1 #!/usr/bin/env python 2 3 """ 4 Inspect source files, obtaining details of classes and attributes. 5 6 Copyright (C) 2007, 2008, 2009 Paul Boddie <paul@boddie.org.uk> 7 8 This program is free software; you can redistribute it and/or modify it under 9 the terms of the GNU General Public License as published by the Free Software 10 Foundation; either version 3 of the License, or (at your option) any later 11 version. 12 13 This program is distributed in the hope that it will be useful, but WITHOUT 14 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS 15 FOR A PARTICULAR PURPOSE. See the GNU General Public License for more 16 details. 17 18 You should have received a copy of the GNU General Public License along with 19 this program. If not, see <http://www.gnu.org/licenses/>. 20 21 -------- 22 23 The results of inspecting a module are as follows: 24 25 Constants 26 --------- 27 28 All constants identified within the code shall be registered. 29 30 Classes 31 ------- 32 33 All global classes shall be registered; local classes (within functions) or 34 nested classes (within classes) are not currently registered. 35 36 Base classes must be detected and constant. 37 38 All classes without bases are made to inherit from __builtins__.object in order 39 to support some standard methods. 40 41 Functions 42 --------- 43 44 All functions and lambda definitions shall be registered. 45 46 Namespaces 47 ---------- 48 49 Modules define their own "global" namespace, within which classes, functions 50 and lambda definitions establish a hierarchy of namespaces. 51 52 Only local, global and built-in namespaces are recognised; closures are not 53 supported. 54 55 Assignments 56 ----------- 57 58 Name assignment and attribute assignment involving modules and classes cause 59 names to be associated with values within namespaces. 60 61 Any assignments within loops are considered to cause the targets of such 62 assignments to provide non-constant values. 63 64 Assignments to names are only really considered to cause the targets of such 65 assignments to provide constant values if the targets reside in the module 66 namespace or in class namespaces, subject to the above conditions. 67 68 Assignments to names within functions are not generally considered to cause the 69 targets of such assignments to provide constant values since functions can be 70 invoked many times with different inputs. However, there may be benefits in 71 considering a local to be constant within a single invocation. 72 """ 73 74 from micropython.common import * 75 from micropython.data import * 76 import compiler.ast 77 from compiler.visitor import ASTVisitor 78 79 # Program visitors. 80 81 class InspectedModule(ASTVisitor, Module): 82 83 """ 84 An inspected module, providing core details via the Module superclass, but 85 capable of being used as an AST visitor. 86 """ 87 88 def __init__(self, name, importer): 89 90 """ 91 Initialise this visitor with a module 'name' and an 'importer' which is 92 used to provide access to other modules when required. 93 """ 94 95 ASTVisitor.__init__(self) 96 Module.__init__(self, name) 97 self.visitor = self 98 99 # Import machinery links. 100 101 self.importer = importer 102 self.optimisations = importer.optimisations 103 self.builtins = self.importer.modules.get("__builtins__") 104 self.loaded = 0 105 106 # Current expression state. 107 108 self.expr = None 109 110 # Namespace state. 111 112 self.in_init = 0 # Find instance attributes in __init__ methods. 113 self.in_method = 0 # Find instance attributes in all methods. 114 self.in_loop = 0 # Note loop "membership", affecting assignments. 115 self.namespaces = [] 116 self.module = None 117 118 def parse(self, filename): 119 120 "Parse the file having the given 'filename'." 121 122 module = compiler.parseFile(filename) 123 self.process(module) 124 125 def process(self, module): 126 127 "Process the given 'module'." 128 129 self.astnode = self.module = module 130 processed = self.dispatch(module) 131 if self.has_key("__all__"): 132 all = self["__all__"] 133 if isinstance(all, compiler.ast.List): 134 for n in all.nodes: 135 self.store(n.value, self.importer.add_module(self.name + "." + n.value)) 136 return processed 137 138 def vacuum(self): 139 140 """ 141 Vacuum the module namespace, removing unreferenced objects and unused 142 names. 143 """ 144 145 for name, value in self.items(): 146 147 if self.should_optimise_unused_objects(): 148 149 # Remove entries for unreferenced objects. 150 # This, due to the nature of the referenced attribute, assumes 151 # that only explicitly mentioned classes and functions are 152 # employed in the final program. 153 154 if isinstance(value, Attr): 155 156 # Only remove entries for classes and functions, not methods. 157 158 for attr_value in value.get_values(): 159 if (isinstance(attr_value, Function) and not attr_value.is_method() or 160 isinstance(attr_value, Class)) and not attr_value.referenced: 161 pass 162 else: 163 break 164 else: 165 del self[name] 166 167 # Complain about globals not initialised at the module level. 168 169 if isinstance(value, Global): 170 print "Warning: global %r in module %r not initialised at the module level." % (name, self.name) 171 172 # Remove unreferenced objects. 173 174 if self.should_optimise_unused_objects(): 175 176 all_objects = list(self.all_objects) 177 178 for obj in all_objects: 179 180 # Only remove entries for classes and functions, not methods. 181 182 if (isinstance(obj, Function) and not obj.is_method() or 183 isinstance(obj, Class)) and not obj.referenced: 184 185 self.all_objects.remove(obj) 186 187 # Remove unused entries from classes plus associated methods. 188 189 if isinstance(obj, Class): 190 for name, attr in obj.class_attributes().items(): 191 192 # Methods can only be deleted if they are the only 193 # assigned object to the class and are unreferenced. 194 195 if name not in self.importer.names_used and \ 196 attr.assignments == 1 and isinstance(attr.get_value(), Function) and \ 197 attr.get_value().is_method() and not attr.get_value().referenced: 198 199 self.all_objects.remove(attr.get_value()) 200 del obj[name] 201 202 def finalise(self): 203 204 "Finalise the module." 205 206 for obj in self.all_objects: 207 if isinstance(obj, (Class, Function)): 208 obj.finalise_attributes() 209 210 def add_object(self, obj, any_scope=0): 211 212 """ 213 Record 'obj' if non-local or if the optional 'any_scope' is set to a 214 true value. 215 """ 216 217 if any_scope or not (self.namespaces and isinstance(self.namespaces[-1], Function)): 218 self.all_objects.add(obj) 219 220 # Optimisation tests. 221 222 def should_optimise_unused_objects(self): 223 return "unused_objects" in self.optimisations 224 225 # Namespace methods. 226 227 def store(self, name, obj): 228 229 "Record attribute or local 'name', storing 'obj'." 230 231 if not self.namespaces: 232 self.set(name, obj, not self.in_loop) 233 else: 234 self.namespaces[-1].set(name, obj, not self.in_loop) 235 236 def store_lambda(self, obj): 237 238 "Store a lambda function 'obj'." 239 240 self.add_object(obj) 241 242 def store_module_attr(self, name, module): 243 244 """ 245 Record module attribute 'name' in the given 'module' using the current 246 expression. 247 """ 248 249 module.set(name, self.expr, 0) 250 251 def store_class_attr(self, name): 252 253 """ 254 Record class attribute 'name' in the current class using the current 255 expression. 256 """ 257 258 if self.in_method and self.namespaces[-2].has_key(name): 259 self.namespaces[-2].set(name, self.expr, 0) 260 return 1 261 262 return 0 263 264 def store_instance_attr(self, name): 265 266 "Record instance attribute 'name' in the current class." 267 268 if self.in_method: 269 270 # Current namespace is the function. 271 # Previous namespace is the class. 272 273 self.namespaces[-2].add_instance_attribute(name) 274 275 def get_parent(self): 276 277 "Return the parent (or most recent) namespace currently exposed." 278 279 return (self.namespaces[-1:] or [self])[0] 280 281 # Visitor methods. 282 283 def default(self, node, *args): 284 raise InspectError(self.full_name(), node, "Node class %r is not supported." % node.__class__) 285 286 def dispatch(self, node, *args): 287 return ASTVisitor.dispatch(self, node, *args) 288 289 def NOP(self, node): 290 for n in node.getChildNodes(): 291 self.dispatch(n) 292 return None 293 294 def OP(self, node): 295 for n in node.getChildNodes(): 296 self.dispatch(n) 297 return Instance() 298 299 def _visitUnary(self, node): 300 301 "Accounting method for the unary operator 'node'." 302 303 method = unary_methods[node.__class__.__name__] 304 self.importer.use_name(method) 305 return self.OP(node) 306 307 def _visitBinary(self, node): 308 309 "Accounting method for the binary operator 'node'." 310 311 left_method, right_method = binary_methods[node.__class__.__name__] 312 self.importer.use_name(left_method) 313 self.importer.use_name(right_method) 314 return self.OP(node) 315 316 def _visitFunction(self, node, name): 317 318 """ 319 Return a function object for the function defined by 'node' with the 320 given 'name'. If a lambda expression is being visited, 'name' should be 321 None. 322 """ 323 324 function = Function( 325 name, 326 self.get_parent(), 327 node.argnames, 328 node.defaults, 329 (node.flags & 4 != 0), 330 (node.flags & 8 != 0), 331 self, 332 node 333 ) 334 335 # Make a back reference from the node for code generation. 336 337 node.unit = function 338 339 # Process the defaults. 340 341 for n in node.defaults: 342 self.expr = self.dispatch(n) 343 function.store_default(self.expr) 344 345 # Enter the function. 346 347 self.namespaces.append(function) 348 349 # Current namespace is the function. 350 # Previous namespace is the class. 351 352 if len(self.namespaces) > 1 and isinstance(self.namespaces[-2], Class): 353 if name == "__init__": 354 self.in_init = 1 355 self.in_method = 1 356 357 self.dispatch(node.code) 358 self.in_init = 0 359 self.in_method = 0 360 self.namespaces.pop() 361 362 if name is not None: 363 self.store(name, function) 364 else: 365 self.store_lambda(function) 366 367 # Lambda functions are always assumed to be referenced. This is 368 # because other means of discovering the referencing of objects rely 369 # on the original names inherent in the definition of those objects. 370 371 function.set_referenced() 372 373 # Where defaults exist, an instance needs creating. Thus, it makes 374 # no sense to return a reference to the function here, since the 375 # recipient will not be referencing the function itself. 376 377 if node.defaults: 378 return Instance() # indicates no known target 379 380 self.add_object(function, any_scope=1) 381 return function 382 383 # Specific handler methods. 384 385 visitAdd = _visitBinary 386 387 visitAnd = OP 388 389 visitAssert = NOP 390 391 def visitAssign(self, node): 392 self.expr = self.dispatch(node.expr) 393 for n in node.nodes: 394 self.dispatch(n) 395 return None 396 397 def visitAssAttr(self, node): 398 expr = self.dispatch(node.expr) 399 if isinstance(expr, Attr): 400 if expr.name == "self": 401 if not self.store_class_attr(node.attrname): 402 self.store_instance_attr(node.attrname) 403 elif isinstance(expr.get_value(), Module): 404 self.store_module_attr(node.attrname, expr.get_value()) 405 print "Warning: attribute %r of module %r set outside the module." % (node.attrname, expr.get_value().name) 406 return None 407 408 def visitAssList(self, node): 409 for i, n in enumerate(node.nodes): 410 self.dispatch(n) 411 self.importer.make_constant(i) # for __getitem__(i) at run-time 412 return None 413 414 def visitAssName(self, node): 415 self.store(node.name, self.expr) 416 return None 417 418 visitAssTuple = visitAssList 419 420 def visitAugAssign(self, node): 421 422 # Accounting. 423 424 aug_method, (left_method, right_method) = augassign_methods[node.op] 425 self.importer.use_name(aug_method) 426 self.importer.use_name(left_method) 427 self.importer.use_name(right_method) 428 return self.OP(node) 429 430 visitBackquote = OP 431 432 visitBitand = _visitBinary 433 434 visitBitor = _visitBinary 435 436 visitBitxor = _visitBinary 437 438 visitBreak = NOP 439 440 visitCallFunc = OP 441 442 def visitClass(self, node): 443 444 """ 445 Register the class at the given 'node' subject to the restrictions 446 mentioned in the module docstring. 447 """ 448 449 if self.namespaces: 450 print "Class %r in %r is not global: ignored." % (node.name, self.namespaces[-1].full_name()) 451 return None 452 else: 453 cls = Class(node.name, self.get_parent(), self, node) 454 455 # Visit the base class expressions, attempting to find concrete 456 # definitions of classes. 457 458 for base in node.bases: 459 expr = self.dispatch(base) 460 if isinstance(expr, Attr): 461 if expr.assignments != 1: 462 raise InspectError(self.full_name(), node, 463 "Base class %r for %r is not constant." % (base, cls.full_name())) 464 else: 465 cls.add_base(expr.get_value()) 466 else: # if expr is None: 467 raise InspectError(self.full_name(), node, 468 "Base class %r for %r is not found: it may be hidden in some way." % (base, cls.full_name())) 469 470 # NOTE: Potentially dubious measure to permit __init__ availability. 471 # If no bases exist, adopt the 'object' class. 472 473 if not node.bases and not (self.name == "__builtins__" and node.name == "object") : 474 expr = self.dispatch(compiler.ast.Name("object")) 475 cls.add_base(expr.get_value()) 476 477 # Make a back reference from the node for code generation. 478 479 node.unit = cls 480 481 # Make an entry for the class. 482 483 self.store(node.name, cls) 484 self.add_object(cls) 485 486 # Process the class body. 487 488 self.namespaces.append(cls) 489 self.dispatch(node.code) 490 self.namespaces.pop() 491 492 return cls 493 494 def visitCompare(self, node): 495 496 # Accounting. 497 # NOTE: Replicates some code in micropython.ast.visitCompare. 498 499 for op in node.ops: 500 op_name, next_node = op 501 methods = comparison_methods[op_name] 502 if methods is not None: 503 self.importer.use_name(methods[0]) 504 self.importer.use_name(methods[1]) 505 elif op_name.endswith("in"): 506 self.importer.use_name("__contains__") 507 508 return self.OP(node) 509 510 def visitConst(self, node): 511 512 # Accounting. 513 514 name = node.value.__class__.__name__ 515 if self.builtins is not None and self.builtins.has_key(name): 516 attr = self.builtins[name] 517 attr.set_referenced() 518 519 # Register the constant, if necessary, returning the resulting object. 520 521 return self.importer.make_constant(node.value) 522 523 visitContinue = NOP 524 525 visitDecorators = NOP 526 527 visitDict = OP 528 529 visitDiscard = NOP 530 531 visitDiv = _visitBinary 532 533 visitEllipsis = NOP 534 535 visitExec = NOP 536 537 visitExpression = OP 538 539 visitFloorDiv = _visitBinary 540 541 def visitFor(self, node): 542 self.in_loop = 1 543 self.NOP(node) 544 self.in_loop = 0 545 return None 546 547 def visitFrom(self, node): 548 module = self.importer.load(node.modname, 1) 549 550 if module is not None: 551 module.set_referenced() 552 553 #if module is None: 554 # print "Warning:", node.modname, "not imported." 555 556 for name, alias in node.names: 557 if name != "*": 558 if module is not None and module.namespace.has_key(name): 559 attr = module[name] 560 self.store(alias or name, attr) 561 if isinstance(attr.get_value(), Module) and not attr.get_value().loaded: 562 self.importer.load(attr.get_value().name) 563 564 # Support the import of names from missing modules. 565 566 else: 567 self.store(alias or name, UnresolvedName(name, node.modname, self)) 568 else: 569 if module is not None: 570 for n in module.namespace.keys(): 571 attr = module[n] 572 self.store(n, attr) 573 if isinstance(attr.get_value(), Module) and not attr.get_value().loaded: 574 self.importer.load(attr.get_value().name) 575 576 return None 577 578 def visitFunction(self, node): 579 return self._visitFunction(node, node.name) 580 581 visitGenExpr = OP 582 583 visitGenExprFor = NOP 584 585 visitGenExprIf = NOP 586 587 visitGenExprInner = NOP 588 589 def visitGetattr(self, node): 590 expr = self.dispatch(node.expr) 591 attrname = node.attrname 592 593 if isinstance(expr, Attr): 594 value = expr.get_value() 595 if isinstance(value, (Class, Module)): 596 attr = value.namespace.get(attrname) 597 elif isinstance(value, UnresolvedName): 598 attr = UnresolvedName(attrname, value.full_name(), self) 599 else: 600 attr = None 601 elif self.builtins is not None: 602 attr = self.builtins.get(attrname) 603 else: 604 attr = UnresolvedName(attrname, value.full_name(), self) 605 606 # Accounting. 607 608 if attr is not None: 609 attr.set_referenced() 610 611 self.importer.use_name(attrname) 612 613 return attr 614 615 def visitGlobal(self, node): 616 if self.namespaces: 617 for name in node.names: 618 ns = self.namespaces[-1] 619 if not ns.make_global(name): 620 raise InspectError(ns.full_name(), node, "Name %r is global and local in %r" % (name, ns.full_name())) 621 622 # Record a global entry for the name in the module. 623 624 if not self.has_key(name): 625 self[name] = Global() 626 627 def visitIf(self, node): 628 for test, body in node.tests: 629 self.dispatch(test) 630 self.dispatch(body) 631 if node.else_ is not None: 632 self.dispatch(node.else_) 633 return None 634 635 visitIfExp = NOP 636 637 def visitImport(self, node): 638 for name, alias in node.names: 639 if alias is not None: 640 module = self.importer.load(name, 1) or UnresolvedName(None, name, self) 641 self.store(alias, module) 642 else: 643 module = self.importer.load(name) or UnresolvedName(None, name.split(".")[0], self) 644 self.store(name.split(".")[0], module) 645 module.set_referenced() 646 647 return None 648 649 visitInvert = _visitUnary 650 651 def visitKeyword(self, node): 652 self.dispatch(node.expr) 653 self.importer.make_constant(node.name) 654 self.keyword_names.add(node.name) 655 return None 656 657 def visitLambda(self, node): 658 return self._visitFunction(node, None) 659 660 visitLeftShift = _visitBinary 661 662 visitList = OP 663 664 visitListComp = OP 665 666 visitListCompFor = NOP 667 668 visitListCompIf = NOP 669 670 visitMod = _visitBinary 671 672 def visitModule(self, node): 673 674 # Make a back reference from the node for code generation. 675 676 node.unit = self 677 return self.dispatch(node.node) 678 679 visitMul = _visitBinary 680 681 def visitName(self, node): 682 name = node.name 683 684 if self.importer.predefined_constants.has_key(name): 685 attr = self.importer.get_predefined_constant(name) 686 elif self.namespaces and self.namespaces[-1].has_key(name): 687 attr = self.namespaces[-1][name] 688 elif self.has_key(name): 689 attr = self[name] 690 elif self.builtins is not None and self.builtins.has_key(name): 691 attr = self.builtins[name] 692 else: 693 attr = None 694 695 # Accounting. 696 697 if attr is not None: 698 attr.set_referenced() 699 700 self.importer.use_name(name) 701 702 return attr 703 704 visitNot = OP 705 706 visitOr = OP 707 708 visitPass = NOP 709 710 visitPower = _visitBinary 711 712 visitPrint = NOP 713 714 visitPrintnl = NOP 715 716 visitRaise = NOP 717 718 visitReturn = NOP 719 720 visitRightShift = _visitBinary 721 722 visitSlice = OP 723 724 visitSliceobj = OP 725 726 def visitStmt(self, node): 727 for n in node.nodes: 728 self.dispatch(n) 729 return None 730 731 visitSub = _visitBinary 732 733 visitSubscript = OP 734 735 def visitTryExcept(self, node): 736 self.dispatch(node.body) 737 for name, var, n in node.handlers: 738 739 # Establish the local for the handler. 740 741 if var is not None: 742 self.dispatch(var) 743 if n is not None: 744 self.dispatch(n) 745 if node.else_ is not None: 746 self.dispatch(node.else_) 747 return None 748 749 visitTryFinally = NOP 750 751 visitTuple = OP 752 753 visitUnaryAdd = _visitUnary 754 755 visitUnarySub = _visitUnary 756 757 def visitWhile(self, node): 758 self.in_loop = 1 759 self.NOP(node) 760 self.in_loop = 0 761 return None 762 763 visitWith = NOP 764 765 visitYield = NOP 766 767 class Global: 768 769 """ 770 A reference to an object assigned to a global from outside the module 771 top-level. 772 """ 773 774 pass 775 776 # vim: tabstop=4 expandtab shiftwidth=4