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.process_structure(n) 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 self.stop_tracking() 658 659 # Exit to the parent. 660 661 self.exit_namespace() 662 663 # Update flags. 664 665 self.in_function = in_function 666 self.in_conditional = in_conditional 667 668 # Define the function using the appropriate name. 669 670 self.set_definition(name, "<function>") 671 672 # Where a function is set conditionally, assign the name. 673 674 if original_name: 675 self.process_assignment_for_object(original_name, compiler.ast.Name(name)) 676 677 def process_global_node(self, n): 678 679 """ 680 Process the given "global" node 'n'. 681 """ 682 683 path = self.get_namespace_path() 684 685 if path != self.name: 686 self.scope_globals[path].update(n.names) 687 688 def process_if_node(self, n): 689 690 """ 691 Process the given "if" node 'n'. 692 """ 693 694 tracker = self.trackers[-1] 695 tracker.new_branchpoint() 696 697 for test, body in n.tests: 698 self.process_structure_node(test) 699 700 tracker.new_branch() 701 702 in_conditional = self.in_conditional 703 self.in_conditional = True 704 self.process_structure_node(body) 705 self.in_conditional = in_conditional 706 707 tracker.shelve_branch() 708 709 # Maintain a branch for the else clause. 710 711 tracker.new_branch() 712 if n.else_: 713 self.process_structure_node(n.else_) 714 tracker.shelve_branch() 715 716 tracker.merge_branches() 717 718 def process_import_node(self, n): 719 720 "Process the given import node 'n'." 721 722 path = self.get_namespace_path() 723 724 # Load the mentioned module. 725 726 for name, alias in n.names: 727 if name == self.name: 728 raise InspectError("Cannot import the current module.", path, n) 729 730 self.set_module(alias or name.split(".")[-1], name) 731 self.queue_module(name, True) 732 733 def process_invocation_node(self, n): 734 735 "Process the given invocation node 'n'." 736 737 path = self.get_namespace_path() 738 739 in_invocation = self.in_invocation 740 self.in_invocation = None 741 742 # Process the arguments. 743 744 keywords = set() 745 746 for arg in n.args: 747 self.process_structure_node(arg) 748 if isinstance(arg, compiler.ast.Keyword): 749 keywords.add(arg.name) 750 751 keywords = list(keywords) 752 keywords.sort() 753 754 # Communicate to the invocation target expression that it forms the 755 # target of an invocation, potentially affecting attribute accesses. 756 757 self.in_invocation = len(n.args), keywords 758 759 # Process the expression, obtaining any identified reference. 760 761 name_ref = self.process_structure_node(n.node) 762 self.in_invocation = in_invocation 763 764 # Detect class invocations. 765 766 if isinstance(name_ref, ResolvedNameRef) and name_ref.has_kind("<class>"): 767 return InstanceRef(name_ref.reference().instance_of()) 768 769 elif isinstance(name_ref, NameRef): 770 return InvocationRef(name_ref) 771 772 # Provide a general reference to indicate that something is produced 773 # by the invocation, useful for retaining assignment expression 774 # details. 775 776 return VariableRef() 777 778 def process_lambda_node(self, n): 779 780 "Process the given lambda node 'n'." 781 782 name = self.get_lambda_name() 783 self.process_function_node(n, name) 784 785 origin = self.get_object_path(name) 786 787 if self.function_defaults.get(origin): 788 return None 789 else: 790 return ResolvedNameRef(name, Reference("<function>", origin)) 791 792 def process_logical_node(self, n): 793 794 "Process the given operator node 'n'." 795 796 self.process_operator_chain(n.nodes, self.process_structure_node) 797 798 def process_name_node(self, n): 799 800 "Process the given name node 'n'." 801 802 path = self.get_namespace_path() 803 804 # Find predefined constant names before anything else. 805 806 if n.name in predefined_constants: 807 ref = self.get_builtin(n.name) 808 value = ResolvedNameRef(n.name, ref) 809 return value 810 811 # Special names that have already been identified. 812 813 if n.name.startswith("$"): 814 value = self.get_special(n.name) 815 if value: 816 return value 817 818 # Special case for operator functions introduced through code 819 # transformations. 820 821 if n.name.startswith("$op"): 822 823 # Obtain the location of the actual function defined in the operator 824 # package. 825 826 op = n.name[len("$op"):] 827 828 # Attempt to get a reference. 829 830 ref = self.import_name_from_module(op, "operator") 831 832 # Record the imported name and provide the resolved name reference. 833 834 value = ResolvedNameRef(n.name, ref) 835 self.set_special(n.name, value) 836 return value 837 838 # Special case for print operations. 839 840 elif n.name.startswith("$print"): 841 842 # Attempt to get a reference. 843 844 ref = self.get_builtin("print_") 845 846 # Record the imported name and provide the resolved name reference. 847 848 value = ResolvedNameRef(n.name, ref) 849 self.set_special(n.name, value) 850 return value 851 852 # Test for self usage, which is only allowed in methods. 853 854 if n.name == "self" and not (self.in_function and self.in_class): 855 raise InspectError("Use of self is only allowed in methods.", path, n) 856 857 # Record usage of the name. 858 859 self.record_name(n.name) 860 861 # Search for unknown names in non-function scopes immediately. 862 # External names in functions are resolved later. 863 864 ref = self.find_name(n.name) 865 if ref: 866 self.record_name_access(n.name, True) 867 return ResolvedNameRef(n.name, ref, is_global=True) 868 869 # Explicitly-declared global names. 870 871 elif self.in_function and n.name in self.scope_globals[path]: 872 self.record_name_access(n.name, True) 873 return NameRef(n.name, is_global=True) 874 875 # Examine other names. 876 877 else: 878 879 # Check local names. 880 881 access_number = self.record_name_access(n.name) 882 883 # Local name. 884 885 if access_number is not None: 886 return LocalNameRef(n.name, access_number) 887 888 # Possible global or built-in name. 889 890 else: 891 self.record_name_access(n.name, True) 892 return NameRef(n.name, is_global=True) 893 894 def record_name_access(self, name, is_global=False): 895 896 """ 897 Record an access involving 'name' if the name is being tracked, using 898 'is_global' to indicate whether the name is global. 899 """ 900 901 name = self.get_name_for_tracking(name, is_global=is_global) 902 branches = self.trackers[-1].tracking_name(name) 903 if branches: 904 self.record_branches_for_access(branches, name, None) 905 return self.record_access_details(name, None, None, None) 906 return None 907 908 def process_operator_chain(self, nodes, fn): 909 910 """ 911 Process the given chain of 'nodes', applying 'fn' to each node or item. 912 Each node starts a new conditional region, effectively making a deeply- 913 nested collection of if-like statements. 914 """ 915 916 tracker = self.trackers[-1] 917 918 for item in nodes: 919 tracker.new_branchpoint() 920 tracker.new_branch() 921 fn(item) 922 923 for item in nodes[:-1]: 924 tracker.shelve_branch() 925 tracker.new_branch() 926 tracker.shelve_branch() 927 tracker.merge_branches() 928 929 tracker.shelve_branch() 930 tracker.merge_branches() 931 932 def process_try_node(self, n): 933 934 """ 935 Process the given "try...except" node 'n'. 936 """ 937 938 self.record_exception_handler() 939 940 tracker = self.trackers[-1] 941 tracker.new_branchpoint() 942 943 self.process_structure_node(n.body) 944 945 for name, var, handler in n.handlers: 946 if name is not None: 947 self.process_structure_node(name) 948 949 # Any abandoned branches from the body can now be resumed in a new 950 # branch. 951 952 tracker.resume_abandoned_branches() 953 954 # Establish the local for the handler. 955 956 if var is not None: 957 self.process_assignment_node(var, None) 958 if handler is not None: 959 self.process_structure_node(handler) 960 961 tracker.shelve_branch() 962 963 # The else clause maintains the usage from the body but without the 964 # abandoned branches since they would never lead to the else clause 965 # being executed. 966 967 if n.else_: 968 tracker.new_branch() 969 self.process_structure_node(n.else_) 970 tracker.shelve_branch() 971 972 # Without an else clause, a null branch propagates the successful 973 # outcome. 974 975 else: 976 tracker.new_branch() 977 tracker.shelve_branch() 978 979 tracker.merge_branches() 980 981 def process_try_finally_node(self, n): 982 983 """ 984 Process the given "try...finally" node 'n'. 985 """ 986 987 self.record_exception_handler() 988 989 tracker = self.trackers[-1] 990 self.process_structure_node(n.body) 991 992 # Any abandoned branches from the body can now be resumed. 993 994 branches = tracker.resume_all_abandoned_branches() 995 self.process_structure_node(n.final) 996 997 # At the end of the finally clause, abandoned branches are discarded. 998 999 tracker.restore_active_branches(branches) 1000 1001 def process_while_node(self, n): 1002 1003 "Process the given while node 'n'." 1004 1005 tracker = self.trackers[-1] 1006 tracker.new_branchpoint(loop_node=True) 1007 1008 # Evaluate any test or iterator outside the loop. 1009 1010 self.process_structure_node(n.test) 1011 1012 # Propagate attribute usage to branches. 1013 1014 tracker.new_branch(loop_node=True) 1015 1016 # Enter the loop. 1017 1018 in_conditional = self.in_conditional 1019 self.in_conditional = True 1020 self.process_structure_node(n.body) 1021 self.in_conditional = in_conditional 1022 1023 # Continuing branches are resumed before any test. 1024 1025 tracker.resume_continuing_branches() 1026 1027 # Evaluate any continuation test within the body. 1028 1029 self.process_structure_node(n.test) 1030 1031 tracker.shelve_branch(loop_node=True) 1032 1033 # Support the non-looping condition. 1034 1035 tracker.new_branch() 1036 tracker.shelve_branch() 1037 1038 tracker.merge_branches() 1039 1040 # Evaluate any else clause outside branches. 1041 1042 if n.else_: 1043 self.process_structure_node(n.else_) 1044 1045 # Connect broken branches to the code after any loop. 1046 1047 tracker.resume_broken_branches() 1048 1049 # Branch tracking methods. 1050 1051 def start_tracking(self, names): 1052 1053 """ 1054 Start tracking attribute usage for names in the current namespace, 1055 immediately registering the given 'names'. 1056 """ 1057 1058 path = self.get_namespace_path() 1059 parent = self.trackers[-1] 1060 tracker = BranchTracker() 1061 self.trackers.append(tracker) 1062 1063 # Record the given names established as new branches. 1064 1065 tracker.assign_names(names) 1066 1067 def assign_name(self, name, name_ref): 1068 1069 "Assign to 'name' the given 'name_ref' in the current namespace." 1070 1071 name = self.get_name_for_tracking(name) 1072 self.trackers[-1].assign_names([name], [name_ref]) 1073 1074 def stop_tracking(self): 1075 1076 """ 1077 Stop tracking attribute usage, recording computed usage for the current 1078 namespace. 1079 """ 1080 1081 path = self.get_namespace_path() 1082 tracker = self.trackers.pop() 1083 self.record_assignments_for_access(tracker) 1084 1085 self.attr_usage[path] = tracker.get_all_usage() 1086 self.name_initialisers[path] = tracker.get_all_values() 1087 1088 def start_tracking_in_module(self): 1089 1090 "Start tracking attribute usage in the module." 1091 1092 tracker = BranchTracker() 1093 self.trackers.append(tracker) 1094 1095 def stop_tracking_in_module(self): 1096 1097 "Stop tracking attribute usage in the module." 1098 1099 tracker = self.trackers[0] 1100 self.record_assignments_for_access(tracker) 1101 self.attr_usage[self.name] = tracker.get_all_usage() 1102 self.name_initialisers[self.name] = tracker.get_all_values() 1103 1104 def record_assignments_for_access(self, tracker): 1105 1106 """ 1107 For the current path, use the given 'tracker' to record assignment 1108 version information for attribute accesses. 1109 """ 1110 1111 path = self.get_path_for_access() 1112 1113 if not self.attr_accessor_branches.has_key(path): 1114 return 1115 1116 init_item(self.attr_accessors, path, dict) 1117 attr_accessors = self.attr_accessors[path] 1118 1119 # Obtain the branches applying during each access. 1120 1121 for access, all_branches in self.attr_accessor_branches[path].items(): 1122 name, attrnames = access 1123 init_item(attr_accessors, access, list) 1124 1125 # Obtain the assignments applying to each branch. 1126 1127 for branches in all_branches: 1128 positions = tracker.get_assignment_positions_for_branches(name, branches) 1129 1130 # Detect missing name information. 1131 1132 if None in positions: 1133 globals = self.global_attr_accesses.get(path) 1134 accesses = globals and globals.get(name) 1135 if not accesses: 1136 print >>sys.stderr, "In %s, %s may not be defined when used." % ( 1137 self.get_namespace_path(), name) 1138 positions.remove(None) 1139 1140 attr_accessors[access].append(positions) 1141 1142 def record_branches_for_access(self, branches, name, attrnames): 1143 1144 """ 1145 Record the given 'branches' for an access involving the given 'name' and 1146 'attrnames'. 1147 """ 1148 1149 access = name, attrnames 1150 path = self.get_path_for_access() 1151 1152 init_item(self.attr_accessor_branches, path, dict) 1153 attr_accessor_branches = self.attr_accessor_branches[path] 1154 1155 init_item(attr_accessor_branches, access, list) 1156 attr_accessor_branches[access].append(branches) 1157 1158 def record_access_details(self, name, attrnames, assignment, invocation): 1159 1160 """ 1161 For the given 'name' and 'attrnames', record an access indicating 1162 whether an 'assignment' or an 'invocation' is occurring. 1163 1164 These details correspond to accesses otherwise recorded by the attribute 1165 accessor and attribute access dictionaries. 1166 """ 1167 1168 access = name, attrnames 1169 path = self.get_path_for_access() 1170 1171 init_item(self.attr_access_modifiers, path, dict) 1172 init_item(self.attr_access_modifiers[path], access, list) 1173 1174 access_number = len(self.attr_access_modifiers[path][access]) 1175 self.attr_access_modifiers[path][access].append((assignment, invocation)) 1176 return access_number 1177 1178 def record_global_access_details(self, name, attrnames): 1179 1180 """ 1181 Record details of a global access via the given 'name' involving the 1182 indicated 'attrnames'. 1183 """ 1184 1185 path = self.get_namespace_path() 1186 1187 init_item(self.global_attr_accesses, path, dict) 1188 init_item(self.global_attr_accesses[path], name, set) 1189 self.global_attr_accesses[path][name].add(attrnames) 1190 1191 # Namespace modification. 1192 1193 def record_name(self, name): 1194 1195 "Record the use of 'name' in a namespace." 1196 1197 path = self.get_namespace_path() 1198 init_item(self.names_used, path, set) 1199 self.names_used[path].add(name) 1200 1201 def set_module(self, name, module_name): 1202 1203 """ 1204 Set a module in the current namespace using the given 'name' associated 1205 with the corresponding 'module_name'. 1206 """ 1207 1208 if name: 1209 self.set_general_local(name, Reference("<module>", module_name)) 1210 1211 def set_definition(self, name, kind): 1212 1213 """ 1214 Set the definition having the given 'name' and 'kind'. 1215 1216 Definitions are set in the static namespace hierarchy, but they can also 1217 be recorded for function locals. 1218 """ 1219 1220 if self.is_global(name): 1221 print >>sys.stderr, "In %s, %s is defined as being global." % ( 1222 self.get_namespace_path(), name) 1223 1224 path = self.get_object_path(name) 1225 self.set_object(path, kind) 1226 1227 ref = self.get_object(path) 1228 if ref.get_kind() == "<var>": 1229 print >>sys.stderr, "In %s, %s is defined more than once." % ( 1230 self.get_namespace_path(), name) 1231 1232 if not self.is_global(name) and self.in_function: 1233 self.set_function_local(name, ref) 1234 1235 def set_function_local(self, name, ref=None): 1236 1237 "Set the local with the given 'name' and optional 'ref'." 1238 1239 locals = self.function_locals[self.get_namespace_path()] 1240 multiple = not ref or locals.has_key(name) and locals[name] != ref 1241 locals[name] = multiple and Reference("<var>") or ref 1242 1243 def assign_general_local(self, name, name_ref): 1244 1245 """ 1246 Set for 'name' the given 'name_ref', recording the name for attribute 1247 usage tracking. 1248 """ 1249 1250 self.set_general_local(name, name_ref) 1251 self.assign_name(name, name_ref) 1252 1253 def set_general_local(self, name, value=None): 1254 1255 """ 1256 Set the 'name' with optional 'value' in any kind of local namespace, 1257 where the 'value' should be a reference if specified. 1258 """ 1259 1260 init_value = self.get_initialising_value(value) 1261 1262 # Module global names. 1263 1264 if self.is_global(name): 1265 path = self.get_global_path(name) 1266 self.set_object(path, init_value) 1267 1268 # Function local names. 1269 1270 elif self.in_function: 1271 path = self.get_object_path(name) 1272 self.set_function_local(name, init_value) 1273 1274 # Other namespaces (classes). 1275 1276 else: 1277 path = self.get_object_path(name) 1278 self.set_name(name, init_value) 1279 1280 def set_name(self, name, ref=None): 1281 1282 "Attach the 'name' with optional 'ref' to the current namespace." 1283 1284 self.set_object(self.get_object_path(name), ref) 1285 1286 def set_instance_attr(self, name, ref=None): 1287 1288 """ 1289 Add an instance attribute of the given 'name' to the current class, 1290 using the optional 'ref'. 1291 """ 1292 1293 self._set_instance_attr(self.in_class, name, ref) 1294 1295 def _set_instance_attr(self, path, name, ref=None): 1296 1297 init_item(self.instance_attrs, path, set) 1298 self.instance_attrs[path].add(name) 1299 1300 if ref: 1301 init_item(self.instance_attr_constants, path, dict) 1302 self.instance_attr_constants[path][name] = ref 1303 1304 def get_initialising_value(self, value): 1305 1306 "Return a suitable initialiser reference for 'value'." 1307 1308 # Includes LiteralSequenceRef, ResolvedNameRef... 1309 1310 if isinstance(value, (NameRef, AccessRef, InstanceRef)): 1311 return value.reference() 1312 1313 # In general, invocations do not produce known results. However, the 1314 # name initialisers are resolved once a module has been inspected. 1315 1316 elif isinstance(value, InvocationRef): 1317 return value.reference() 1318 1319 # Variable references are unknown results. 1320 1321 elif isinstance(value, VariableRef): 1322 return value.reference() 1323 1324 else: 1325 return value 1326 1327 # Static, program-relative naming. 1328 1329 def find_name(self, name): 1330 1331 """ 1332 Return the qualified name for the given 'name' used in the current 1333 non-function namespace. 1334 """ 1335 1336 path = self.get_namespace_path() 1337 ref = None 1338 1339 if not self.in_function and name not in predefined_constants: 1340 if self.in_class: 1341 ref = self.get_object(self.get_object_path(name), False) 1342 if not ref: 1343 ref = self.get_global_or_builtin(name) 1344 1345 return ref 1346 1347 def get_class(self, node): 1348 1349 """ 1350 Use the given 'node' to obtain the identity of a class. Return a 1351 reference for the class. Unresolved dependencies are permitted and must 1352 be resolved later. 1353 """ 1354 1355 ref = self._get_class(node) 1356 return ref.has_kind(["<class>", "<depends>"]) and ref or None 1357 1358 def _get_class(self, node): 1359 1360 """ 1361 Use the given 'node' to find a class definition. Return a reference to 1362 the class. 1363 """ 1364 1365 if isinstance(node, compiler.ast.Getattr): 1366 1367 # Obtain the identity of the access target. 1368 1369 ref = self._get_class(node.expr) 1370 1371 # Where the target is a class or module, obtain the identity of the 1372 # attribute. 1373 1374 if ref.has_kind(["<function>", "<var>"]): 1375 return None 1376 else: 1377 attrname = "%s.%s" % (ref.get_origin(), node.attrname) 1378 return self.get_object(attrname) 1379 1380 # Names can be module-level or built-in. 1381 1382 elif isinstance(node, compiler.ast.Name): 1383 1384 # Record usage of the name and attempt to identify it. 1385 1386 self.record_name(node.name) 1387 return self.find_name(node.name) 1388 else: 1389 return None 1390 1391 def get_constant(self, name, value): 1392 1393 "Return a constant reference for the given type 'name' and 'value'." 1394 1395 ref = self.get_builtin_class(name) 1396 return self.get_constant_reference(ref, value) 1397 1398 def get_literal_instance(self, n, name=None): 1399 1400 """ 1401 For node 'n', return a reference to an instance of 'name', or if 'name' 1402 is not specified, deduce the type from the value. 1403 """ 1404 1405 # Handle stray None constants (Sliceobj seems to produce them). 1406 1407 if name == "NoneType": 1408 return self.process_name_node(compiler.ast.Name("None")) 1409 1410 # Obtain the details of the literal itself. 1411 # An alias to the type is generated for sequences. 1412 1413 if name in ("dict", "list", "tuple"): 1414 ref = self.get_builtin_class(name) 1415 self.set_special_literal(name, ref) 1416 return self.process_literal_sequence_node(n, name, ref, LiteralSequenceRef) 1417 1418 # Constant values are independently recorded. 1419 1420 else: 1421 value, typename, encoding = self.get_constant_value(n.value, n.literals) 1422 ref = self.get_builtin_class(typename) 1423 return self.get_constant_reference(ref, value, encoding) 1424 1425 # Special names. 1426 1427 def get_special(self, name): 1428 1429 "Return any stored value for the given special 'name'." 1430 1431 value = self.special.get(name) 1432 if value: 1433 ref, paths = value 1434 else: 1435 ref = None 1436 return ref 1437 1438 def set_special(self, name, value): 1439 1440 """ 1441 Set a special 'name' that merely tracks the use of an implicit object 1442 'value'. 1443 """ 1444 1445 if not self.special.has_key(name): 1446 paths = set() 1447 self.special[name] = value, paths 1448 else: 1449 _ref, paths = self.special[name] 1450 1451 paths.add(self.get_namespace_path()) 1452 1453 def set_special_literal(self, name, ref): 1454 1455 """ 1456 Set a special name for the literal type 'name' having type 'ref'. Such 1457 special names provide a way of referring to literal object types. 1458 """ 1459 1460 literal_name = "$L%s" % name 1461 value = ResolvedNameRef(literal_name, ref) 1462 self.set_special(literal_name, value) 1463 1464 # Exceptions. 1465 1466 def record_exception_handler(self): 1467 1468 "Record the current namespace as employing an exception handler." 1469 1470 self.exception_namespaces.add(self.get_namespace_path()) 1471 1472 # vim: tabstop=4 expandtab shiftwidth=4