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