1 #!/usr/bin/env python 2 3 """ 4 Encoder functions, producing representations of program objects. 5 6 Copyright (C) 2016, 2017 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 from common import first, InstructionSequence 23 24 # Output encoding and decoding for the summary files. 25 26 def encode_attrnames(attrnames): 27 28 "Encode the 'attrnames' representing usage." 29 30 return ", ".join(attrnames) or "{}" 31 32 def encode_constrained(constrained): 33 34 "Encode the 'constrained' status for program summaries." 35 36 return constrained and "constrained" or "deduced" 37 38 def encode_usage(usage): 39 40 "Encode attribute details from 'usage'." 41 42 all_attrnames = [] 43 for t in usage: 44 attrname, invocation, assignment = t 45 all_attrnames.append("%s%s" % (attrname, invocation and "!" or assignment and "=" or "")) 46 return ", ".join(all_attrnames) or "{}" 47 48 def decode_usage(s): 49 50 "Decode attribute details from 's'." 51 52 all_attrnames = set() 53 for attrname_str in s.split(", "): 54 all_attrnames.add((attrname_str.rstrip("!="), attrname_str.endswith("!"), attrname_str.endswith("="))) 55 56 all_attrnames = list(all_attrnames) 57 all_attrnames.sort() 58 return tuple(all_attrnames) 59 60 def encode_access_location(t): 61 62 "Encode the access location 't'." 63 64 path, name, attrname, version = t 65 return "%s %s %s:%d" % (path, name or "{}", attrname, version) 66 67 def encode_location(t): 68 69 "Encode the general location 't' in a concise form." 70 71 path, name, attrname, version = t 72 if name is not None and version is not None: 73 return "%s %s:%d" % (path, name, version) 74 elif name is not None: 75 return "%s %s" % (path, name) 76 else: 77 return "%s :%s" % (path, attrname) 78 79 def encode_modifiers(modifiers): 80 81 "Encode assignment and invocation details from 'modifiers'." 82 83 all_modifiers = [] 84 for t in modifiers: 85 all_modifiers.append(encode_modifier_term(t)) 86 return "".join(all_modifiers) 87 88 def encode_modifier_term(t): 89 90 "Encode modifier 't' representing an assignment or an invocation." 91 92 assignment, invocation = t 93 if assignment: 94 return "=" 95 elif invocation is not None: 96 arguments, keywords = invocation 97 return "(%d;%s)" % (arguments, ",".join(keywords)) 98 else: 99 return "_" 100 101 def decode_modifiers(s): 102 103 "Decode 's' containing modifiers." 104 105 i = 0 106 end = len(s) 107 108 modifiers = [] 109 110 while i < end: 111 if s[i] == "=": 112 modifiers.append((True, None)) 113 i += 1 114 elif s[i] == "(": 115 j = s.index(";", i) 116 arguments = int(s[i+1:j]) 117 i = j 118 j = s.index(")", i) 119 keywords = s[i+1:j] 120 keywords = keywords and keywords.split(",") or [] 121 modifiers.append((False, (arguments, keywords))) 122 i = j + 1 123 else: 124 modifiers.append((False, None)) 125 i += 1 126 127 return modifiers 128 129 130 131 # Test generation functions. 132 133 def get_kinds(all_types): 134 135 """ 136 Return object kind details for 'all_types', being a collection of 137 references for program types. 138 """ 139 140 return map(lambda ref: ref.get_kind(), all_types) 141 142 def test_label_for_kind(kind): 143 144 "Return the label used for 'kind' in test details." 145 146 return kind == "<instance>" and "instance" or "type" 147 148 def test_label_for_type(ref): 149 150 "Return the label used for 'ref' in test details." 151 152 return test_label_for_kind(ref.get_kind()) 153 154 155 156 # Instruction representation encoding. 157 158 def encode_instruction(instruction): 159 160 """ 161 Encode the 'instruction' - a sequence starting with an operation and 162 followed by arguments, each of which may be an instruction sequence or a 163 plain value - to produce a function call string representation. 164 """ 165 166 op = instruction[0] 167 args = instruction[1:] 168 169 if args: 170 a = [] 171 for arg in args: 172 if isinstance(arg, tuple): 173 a.append(encode_instruction(arg)) 174 else: 175 a.append(arg or "{}") 176 argstr = "(%s)" % ", ".join(a) 177 return "%s%s" % (op, argstr) 178 else: 179 return op 180 181 182 183 # Output program encoding. 184 185 attribute_loading_ops = ( 186 "__load_via_class", "__load_via_object", "__get_class_and_load", 187 ) 188 189 attribute_ops = attribute_loading_ops + ( 190 "__store_via_object", 191 ) 192 193 checked_loading_ops = ( 194 "__check_and_load_via_class", "__check_and_load_via_object", "__check_and_load_via_any", 195 ) 196 197 checked_ops = checked_loading_ops + ( 198 "__check_and_store_via_class", "__check_and_store_via_object", "__check_and_store_via_any", 199 ) 200 201 typename_ops = ( 202 "__test_common_instance", "__test_common_object", "__test_common_type", 203 ) 204 205 type_ops = ( 206 "__test_specific_instance", "__test_specific_object", "__test_specific_type", 207 ) 208 209 static_ops = ( 210 "__load_static", 211 ) 212 213 reference_acting_ops = attribute_ops + checked_ops + typename_ops 214 attribute_producing_ops = attribute_loading_ops + checked_loading_ops 215 216 def encode_access_instruction(instruction, subs): 217 218 """ 219 Encode the 'instruction' - a sequence starting with an operation and 220 followed by arguments, each of which may be an instruction sequence or a 221 plain value - to produce a function call string representation. 222 223 The 'subs' parameter defines a mapping of substitutions for special values 224 used in instructions. 225 226 Return both the encoded instruction and a collection of substituted names. 227 """ 228 229 op = instruction[0] 230 args = instruction[1:] 231 substituted = set() 232 233 if not args: 234 argstr = "" 235 236 else: 237 # Encode the arguments. 238 239 a = [] 240 converting_op = op 241 for arg in args: 242 s, _substituted = encode_access_instruction_arg(arg, subs, converting_op) 243 substituted.update(_substituted) 244 a.append(s) 245 converting_op = None 246 247 # Modify certain arguments. 248 249 # Convert attribute name arguments to position symbols. 250 251 if op in attribute_ops: 252 arg = a[1] 253 a[1] = encode_symbol("pos", arg) 254 255 # Convert attribute name arguments to position and code symbols. 256 257 elif op in checked_ops: 258 arg = a[1] 259 a[1] = encode_symbol("pos", arg) 260 a.insert(2, encode_symbol("code", arg)) 261 262 # Convert type name arguments to position and code symbols. 263 264 elif op in typename_ops: 265 arg = encode_type_attribute(args[1]) 266 a[1] = encode_symbol("pos", arg) 267 a.insert(2, encode_symbol("code", arg)) 268 269 # Obtain addresses of type arguments. 270 271 elif op in type_ops: 272 a[1] = "&%s" % a[1] 273 274 # Obtain addresses of static objects. 275 276 elif op in static_ops: 277 a[0] = "&%s" % a[0] 278 a[1] = "&%s" % a[1] 279 280 argstr = "(%s)" % ", ".join(map(str, a)) 281 282 # Substitute the first element of the instruction, which may not be an 283 # operation at all. 284 285 if subs.has_key(op): 286 substituted.add(op) 287 288 # Break accessor initialisation into initialisation and value-yielding 289 # parts: 290 291 if op == "<set_accessor>" and isinstance(a[0], InstructionSequence): 292 ops = [] 293 ops += a[0].get_init_instructions() 294 ops.append("%s(%s)" % (subs[op], a[0].get_value_instruction())) 295 return ", ".join(map(str, ops)), substituted 296 297 op = subs[op] 298 299 elif not args: 300 op = "&%s" % encode_path(op) 301 302 return "%s%s" % (op, argstr), substituted 303 304 def encode_access_instruction_arg(arg, subs, op): 305 306 """ 307 Encode 'arg' using 'subs' to define substitutions, returning a tuple 308 containing the encoded form of 'arg' along with a collection of any 309 substituted values. 310 """ 311 312 if isinstance(arg, tuple): 313 encoded, substituted = encode_access_instruction(arg, subs) 314 315 # Convert attribute results to references where required. 316 317 if op and op in reference_acting_ops and arg[0] in attribute_producing_ops: 318 return "%s.value" % encoded, substituted 319 else: 320 return encoded, substituted 321 322 # Special values only need replacing, not encoding. 323 324 elif subs.has_key(arg): 325 return subs.get(arg), set([arg]) 326 327 # Convert static references to the appropriate type. 328 329 elif op and op in reference_acting_ops and arg != "<accessor>": 330 return "&%s" % encode_path(arg), set() 331 332 # Other values may need encoding. 333 334 else: 335 return encode_path(arg), set() 336 337 def encode_function_pointer(path): 338 339 "Encode 'path' as a reference to an output program function." 340 341 return "__fn_%s" % encode_path(path) 342 343 def encode_instantiator_pointer(path): 344 345 "Encode 'path' as a reference to an output program instantiator." 346 347 return "__new_%s" % encode_path(path) 348 349 def encode_instructions(instructions): 350 351 "Encode 'instructions' as a sequence." 352 353 if len(instructions) == 1: 354 return instructions[0] 355 else: 356 return "(\n%s\n)" % ",\n".join(instructions) 357 358 def encode_literal_constant(n): 359 360 "Encode a name for the literal constant with the number 'n'." 361 362 return "__const%d" % n 363 364 def encode_literal_constant_size(value): 365 366 "Encode a size for the literal constant with the given 'value'." 367 368 if isinstance(value, basestring): 369 return len(value) 370 else: 371 return 0 372 373 def encode_literal_constant_member(value): 374 375 "Encode the member name for the 'value' in the final program." 376 377 return "%svalue" % value.__class__.__name__ 378 379 def encode_literal_constant_value(value): 380 381 "Encode the given 'value' in the final program." 382 383 if isinstance(value, (int, float)): 384 return str(value) 385 else: 386 l = [] 387 388 # Encode characters including non-ASCII ones. 389 390 for c in str(value): 391 if c == '"': l.append('\\"') 392 elif c == '\n': l.append('\\n') 393 elif c == '\t': l.append('\\t') 394 elif c == '\r': l.append('\\r') 395 elif c == '\\': l.append('\\\\') 396 elif 0x20 <= ord(c) < 0x80: l.append(c) 397 else: l.append("\\x%02x" % ord(c)) 398 399 return '"%s"' % "".join(l) 400 401 def encode_literal_data_initialiser(style): 402 403 """ 404 Encode a reference to a function populating the data for a literal having 405 the given 'style' ("mapping" or "sequence"). 406 """ 407 408 return "__newdata_%s" % style 409 410 def encode_literal_instantiator(path): 411 412 """ 413 Encode a reference to an instantiator for a literal having the given 'path'. 414 """ 415 416 return "__newliteral_%s" % encode_path(path) 417 418 def encode_literal_reference(n): 419 420 "Encode a reference to a literal constant with the number 'n'." 421 422 return "__constvalue%d" % n 423 424 425 426 # Track all encoded paths, detecting and avoiding conflicts. 427 428 all_encoded_paths = {} 429 430 def encode_path(path): 431 432 "Encode 'path' as an output program object, translating special symbols." 433 434 if path in reserved_words: 435 return "__%s" % path 436 else: 437 part_encoded = path.replace("#", "__").replace("$", "__") 438 439 if "." not in path: 440 return part_encoded 441 442 encoded = part_encoded.replace(".", "_") 443 444 # Test for a conflict with the encoding of a different path, re-encoding 445 # if necessary. 446 447 previous = all_encoded_paths.get(encoded) 448 replacement = "_" 449 450 while previous: 451 if path == previous: 452 return encoded 453 replacement += "_" 454 encoded = part_encoded.replace(".", replacement) 455 previous = all_encoded_paths.get(encoded) 456 457 # Store any new or re-encoded path. 458 459 all_encoded_paths[encoded] = path 460 return encoded 461 462 def encode_predefined_reference(path): 463 464 "Encode a reference to a predefined constant value for 'path'." 465 466 return "__predefined_%s" % encode_path(path) 467 468 def encode_size(kind, path=None): 469 470 """ 471 Encode a structure size reference for the given 'kind' of structure, with 472 'path' indicating a specific structure name. 473 """ 474 475 return "__%ssize%s" % (structure_size_prefixes.get(kind, kind), path and "_%s" % encode_path(path) or "") 476 477 def encode_symbol(symbol_type, path=None): 478 479 "Encode a symbol with the given 'symbol_type' and optional 'path'." 480 481 return "__%s%s" % (symbol_type, path and "_%s" % encode_path(path) or "") 482 483 def encode_tablename(kind, path): 484 485 """ 486 Encode a table reference for the given 'kind' of table structure, indicating 487 a 'path' for the specific object concerned. 488 """ 489 490 return "__%sTable_%s" % (table_name_prefixes[kind], encode_path(path)) 491 492 def encode_type_attribute(path): 493 494 "Encode the special type attribute for 'path'." 495 496 return "#%s" % path 497 498 def decode_type_attribute(s): 499 500 "Decode the special type attribute 's'." 501 502 return s[1:] 503 504 def is_type_attribute(s): 505 506 "Return whether 's' is a type attribute name." 507 508 return s.startswith("#") 509 510 511 512 # A mapping from kinds to structure size reference prefixes. 513 514 structure_size_prefixes = { 515 "<class>" : "c", 516 "<module>" : "m", 517 "<instance>" : "i" 518 } 519 520 # A mapping from kinds to table name prefixes. 521 522 table_name_prefixes = { 523 "<class>" : "Class", 524 "<function>" : "Function", 525 "<module>" : "Module", 526 "<instance>" : "Instance" 527 } 528 529 530 531 # Output language reserved words. 532 533 reserved_words = [ 534 "break", "char", "const", "continue", 535 "default", "double", "else", 536 "float", "for", 537 "if", "int", "long", 538 "NULL", 539 "return", "struct", 540 "typedef", 541 "void", "while", 542 ] 543 544 # vim: tabstop=4 expandtab shiftwidth=4