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