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