supported prefix: %s" % s 1037 1038 if "b" in prefix: 1039 typename = "str" 1040 else: 1041 typename = "unicode" 1042 1043 # Identify triple quotes or single quotes. 1044 1045 if len(s) >= 6 and s[-2] == quote_type and s[-3] == quote_type: 1046 quote = s[prefix_end:prefix_end+3] 1047 current = prefix_end + 3 1048 end = len(s) - 3 1049 else: 1050 quote = s[prefix_end] 1051 current = prefix_end + 1 1052 end = len(s) - 1 1053 1054 # Conversions of some quoted values. 1055 1056 searches = { 1057 "u" : (6, 16), 1058 "U" : (10, 16), 1059 "x" : (4, 16), 1060 } 1061 1062 octal_digits = map(str, range(0, 8)) 1063 1064 # Translations of some quoted values. 1065 1066 escaped = { 1067 "\\" : "\\", "'" : "'", '"' : '"', 1068 "a" : "\a", "b" : "\b", "f" : "\f", 1069 "n" : "\n", "r" : "\r", "t" : "\t", 1070 } 1071 1072 while current < end: 1073 1074 # Look for quoted values. 1075 1076 index = s.find("\\", current) 1077 if index == -1 or index + 1 == end: 1078 l.append(s[current:end]) 1079 break 1080 1081 # Add the preceding text. 1082 1083 l.append(s[current:index]) 1084 1085 # Handle quoted text. 1086 1087 term = s[index+1] 1088 1089 # Add Unicode values. Where a string is u-prefixed, even \o and \x 1090 # produce Unicode values. 1091 1092 if typename == "unicode" and ( 1093 term in ("u", "U") or 1094 "u" in prefix and (term == "x" or term in octal_digits)): 1095 1096 needed, base = searches.get(term, (4, 8)) 1097 value = convert_quoted_value(s, index, needed, end, base, unichr) 1098 l.append(value) 1099 current = index + needed 1100 1101 # Add raw byte values, changing the string type. 1102 1103 elif "r" not in prefix and ( 1104 term == "x" or term in octal_digits): 1105 1106 needed, base = searches.get(term, (4, 8)) 1107 value = convert_quoted_value(s, index, needed, end, base, chr) 1108 l.append(value) 1109 typename = "str" 1110 current = index + needed 1111 1112 # Add other escaped values. 1113 1114 elif "r" not in prefix and escaped.has_key(term): 1115 l.append(escaped[term]) 1116 current = index + 2 1117 1118 # Add other text as found. 1119 1120 else: 1121 l.append(s[index:index+2]) 1122 current = index + 2 1123 1124 # Collect the components into a single Unicode object. Since the literal 1125 # text was already in UTF-8 form, interpret plain strings as UTF-8 1126 # sequences. 1127 1128 out = [] 1129 1130 for value in l: 1131 if isinstance(value, unicode): 1132 out.append(value) 1133 else: 1134 out.append(unicode(value, "utf-8")) 1135 1136 return "".join(out), typename 1137 1138 def convert_quoted_value(s, index, needed, end, base, fn): 1139 1140 """ 1141 Interpret a quoted value in 's' at 'index' with the given 'needed' number of 1142 positions, and with the given 'end' indicating the first position after the 1143 end of the actual string content. 1144 1145 Use 'base' as the numerical base when interpreting the value, and use 'fn' 1146 to convert the value to an appropriate type. 1147 """ 1148 1149 s = s[index:min(index+needed, end)] 1150 1151 # Not a complete occurrence. 1152 1153 if len(s) < needed: 1154 return s 1155 1156 # Test for a well-formed value. 1157 1158 try: 1159 first = base == 8 and 1 or 2 1160 value = int(s[first:needed], base) 1161 except ValueError: 1162 return s 1163 else: 1164 return fn(value) 1165 1166 # Attribute chain decoding. 1167 1168 def get_attrnames(attrnames): 1169 1170 """ 1171 Split the qualified attribute chain 'attrnames' into its components, 1172 handling special attributes starting with "#" that indicate type 1173 conformance. 1174 """ 1175 1176 if attrnames.startswith("#"): 1177 return [attrnames] 1178 else: 1179 return attrnames.split(".") 1180 1181 def get_attrname_from_location(location): 1182 1183 """ 1184 Extract the first attribute from the attribute names employed in a 1185 'location'. 1186 """ 1187 1188 path, name, attrnames, access = location 1189 if not attrnames: 1190 return attrnames 1191 return get_attrnames(attrnames)[0] 1192 1193 def get_name_path(path, name): 1194 1195 "Return a suitable qualified name from the given 'path' and 'name'." 1196 1197 if "." in name: 1198 return name 1199 else: 1200 return "%s.%s" % (path, name) 1201 1202 # Usage-related functions. 1203 1204 def get_types_for_usage(attrnames, objects): 1205 1206 """ 1207 Identify the types that can support the given 'attrnames', using the 1208 given 'objects' as the catalogue of type details. 1209 """ 1210 1211 types = [] 1212 for name, _attrnames in objects.items(): 1213 if set(attrnames).issubset(_attrnames): 1214 types.append(name) 1215 return types 1216 1217 def get_invoked_attributes(usage): 1218 1219 "Obtain invoked attribute from the given 'usage'." 1220 1221 invoked = [] 1222 if usage: 1223 for attrname, invocation, assignment in usage: 1224 if invocation: 1225 invoked.append(attrname) 1226 return invoked 1227 1228 def get_assigned_attributes(usage): 1229 1230 "Obtain assigned attribute from the given 'usage'." 1231 1232 assigned = [] 1233 if usage: 1234 for attrname, invocation, assignment in usage: 1235 if assignment: 1236 assigned.append(attrname) 1237 return assigned 1238 1239 # Type and module functions. 1240 # NOTE: This makes assumptions about the __builtins__ structure. 1241 1242 def get_builtin_module(name): 1243 1244 "Return the module name containing the given type 'name'." 1245 1246 if name == "string": 1247 modname = "str" 1248 elif name == "utf8string": 1249 modname = "unicode" 1250 elif name == "NoneType": 1251 modname = "none" 1252 else: 1253 modname = name 1254 1255 return "__builtins__.%s" % modname 1256 1257 def get_builtin_type(name): 1258 1259 "Return the type name provided by the given Python value 'name'." 1260 1261 if name == "str": 1262 return "string" 1263 elif name == "unicode": 1264 return "utf8string" 1265 else: 1266 return name 1267 1268 def get_builtin_class(name): 1269 1270 "Return the full name of the built-in class having the given 'name'." 1271 1272 typename = get_builtin_type(name) 1273 module = get_builtin_module(typename) 1274 return "%s.%s" % (module, typename) 1275 1276 # Useful data. 1277 1278 predefined_constants = "False", "None", "NotImplemented", "True" 1279 1280 operator_functions = { 1281 1282 # Fundamental operations. 1283 1284 "is" : "is_", 1285 "is not" : "is_not", 1286 1287 # Binary operations. 1288 1289 "in" : "in_", 1290 "not in" : "not_in", 1291 "Add" : "add", 1292 "Bitand" : "and_", 1293 "Bitor" : "or_", 1294 "Bitxor" : "xor", 1295 "Div" : "div", 1296 "FloorDiv" : "floordiv", 1297 "LeftShift" : "lshift", 1298 "Mod" : "mod", 1299 "Mul" : "mul", 1300 "Power" : "pow", 1301 "RightShift" : "rshift", 1302 "Sub" : "sub", 1303 1304 # Unary operations. 1305 1306 "Invert" : "invert", 1307 "UnaryAdd" : "pos", 1308 "UnarySub" : "neg", 1309 1310 # Augmented assignment. 1311 1312 "+=" : "iadd", 1313 "-=" : "isub", 1314 "*=" : "imul", 1315 "/=" : "idiv", 1316 "//=" : "ifloordiv", 1317 "%=" : "imod", 1318 "**=" : "ipow", 1319 "<<=" : "ilshift", 1320 ">>=" : "irshift", 1321 "&=" : "iand", 1322 "^=" : "ixor", 1323 "|=" : "ior", 1324 1325 # Comparisons. 1326 1327 "==" : "eq", 1328 "!=" : "ne", 1329 "<" : "lt", 1330 "<=" : "le", 1331 ">=" : "ge", 1332 ">" : "gt", 1333 } 1334 1335 # vim: tabstop=4 expandtab shiftwidth=4