1 #!/usr/bin/env python 2 3 """ 4 Inspect and obtain module structure. 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 branching import BranchTracker 24 from common import CommonModule, get_argnames, init_item, predefined_constants 25 from modules import BasicModule, CacheWritingModule, InspectionNaming 26 from errors import InspectError 27 from referencing import Reference 28 from resolving import NameResolving 29 from results import AccessRef, InstanceRef, InvocationRef, LiteralSequenceRef, \ 30 LocalNameRef, NameRef, ResolvedNameRef, VariableRef 31 import compiler 32 import sys 33 34 class InspectedModule(BasicModule, CacheWritingModule, NameResolving, InspectionNaming): 35 36 "A module inspector." 37 38 def __init__(self, name, importer): 39 40 "Initialise the module with basic details." 41 42 BasicModule.__init__(self, name, importer) 43 44 self.in_class = False 45 self.in_conditional = False 46 47 # Accesses to global attributes. 48 49 self.global_attr_accesses = {} 50 51 # Usage tracking. 52 53 self.trackers = [] 54 self.attr_accessor_branches = {} 55 56 def __repr__(self): 57 return "InspectedModule(%r, %r)" % (self.name, self.importer) 58 59 # Principal methods. 60 61 def parse(self, filename): 62 63 "Parse the file having the given 'filename'." 64 65 self.parse_file(filename) 66 67 # Inspect the module. 68 69 self.start_tracking_in_module() 70 71 # Detect and record imports and globals declared in the module. 72 73 self.process_structure(self.astnode) 74 75 # Set the class of the module after the definition has occurred. 76 77 ref = self.get_builtin("module") 78 self.set_name("__class__", ref) 79 self.set_name("__name__", self.get_constant("string", self.name).reference()) 80 self.set_name("__file__", self.get_constant("string", filename).reference()) 81 82 # Reserve a constant for the encoding. 83 84 if self.encoding: 85 self.get_constant("string", self.encoding) 86 87 # Get module-level attribute usage details. 88 89 self.stop_tracking_in_module() 90 91 # Collect external name references. 92 93 self.collect_names() 94 95 def complete(self): 96 97 "Complete the module inspection." 98 99 # Resolve names not definitively mapped to objects. 100 101 self.resolve() 102 103 # Propagate to the importer information needed in subsequent activities. 104 105 self.propagate() 106 107 # Accessory methods. 108 109 def collect_names(self): 110 111 "Collect the names used by each scope." 112 113 for path in self.names_used.keys(): 114 self.collect_names_for_path(path) 115 116 def collect_names_for_path(self, path): 117 118 """ 119 Collect the names used by the given 'path'. These are propagated to the 120 importer in advance of any dependency resolution. 121 """ 122 123 names = self.names_used.get(path) 124 if not names: 125 return 126 127 in_function = self.function_locals.has_key(path) 128 129 for name in names: 130 if in_function and name in self.function_locals[path]: 131 continue 132 133 key = "%s.%s" % (path, name) 134 135 # Find local definitions (within dynamic namespaces). 136 137 ref = self.get_resolved_object(key) 138 if ref: 139 self.set_name_reference(key, ref) 140 continue 141 142 # Find global. 143 144 ref = self.get_global(name) 145 if ref: 146 self.set_name_reference(key, ref) 147 continue 148 149 # Find presumed built-in definitions. 150 151 ref = self.get_builtin(name) 152 self.set_name_reference(key, ref) 153 154 def set_name_reference(self, path, ref): 155 156 "Map the given name 'path' to 'ref'." 157 158 self.importer.all_name_references[path] = self.name_references[path] = ref 159 160 # Module structure traversal. 161 162 def process_structure_node(self, n): 163 164 "Process the individual node 'n'." 165 166 path = self.get_namespace_path() 167 168 # Module global detection. 169 170 if isinstance(n, compiler.ast.Global): 171 self.process_global_node(n) 172 173 # Module import declarations. 174 175 elif isinstance(n, compiler.ast.From): 176 self.process_from_node(n) 177 178 elif isinstance(n, compiler.ast.Import): 179 self.process_import_node(n) 180 181 # Nodes using operator module functions. 182 183 elif isinstance(n, compiler.ast.Operator): 184 return self.process_operator_node(n) 185 186 elif isinstance(n, compiler.ast.AugAssign): 187 self.process_augassign_node(n) 188 189 elif isinstance(n, compiler.ast.Compare): 190 return self.process_compare_node(n) 191 192 elif isinstance(n, compiler.ast.Slice): 193 return self.process_slice_node(n) 194 195 elif isinstance(n, compiler.ast.Sliceobj): 196 return self.process_sliceobj_node(n) 197 198 elif isinstance(n, compiler.ast.Subscript): 199 return self.process_subscript_node(n) 200 201 # Namespaces within modules. 202 203 elif isinstance(n, compiler.ast.Class): 204 self.process_class_node(n) 205 206 elif isinstance(n, compiler.ast.Function): 207 self.process_function_node(n, n.name) 208 209 elif isinstance(n, compiler.ast.Lambda): 210 return self.process_lambda_node(n) 211 212 # Assignments. 213 214 elif isinstance(n, compiler.ast.Assign): 215 216 # Handle each assignment node. 217 218 for node in n.nodes: 219 self.process_assignment_node(node, n.expr) 220 221 # Assignments within non-Assign nodes. 222 223 elif isinstance(n, compiler.ast.AssName): 224 raise InspectError("Name assignment appearing outside assignment statement.", path, n) 225 226 elif isinstance(n, compiler.ast.AssAttr): 227 raise InspectError("Attribute assignment appearing outside assignment statement.", path, n) 228 229 # Accesses. 230 231 elif isinstance(n, compiler.ast.Getattr): 232 return self.process_attribute_access(n) 233 234 # Name recording for later testing. 235 236 elif isinstance(n, compiler.ast.Name): 237 return self.process_name_node(n) 238 239 # Conditional statement tracking. 240 241 elif isinstance(n, compiler.ast.For): 242 self.process_for_node(n) 243 244 elif isinstance(n, compiler.ast.While): 245 self.process_while_node(n) 246 247 elif isinstance(n, compiler.ast.If): 248 self.process_if_node(n) 249 250 elif isinstance(n, (compiler.ast.And, compiler.ast.Or)): 251 return self.process_logical_node(n) 252 253 # Exception control-flow tracking. 254 255 elif isinstance(n, compiler.ast.TryExcept): 256 self.process_try_node(n) 257 258 elif isinstance(n, compiler.ast.TryFinally): 259 self.process_try_finally_node(n) 260 261 # Control-flow modification statements. 262 263 elif isinstance(n, compiler.ast.Break): 264 self.trackers[-1].suspend_broken_branch() 265 266 elif isinstance(n, compiler.ast.Continue): 267 self.trackers[-1].suspend_continuing_branch() 268 269 elif isinstance(n, compiler.ast.Raise): 270 self.process_structure(n) 271 self.trackers[-1].abandon_branch() 272 273 elif isinstance(n, compiler.ast.Return): 274 self.record_return_value(self.process_structure_node(n.value)) 275 self.trackers[-1].abandon_returning_branch() 276 277 # Print statements. 278 279 elif isinstance(n, (compiler.ast.Print, compiler.ast.Printnl)): 280 self.process_print_node(n) 281 282 # Invocations. 283 284 elif isinstance(n, compiler.ast.CallFunc): 285 return self.process_invocation_node(n) 286 287 # Constant usage. 288 289 elif isinstance(n, compiler.ast.Const): 290 return self.get_literal_instance(n) 291 292 elif isinstance(n, compiler.ast.Dict): 293 return self.get_literal_instance(n, "dict") 294 295 elif isinstance(n, compiler.ast.List): 296 return self.get_literal_instance(n, "list") 297 298 elif isinstance(n, compiler.ast.Tuple): 299 return self.get_literal_instance(n, "tuple") 300 301 # All other nodes are processed depth-first. 302 303 else: 304 self.process_structure(n) 305 306 # By default, no expression details are returned. 307 308 return None 309 310 # Specific node handling. 311 312 def process_assignment_node(self, n, expr): 313 314 "Process the individual node 'n' to be assigned the contents of 'expr'." 315 316 # Names and attributes are assigned the entire expression. 317 318 if isinstance(n, compiler.ast.AssName): 319 if n.name == "self": 320 raise InspectError("Redefinition of self is not allowed.", self.get_namespace_path(), n) 321 322 name_ref = expr and self.process_structure_node(expr) 323 324 # Name assignments populate either function namespaces or the 325 # general namespace hierarchy. 326 327 self.assign_general_local(n.name, name_ref) 328 329 # Record usage of the name. 330 331 self.record_name(n.name) 332 333 elif isinstance(n, compiler.ast.AssAttr): 334 if expr: 335 expr = self.process_structure_node(expr) 336 337 in_assignment = self.in_assignment 338 self.in_assignment = True 339 self.process_attribute_access(n) 340 self.in_assignment = in_assignment 341 342 # Lists and tuples are matched against the expression and their 343 # items assigned to expression items. 344 345 elif isinstance(n, (compiler.ast.AssList, compiler.ast.AssTuple)): 346 self.process_assignment_node_items(n, expr) 347 348 # Slices and subscripts are permitted within assignment nodes. 349 350 elif isinstance(n, compiler.ast.Slice): 351 self.process_slice_node(n, expr) 352 353 elif isinstance(n, compiler.ast.Subscript): 354 self.process_subscript_node(n, expr) 355 356 def process_attribute_access(self, n): 357 358 "Process the given attribute access node 'n'." 359 360 # Obtain any completed chain and return the reference to it. 361 362 name_ref = self.process_attribute_chain(n) 363 364 if self.have_access_expression(n): 365 return name_ref 366 367 # Where the start of the chain of attributes has been reached, determine 368 # the complete access. 369 370 # Given a non-access node, this chain can be handled in its entirety, 371 # either being name-based and thus an access rooted on a name, or being 372 # based on some other node and thus an anonymous access of some kind. 373 374 path = self.get_namespace_path() 375 376 # Start with the the full attribute chain. 377 378 remaining = self.attrs 379 attrnames = ".".join(remaining) 380 381 # If the accessor cannot be identified, or where attributes 382 # remain in an attribute chain, record the anonymous accesses. 383 384 if not isinstance(name_ref, NameRef): # includes ResolvedNameRef 385 386 init_item(self.attr_accesses, path, set) 387 self.attr_accesses[path].add(attrnames) 388 389 self.record_access_details(None, attrnames, self.in_assignment, 390 self.in_invocation) 391 del self.attrs[0] 392 return 393 394 # Name-based accesses will handle the first attribute in a 395 # chain. 396 397 else: 398 attrname = remaining[0] 399 400 # Attribute assignments are used to identify instance attributes. 401 402 if isinstance(n, compiler.ast.AssAttr) and \ 403 self.in_class and self.in_function and n.expr.name == "self": 404 405 self.set_instance_attr(attrname) 406 407 # Record attribute usage using any name local to this namespace, 408 # if assigned in the namespace, or using an external name 409 # (presently just globals within classes). 410 411 name = self.get_name_for_tracking(name_ref.name, name_ref) 412 tracker = self.trackers[-1] 413 414 immediate_access = len(self.attrs) == 1 415 assignment = immediate_access and isinstance(n, compiler.ast.AssAttr) 416 417 # Record global-based chains for subsequent resolution. 418 419 if name_ref.is_global_name(): 420 self.record_global_access_details(name, attrnames) 421 422 # Make sure the name is being tracked: global names will not 423 # already be initialised in a branch and must be added 424 # explicitly. 425 426 if not tracker.have_name(name): 427 tracker.assign_names([name]) 428 if self.in_function: 429 self.scope_globals[path].add(name) 430 431 # Record attribute usage in the tracker, and record the branch 432 # information for the access. 433 434 branches = tracker.use_attribute(name, attrname, 435 self.in_invocation is not None, assignment) 436 437 if not branches: 438 raise InspectError("Name %s is accessed using %s before an assignment." % ( 439 name, attrname), path, n) 440 441 self.record_branches_for_access(branches, name, attrnames) 442 access_number = self.record_access_details(name, attrnames, 443 self.in_assignment, self.in_invocation) 444 445 del self.attrs[0] 446 return AccessRef(name, attrnames, access_number) 447 448 def process_class_node(self, n): 449 450 "Process the given class node 'n'." 451 452 path = self.get_namespace_path() 453 454 # To avoid notions of class "versions" where the same definition 455 # might be parameterised with different state and be referenced 456 # elsewhere (as base classes, for example), classes in functions or 457 # conditions are forbidden. 458 459 if self.in_function or self.in_conditional: 460 print >>sys.stderr, "In %s, class %s in function or conditional statement ignored." % ( 461 path, n.name) 462 return 463 464 # Resolve base classes. 465 466 bases = [] 467 468 for base in n.bases: 469 base_class = self.get_class(base) 470 471 if not base_class: 472 print >>sys.stderr, "In %s, class %s has unidentifiable base class: %s" % ( 473 path, n.name, base) 474 return 475 else: 476 bases.append(base_class) 477 478 # Detect conflicting definitions. Such definitions cause conflicts in 479 # the storage of namespace-related information. 480 481 class_name = self.get_object_path(n.name) 482 ref = self.get_object(class_name, defer=False) 483 484 if ref and ref.static(): 485 raise InspectError("Multiple definitions for the same name are not permitted.", class_name, n) 486 487 # Record bases for the class and retain the class name. 488 # Note that the function class does not inherit from the object class. 489 490 if not bases and class_name != "__builtins__.core.object" and \ 491 class_name != "__builtins__.core.function": 492 493 ref = self.get_object("__builtins__.object") 494 bases.append(ref) 495 496 self.importer.classes[class_name] = self.classes[class_name] = bases 497 self.importer.subclasses[class_name] = set() 498 self.scope_globals[class_name] = set() 499 500 # Set the definition before entering the namespace rather than 501 # afterwards because methods may reference it. In normal Python, 502 # a class is not accessible until the definition is complete, but 503 # methods can generally reference it since upon being called the 504 # class will already exist. 505 506 self.set_definition(n.name, "<class>") 507 508 in_class = self.in_class 509 self.in_class = class_name 510 self.set_instance_attr("__class__", Reference("<class>", class_name)) 511 self.enter_namespace(n.name) 512 513 # Do not provide the special instantiator attributes on the function 514 # class. Function instances provide these attributes. 515 516 if class_name != "__builtins__.core.function": 517 518 self.set_name("__fn__") # special instantiator attribute 519 self.set_name("__args__") # special instantiator attribute 520 521 # Provide leafname, parent and context attributes. 522 523 parent, leafname = class_name.rsplit(".", 1) 524 self.set_name("__name__", self.get_constant("string", leafname).reference()) 525 526 if class_name != "__builtins__.core.function": 527 self.set_name("__parent__") 528 529 self.process_structure_node(n.code) 530 self.exit_namespace() 531 self.in_class = in_class 532 533 def process_from_node(self, n): 534 535 "Process the given node 'n', importing from another module." 536 537 path = self.get_namespace_path() 538 539 module_name, names = self.get_module_name(n) 540 if module_name == self.name: 541 raise InspectError("Cannot import from the current module.", path, n) 542 543 self.queue_module(module_name) 544 545 # Attempt to obtain the referenced objects. 546 547 for name, alias in n.names: 548 if name == "*": 549 raise InspectError("Only explicitly specified names can be imported from modules.", path, n) 550 551 # Explicit names. 552 553 ref = self.import_name_from_module(name, module_name) 554 value = ResolvedNameRef(alias or name, ref) 555 self.set_general_local(alias or name, value) 556 557 def process_function_node(self, n, name): 558 559 """ 560 Process the given function or lambda node 'n' with the given 'name'. 561 """ 562 563 is_lambda = isinstance(n, compiler.ast.Lambda) 564 565 # Where a function is declared conditionally, use a separate name for 566 # the definition, and assign the definition to the stated name. 567 568 if (self.in_conditional or self.in_function) and not is_lambda: 569 original_name = name 570 name = self.get_lambda_name() 571 else: 572 original_name = None 573 574 # Detect conflicting definitions. Such definitions cause conflicts in 575 # the storage of namespace-related information. 576 577 function_name = self.get_object_path(name) 578 ref = self.get_object(function_name, defer=False) 579 580 if ref and ref.static(): 581 raise InspectError("Multiple definitions for the same name are not permitted.", function_name, n) 582 583 # Initialise argument and local records. 584 585 argnames = get_argnames(n.argnames) 586 is_method = self.in_class and not self.in_function 587 588 # Remove explicit "self" from method parameters. 589 590 if is_method and argnames and argnames[0] == "self": 591 del argnames[0] 592 593 # Copy and propagate the parameters. 594 595 self.importer.function_parameters[function_name] = \ 596 self.function_parameters[function_name] = argnames[:] 597 598 # Define all arguments/parameters in the local namespace. 599 600 locals = \ 601 self.importer.function_locals[function_name] = \ 602 self.function_locals[function_name] = {} 603 604 # Insert "self" into method locals. 605 606 if is_method: 607 argnames.insert(0, "self") 608 609 # Define "self" in terms of the class if in a method. 610 # This does not diminish the need for type-narrowing in the deducer. 611 612 if argnames: 613 if self.in_class and not self.in_function and argnames[0] == "self": 614 locals[argnames[0]] = Reference("<instance>", self.in_class) 615 else: 616 locals[argnames[0]] = Reference("<var>") 617 618 for argname in argnames[1:]: 619 locals[argname] = Reference("<var>") 620 621 globals = self.scope_globals[function_name] = set() 622 623 # Process the defaults. 624 625 defaults = self.importer.function_defaults[function_name] = \ 626 self.function_defaults[function_name] = [] 627 628 for argname, default in compiler.ast.get_defaults(n): 629 if default: 630 631 # Obtain any reference for the default. 632 633 name_ref = self.process_structure_node(default) 634 defaults.append((argname, name_ref.is_name() and name_ref.reference() or Reference("<var>"))) 635 636 # Reset conditional tracking to focus on the function contents. 637 638 in_conditional = self.in_conditional 639 self.in_conditional = False 640 641 in_function = self.in_function 642 self.in_function = function_name 643 644 self.enter_namespace(name) 645 646 # Define a leafname attribute value for the function instance. 647 648 ref = self.get_builtin_class("string") 649 self.reserve_constant(function_name, name, ref.get_origin()) 650 651 # Track attribute usage within the namespace. 652 653 path = self.get_namespace_path() 654 655 self.start_tracking(locals) 656 self.process_structure_node(n.code) 657 returns_value = self.stop_tracking() 658 659 # Record any null result. 660 661 is_initialiser = is_method and name == "__init__" 662 663 if not returns_value and not is_initialiser: 664 self.record_return_value(ResolvedNameRef("None", self.get_builtin("None"))) 665 666 # Exit to the parent. 667 668 self.exit_namespace() 669 670 # Update flags. 671 672 self.in_function = in_function 673 self.in_conditional = in_conditional 674 675 # Define the function using the appropriate name. 676 677 self.set_definition(name, "<function>") 678 679 # Where a function is set conditionally, assign the name. 680 681 if original_name: 682 self.process_assignment_for_object(original_name, compiler.ast.Name(name)) 683 684 def process_global_node(self, n): 685 686 """ 687 Process the given "global" node 'n'. 688 """ 689 690 path = self.get_namespace_path() 691 692 if path != self.name: 693 self.scope_globals[path].update(n.names) 694 695 def process_if_node(self, n): 696 697 """ 698 Process the given "if" node 'n'. 699 """ 700 701 tracker = self.trackers[-1] 702 tracker.new_branchpoint() 703 704 for test, body in n.tests: 705 self.process_structure_node(test) 706 707 tracker.new_branch() 708 709 in_conditional = self.in_conditional 710 self.in_conditional = True 711 self.process_structure_node(body) 712 self.in_conditional = in_conditional 713 714 tracker.shelve_branch() 715 716 # Maintain a branch for the else clause. 717 718 tracker.new_branch() 719 if n.else_: 720 self.process_structure_node(n.else_) 721 tracker.shelve_branch() 722 723 tracker.merge_branches() 724 725 def process_import_node(self, n): 726 727 "Process the given import node 'n'." 728 729 path = self.get_namespace_path() 730 731 # Load the mentioned module. 732 733 for name, alias in n.names: 734 if name == self.name: 735 raise InspectError("Cannot import the current module.", path, n) 736 737 self.set_module(alias or name.split(".")[-1], name) 738 self.queue_module(name, True) 739 740 def process_invocation_node(self, n): 741 742 "Process the given invocation node 'n'." 743 744 path = self.get_namespace_path() 745 746 in_invocation = self.in_invocation 747 self.in_invocation = None 748 749 # Process the arguments. 750 751 keywords = set() 752 753 for arg in n.args: 754 self.process_structure_node(arg) 755 if isinstance(arg, compiler.ast.Keyword): 756 keywords.add(arg.name) 757 758 keywords = list(keywords) 759 keywords.sort() 760 761 # Communicate to the invocation target expression that it forms the 762 # target of an invocation, potentially affecting attribute accesses. 763 764 self.in_invocation = len(n.args), keywords 765 766 # Process the expression, obtaining any identified reference. 767 768 name_ref = self.process_structure_node(n.node) 769 self.in_invocation = in_invocation 770 771 # Detect class invocations. 772 773 if isinstance(name_ref, ResolvedNameRef) and name_ref.has_kind("<class>"): 774 return InstanceRef(name_ref.reference().instance_of()) 775 776 elif isinstance(name_ref, (NameRef, AccessRef)): 777 return InvocationRef(name_ref) 778 779 # Provide a general reference to indicate that something is produced 780 # by the invocation, useful for retaining assignment expression 781 # details. 782 783 return VariableRef() 784 785 def process_lambda_node(self, n): 786 787 "Process the given lambda node 'n'." 788 789 name = self.get_lambda_name() 790 self.process_function_node(n, name) 791 792 origin = self.get_object_path(name) 793 794 if self.function_defaults.get(origin): 795 return None 796 else: 797 return ResolvedNameRef(name, Reference("<function>", origin)) 798 799 def process_logical_node(self, n): 800 801 "Process the given operator node 'n'." 802 803 self.process_operator_chain(n.nodes, self.process_structure_node) 804 805 def process_name_node(self, n): 806 807 "Process the given name node 'n'." 808 809 path = self.get_namespace_path() 810 811 # Find predefined constant names before anything else. 812 813 if n.name in predefined_constants: 814 ref = self.get_builtin(n.name) 815 value = ResolvedNameRef(n.name, ref) 816 return value 817 818 # Special names that have already been identified. 819 820 if n.name.startswith("$"): 821 value = self.get_special(n.name) 822 if value: 823 return value 824 825 # Special case for operator functions introduced through code 826 # transformations. 827 828 if n.name.startswith("$op"): 829 830 # Obtain the location of the actual function defined in the operator 831 # package. 832 833 op = n.name[len("$op"):] 834 835 # Attempt to get a reference. 836 837 ref = self.import_name_from_module(op, "operator") 838 839 # Record the imported name and provide the resolved name reference. 840 841 value = ResolvedNameRef(n.name, ref) 842 self.set_special(n.name, value) 843 return value 844 845 # Special case for print operations. 846 847 elif n.name.startswith("$print"): 848 849 # Attempt to get a reference. 850 851 ref = self.get_builtin("print_") 852 853 # Record the imported name and provide the resolved name reference. 854 855 value = ResolvedNameRef(n.name, ref) 856 self.set_special(n.name, value) 857 return value 858 859 # Test for self usage, which is only allowed in methods. 860 861 if n.name == "self" and not (self.in_function and self.in_class): 862 raise InspectError("Use of self is only allowed in methods.", path, n) 863 864 # Record usage of the name. 865 866 self.record_name(n.name) 867 868 # Search for unknown names in non-function scopes immediately. 869 # External names in functions are resolved later. 870 871 ref = self.find_name(n.name) 872 if ref: 873 self.record_name_access(n.name, True) 874 return ResolvedNameRef(n.name, ref, is_global=True) 875 876 # Explicitly-declared global names. 877 878 elif self.in_function and n.name in self.scope_globals[path]: 879 self.record_name_access(n.name, True) 880 return NameRef(n.name, is_global=True) 881 882 # Examine other names. 883 884 else: 885 886 # Check local names. 887 888 access_number = self.record_name_access(n.name) 889 890 # Local name. 891 892 if access_number is not None: 893 return LocalNameRef(n.name, access_number) 894 895 # Possible global or built-in name. 896 897 else: 898 self.record_name_access(n.name, True) 899 return NameRef(n.name, is_global=True) 900 901 def record_name_access(self, name, is_global=False): 902 903 """ 904 Record an access involving 'name' if the name is being tracked, using 905 'is_global' to indicate whether the name is global. 906 """ 907 908 name = self.get_name_for_tracking(name, is_global=is_global) 909 branches = self.trackers[-1].tracking_name(name) 910 if branches: 911 self.record_branches_for_access(branches, name, None) 912 return self.record_access_details(name, None, None, None) 913 return None 914 915 def process_operator_chain(self, nodes, fn): 916 917 """ 918 Process the given chain of 'nodes', applying 'fn' to each node or item. 919 Each node starts a new conditional region, effectively making a deeply- 920 nested collection of if-like statements. 921 """ 922 923 tracker = self.trackers[-1] 924 925 for item in nodes: 926 tracker.new_branchpoint() 927 tracker.new_branch() 928 fn(item) 929 930 for item in nodes[:-1]: 931 tracker.shelve_branch() 932 tracker.new_branch() 933 tracker.shelve_branch() 934 tracker.merge_branches() 935 936 tracker.shelve_branch() 937 tracker.merge_branches() 938 939 def process_try_node(self, n): 940 941 """ 942 Process the given "try...except" node 'n'. 943 """ 944 945 self.record_exception_handler() 946 947 tracker = self.trackers[-1] 948 tracker.new_branchpoint() 949 950 self.process_structure_node(n.body) 951 952 for name, var, handler in n.handlers: 953 if name is not None: 954 self.process_structure_node(name) 955 956 # Any abandoned branches from the body can now be resumed in a new 957 # branch. 958 959 tracker.resume_abandoned_branches() 960 961 # Establish the local for the handler. 962 963 if var is not None: 964 self.process_assignment_node(var, None) 965 if handler is not None: 966 self.process_structure_node(handler) 967 968 tracker.shelve_branch() 969 970 # The else clause maintains the usage from the body but without the 971 # abandoned branches since they would never lead to the else clause 972 # being executed. 973 974 if n.else_: 975 tracker.new_branch() 976 self.process_structure_node(n.else_) 977 tracker.shelve_branch() 978 979 # Without an else clause, a null branch propagates the successful 980 # outcome. 981 982 else: 983 tracker.new_branch() 984 tracker.shelve_branch() 985 986 tracker.merge_branches() 987 988 def process_try_finally_node(self, n): 989 990 """ 991 Process the given "try...finally" node 'n'. 992 """ 993 994 self.record_exception_handler() 995 996 tracker = self.trackers[-1] 997 self.process_structure_node(n.body) 998 999 # Any abandoned branches from the body can now be resumed. 1000 1001 branches = tracker.resume_all_abandoned_branches() 1002 self.process_structure_node(n.final) 1003 1004 # At the end of the finally clause, abandoned branches are discarded. 1005 1006 tracker.restore_active_branches(branches) 1007 1008 def process_while_node(self, n): 1009 1010 "Process the given while node 'n'." 1011 1012 tracker = self.trackers[-1] 1013 tracker.new_branchpoint(loop_node=True) 1014 1015 # Evaluate any test or iterator outside the loop. 1016 1017 self.process_structure_node(n.test) 1018 1019 # Propagate attribute usage to branches. 1020 1021 tracker.new_branch(loop_node=True) 1022 1023 # Enter the loop. 1024 1025 in_conditional = self.in_conditional 1026 self.in_conditional = True 1027 self.process_structure_node(n.body) 1028 self.in_conditional = in_conditional 1029 1030 # Continuing branches are resumed before any test. 1031 1032 tracker.resume_continuing_branches() 1033 1034 # Evaluate any continuation test within the body. 1035 1036 self.process_structure_node(n.test) 1037 1038 tracker.shelve_branch(loop_node=True) 1039 1040 # Support the non-looping condition. 1041 1042 tracker.new_branch() 1043 tracker.shelve_branch() 1044 1045 tracker.merge_branches() 1046 1047 # Evaluate any else clause outside branches. 1048 1049 if n.else_: 1050 self.process_structure_node(n.else_) 1051 1052 # Connect broken branches to the code after any loop. 1053 1054 tracker.resume_broken_branches() 1055 1056 # Branch tracking methods. 1057 1058 def start_tracking(self, names): 1059 1060 """ 1061 Start tracking attribute usage for names in the current namespace, 1062 immediately registering the given 'names'. 1063 """ 1064 1065 path = self.get_namespace_path() 1066 parent = self.trackers[-1] 1067 tracker = BranchTracker() 1068 self.trackers.append(tracker) 1069 1070 # Record the given names established as new branches. 1071 1072 tracker.assign_names(names) 1073 1074 def assign_name(self, name, name_ref): 1075 1076 "Assign to 'name' the given 'name_ref' in the current namespace." 1077 1078 name = self.get_name_for_tracking(name) 1079 self.trackers[-1].assign_names([name], [name_ref]) 1080 1081 def stop_tracking(self): 1082 1083 """ 1084 Stop tracking attribute usage, recording computed usage for the current 1085 namespace. Indicate whether a value is always returned from the 1086 namespace. 1087 """ 1088 1089 path = self.get_namespace_path() 1090 tracker = self.trackers.pop() 1091 self.record_assignments_for_access(tracker) 1092 1093 self.attr_usage[path] = tracker.get_all_usage() 1094 self.name_initialisers[path] = tracker.get_all_values() 1095 1096 return tracker.returns_value() 1097 1098 def start_tracking_in_module(self): 1099 1100 "Start tracking attribute usage in the module." 1101 1102 tracker = BranchTracker() 1103 self.trackers.append(tracker) 1104 1105 def stop_tracking_in_module(self): 1106 1107 "Stop tracking attribute usage in the module." 1108 1109 tracker = self.trackers[0] 1110 self.record_assignments_for_access(tracker) 1111 self.attr_usage[self.name] = tracker.get_all_usage() 1112 self.name_initialisers[self.name] = tracker.get_all_values() 1113 1114 def record_assignments_for_access(self, tracker): 1115 1116 """ 1117 For the current path, use the given 'tracker' to record assignment 1118 version information for attribute accesses. 1119 """ 1120 1121 path = self.get_path_for_access() 1122 1123 if not self.attr_accessor_branches.has_key(path): 1124 return 1125 1126 init_item(self.attr_accessors, path, dict) 1127 attr_accessors = self.attr_accessors[path] 1128 1129 # Obtain the branches applying during each access. 1130 1131 for access, all_branches in self.attr_accessor_branches[path].items(): 1132 name, attrnames = access 1133 init_item(attr_accessors, access, list) 1134 1135 # Obtain the assignments applying to each branch. 1136 1137 for branches in all_branches: 1138 positions = tracker.get_assignment_positions_for_branches(name, branches) 1139 1140 # Detect missing name information. 1141 1142 if None in positions: 1143 globals = self.global_attr_accesses.get(path) 1144 accesses = globals and globals.get(name) 1145 if not accesses: 1146 print >>sys.stderr, "In %s, %s may not be defined when used." % ( 1147 self.get_namespace_path(), name) 1148 positions.remove(None) 1149 1150 attr_accessors[access].append(positions) 1151 1152 def record_branches_for_access(self, branches, name, attrnames): 1153 1154 """ 1155 Record the given 'branches' for an access involving the given 'name' and 1156 'attrnames'. 1157 """ 1158 1159 access = name, attrnames 1160 path = self.get_path_for_access() 1161 1162 init_item(self.attr_accessor_branches, path, dict) 1163 attr_accessor_branches = self.attr_accessor_branches[path] 1164 1165 init_item(attr_accessor_branches, access, list) 1166 attr_accessor_branches[access].append(branches) 1167 1168 def record_access_details(self, name, attrnames, assignment, invocation): 1169 1170 """ 1171 For the given 'name' and 'attrnames', record an access indicating 1172 whether an 'assignment' or an 'invocation' is occurring. 1173 1174 These details correspond to accesses otherwise recorded by the attribute 1175 accessor and attribute access dictionaries. 1176 """ 1177 1178 access = name, attrnames 1179 path = self.get_path_for_access() 1180 1181 init_item(self.attr_access_modifiers, path, dict) 1182 init_item(self.attr_access_modifiers[path], access, list) 1183 1184 access_number = len(self.attr_access_modifiers[path][access]) 1185 self.attr_access_modifiers[path][access].append((assignment, invocation)) 1186 return access_number 1187 1188 def record_global_access_details(self, name, attrnames): 1189 1190 """ 1191 Record details of a global access via the given 'name' involving the 1192 indicated 'attrnames'. 1193 """ 1194 1195 path = self.get_namespace_path() 1196 1197 init_item(self.global_attr_accesses, path, dict) 1198 init_item(self.global_attr_accesses[path], name, set) 1199 self.global_attr_accesses[path][name].add(attrnames) 1200 1201 # Namespace modification. 1202 1203 def record_name(self, name): 1204 1205 "Record the use of 'name' in a namespace." 1206 1207 path = self.get_namespace_path() 1208 init_item(self.names_used, path, set) 1209 self.names_used[path].add(name) 1210 1211 def set_module(self, name, module_name): 1212 1213 """ 1214 Set a module in the current namespace using the given 'name' associated 1215 with the corresponding 'module_name'. 1216 """ 1217 1218 if name: 1219 self.set_general_local(name, Reference("<module>", module_name)) 1220 1221 def set_definition(self, name, kind): 1222 1223 """ 1224 Set the definition having the given 'name' and 'kind'. 1225 1226 Definitions are set in the static namespace hierarchy, but they can also 1227 be recorded for function locals. 1228 """ 1229 1230 if self.is_global(name): 1231 print >>sys.stderr, "In %s, %s is defined as being global." % ( 1232 self.get_namespace_path(), name) 1233 1234 path = self.get_object_path(name) 1235 self.set_object(path, kind) 1236 1237 ref = self.get_object(path) 1238 if ref.get_kind() == "<var>": 1239 print >>sys.stderr, "In %s, %s is defined more than once." % ( 1240 self.get_namespace_path(), name) 1241 1242 if not self.is_global(name) and self.in_function: 1243 self.set_function_local(name, ref) 1244 1245 def set_function_local(self, name, ref=None): 1246 1247 "Set the local with the given 'name' and optional 'ref'." 1248 1249 locals = self.function_locals[self.get_namespace_path()] 1250 multiple = not ref or locals.has_key(name) and locals[name] != ref 1251 locals[name] = multiple and Reference("<var>") or ref 1252 1253 def assign_general_local(self, name, name_ref): 1254 1255 """ 1256 Set for 'name' the given 'name_ref', recording the name for attribute 1257 usage tracking. 1258 """ 1259 1260 self.set_general_local(name, name_ref) 1261 self.assign_name(name, name_ref) 1262 1263 def set_general_local(self, name, value=None): 1264 1265 """ 1266 Set the 'name' with optional 'value' in any kind of local namespace, 1267 where the 'value' should be a reference if specified. 1268 """ 1269 1270 init_value = self.get_initialising_value(value) 1271 1272 # Module global names. 1273 1274 if self.is_global(name): 1275 path = self.get_global_path(name) 1276 self.set_object(path, init_value) 1277 1278 # Function local names. 1279 1280 elif self.in_function: 1281 path = self.get_object_path(name) 1282 self.set_function_local(name, init_value) 1283 1284 # Other namespaces (classes). 1285 1286 else: 1287 path = self.get_object_path(name) 1288 self.set_name(name, init_value) 1289 1290 def set_name(self, name, ref=None): 1291 1292 "Attach the 'name' with optional 'ref' to the current namespace." 1293 1294 self.set_object(self.get_object_path(name), ref) 1295 1296 def set_instance_attr(self, name, ref=None): 1297 1298 """ 1299 Add an instance attribute of the given 'name' to the current class, 1300 using the optional 'ref'. 1301 """ 1302 1303 self._set_instance_attr(self.in_class, name, ref) 1304 1305 def _set_instance_attr(self, path, name, ref=None): 1306 1307 init_item(self.instance_attrs, path, set) 1308 self.instance_attrs[path].add(name) 1309 1310 if ref: 1311 init_item(self.instance_attr_constants, path, dict) 1312 self.instance_attr_constants[path][name] = ref 1313 1314 def get_initialising_value(self, value): 1315 1316 "Return a suitable initialiser reference for 'value'." 1317 1318 # Includes LiteralSequenceRef, ResolvedNameRef... 1319 1320 if isinstance(value, (NameRef, AccessRef, InstanceRef)): 1321 return value.reference() 1322 1323 # In general, invocations do not produce known results. However, the 1324 # name initialisers are resolved once a module has been inspected. 1325 1326 elif isinstance(value, InvocationRef): 1327 return value.reference() 1328 1329 # Variable references are unknown results. 1330 1331 elif isinstance(value, VariableRef): 1332 return value.reference() 1333 1334 else: 1335 return value 1336 1337 # Static, program-relative naming. 1338 1339 def find_name(self, name): 1340 1341 """ 1342 Return the qualified name for the given 'name' used in the current 1343 non-function namespace. 1344 """ 1345 1346 path = self.get_namespace_path() 1347 ref = None 1348 1349 if not self.in_function and name not in predefined_constants: 1350 if self.in_class: 1351 ref = self.get_object(self.get_object_path(name), False) 1352 if not ref: 1353 ref = self.get_global_or_builtin(name) 1354 1355 return ref 1356 1357 def get_class(self, node): 1358 1359 """ 1360 Use the given 'node' to obtain the identity of a class. Return a 1361 reference for the class. Unresolved dependencies are permitted and must 1362 be resolved later. 1363 """ 1364 1365 ref = self._get_class(node) 1366 return ref.has_kind(["<class>", "<depends>"]) and ref or None 1367 1368 def _get_class(self, node): 1369 1370 """ 1371 Use the given 'node' to find a class definition. Return a reference to 1372 the class. 1373 """ 1374 1375 if isinstance(node, compiler.ast.Getattr): 1376 1377 # Obtain the identity of the access target. 1378 1379 ref = self._get_class(node.expr) 1380 1381 # Where the target is a class or module, obtain the identity of the 1382 # attribute. 1383 1384 if ref.has_kind(["<function>", "<var>"]): 1385 return None 1386 else: 1387 attrname = "%s.%s" % (ref.get_origin(), node.attrname) 1388 return self.get_object(attrname) 1389 1390 # Names can be module-level or built-in. 1391 1392 elif isinstance(node, compiler.ast.Name): 1393 1394 # Record usage of the name and attempt to identify it. 1395 1396 self.record_name(node.name) 1397 return self.find_name(node.name) 1398 else: 1399 return None 1400 1401 def get_constant(self, name, value): 1402 1403 "Return a constant reference for the given type 'name' and 'value'." 1404 1405 ref = self.get_builtin_class(name) 1406 return self.get_constant_reference(ref, value) 1407 1408 def get_literal_instance(self, n, name=None): 1409 1410 """ 1411 For node 'n', return a reference to an instance of 'name', or if 'name' 1412 is not specified, deduce the type from the value. 1413 """ 1414 1415 # Handle stray None constants (Sliceobj seems to produce them). 1416 1417 if name == "NoneType": 1418 return self.process_name_node(compiler.ast.Name("None")) 1419 1420 # Obtain the details of the literal itself. 1421 # An alias to the type is generated for sequences. 1422 1423 if name in ("dict", "list", "tuple"): 1424 ref = self.get_builtin_class(name) 1425 self.set_special_literal(name, ref) 1426 return self.process_literal_sequence_node(n, name, ref, LiteralSequenceRef) 1427 1428 # Constant values are independently recorded. 1429 1430 else: 1431 value, typename, encoding = self.get_constant_value(n.value, n.literals) 1432 ref = self.get_builtin_class(typename) 1433 return self.get_constant_reference(ref, value, encoding) 1434 1435 # Special names. 1436 1437 def get_special(self, name): 1438 1439 "Return any stored value for the given special 'name'." 1440 1441 value = self.special.get(name) 1442 if value: 1443 ref, paths = value 1444 else: 1445 ref = None 1446 return ref 1447 1448 def set_special(self, name, value): 1449 1450 """ 1451 Set a special 'name' that merely tracks the use of an implicit object 1452 'value'. 1453 """ 1454 1455 if not self.special.has_key(name): 1456 paths = set() 1457 self.special[name] = value, paths 1458 else: 1459 _ref, paths = self.special[name] 1460 1461 paths.add(self.get_namespace_path()) 1462 1463 def set_special_literal(self, name, ref): 1464 1465 """ 1466 Set a special name for the literal type 'name' having type 'ref'. Such 1467 special names provide a way of referring to literal object types. 1468 """ 1469 1470 literal_name = "$L%s" % name 1471 value = ResolvedNameRef(literal_name, ref) 1472 self.set_special(literal_name, value) 1473 1474 # Exceptions. 1475 1476 def record_exception_handler(self): 1477 1478 "Record the current namespace as employing an exception handler." 1479 1480 self.exception_namespaces.add(self.get_namespace_path()) 1481 1482 # Return values. 1483 1484 def record_return_value(self, expr): 1485 1486 "Record the given return 'expr'." 1487 1488 path = self.get_namespace_path() 1489 init_item(self.return_values, path, set) 1490 self.return_values[path].add(expr) 1491 1492 # vim: tabstop=4 expandtab shiftwidth=4