Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1 | // Copyright 2014 the V8 project authors. All rights reserved. |
| 2 | // Use of this source code is governed by a BSD-style license that can be |
| 3 | // found in the LICENSE file. |
| 4 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 5 | #include "src/runtime/runtime-utils.h" |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 6 | |
| 7 | #include "src/arguments.h" |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 8 | #include "src/conversions-inl.h" |
| 9 | #include "src/isolate-inl.h" |
| 10 | #include "src/regexp/jsregexp-inl.h" |
| 11 | #include "src/regexp/jsregexp.h" |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 12 | #include "src/string-builder.h" |
| 13 | #include "src/string-search.h" |
| 14 | |
| 15 | namespace v8 { |
| 16 | namespace internal { |
| 17 | |
| 18 | |
| 19 | // Perform string match of pattern on subject, starting at start index. |
| 20 | // Caller must ensure that 0 <= start_index <= sub->length(), |
| 21 | // and should check that pat->length() + start_index <= sub->length(). |
| 22 | int StringMatch(Isolate* isolate, Handle<String> sub, Handle<String> pat, |
| 23 | int start_index) { |
| 24 | DCHECK(0 <= start_index); |
| 25 | DCHECK(start_index <= sub->length()); |
| 26 | |
| 27 | int pattern_length = pat->length(); |
| 28 | if (pattern_length == 0) return start_index; |
| 29 | |
| 30 | int subject_length = sub->length(); |
| 31 | if (start_index + pattern_length > subject_length) return -1; |
| 32 | |
| 33 | sub = String::Flatten(sub); |
| 34 | pat = String::Flatten(pat); |
| 35 | |
| 36 | DisallowHeapAllocation no_gc; // ensure vectors stay valid |
| 37 | // Extract flattened substrings of cons strings before getting encoding. |
| 38 | String::FlatContent seq_sub = sub->GetFlatContent(); |
| 39 | String::FlatContent seq_pat = pat->GetFlatContent(); |
| 40 | |
| 41 | // dispatch on type of strings |
| 42 | if (seq_pat.IsOneByte()) { |
| 43 | Vector<const uint8_t> pat_vector = seq_pat.ToOneByteVector(); |
| 44 | if (seq_sub.IsOneByte()) { |
| 45 | return SearchString(isolate, seq_sub.ToOneByteVector(), pat_vector, |
| 46 | start_index); |
| 47 | } |
| 48 | return SearchString(isolate, seq_sub.ToUC16Vector(), pat_vector, |
| 49 | start_index); |
| 50 | } |
| 51 | Vector<const uc16> pat_vector = seq_pat.ToUC16Vector(); |
| 52 | if (seq_sub.IsOneByte()) { |
| 53 | return SearchString(isolate, seq_sub.ToOneByteVector(), pat_vector, |
| 54 | start_index); |
| 55 | } |
| 56 | return SearchString(isolate, seq_sub.ToUC16Vector(), pat_vector, start_index); |
| 57 | } |
| 58 | |
| 59 | |
| 60 | // This may return an empty MaybeHandle if an exception is thrown or |
| 61 | // we abort due to reaching the recursion limit. |
| 62 | MaybeHandle<String> StringReplaceOneCharWithString( |
| 63 | Isolate* isolate, Handle<String> subject, Handle<String> search, |
| 64 | Handle<String> replace, bool* found, int recursion_limit) { |
| 65 | StackLimitCheck stackLimitCheck(isolate); |
| 66 | if (stackLimitCheck.HasOverflowed() || (recursion_limit == 0)) { |
| 67 | return MaybeHandle<String>(); |
| 68 | } |
| 69 | recursion_limit--; |
| 70 | if (subject->IsConsString()) { |
| 71 | ConsString* cons = ConsString::cast(*subject); |
| 72 | Handle<String> first = Handle<String>(cons->first()); |
| 73 | Handle<String> second = Handle<String>(cons->second()); |
| 74 | Handle<String> new_first; |
| 75 | if (!StringReplaceOneCharWithString(isolate, first, search, replace, found, |
| 76 | recursion_limit).ToHandle(&new_first)) { |
| 77 | return MaybeHandle<String>(); |
| 78 | } |
| 79 | if (*found) return isolate->factory()->NewConsString(new_first, second); |
| 80 | |
| 81 | Handle<String> new_second; |
| 82 | if (!StringReplaceOneCharWithString(isolate, second, search, replace, found, |
| 83 | recursion_limit) |
| 84 | .ToHandle(&new_second)) { |
| 85 | return MaybeHandle<String>(); |
| 86 | } |
| 87 | if (*found) return isolate->factory()->NewConsString(first, new_second); |
| 88 | |
| 89 | return subject; |
| 90 | } else { |
| 91 | int index = StringMatch(isolate, subject, search, 0); |
| 92 | if (index == -1) return subject; |
| 93 | *found = true; |
| 94 | Handle<String> first = isolate->factory()->NewSubString(subject, 0, index); |
| 95 | Handle<String> cons1; |
| 96 | ASSIGN_RETURN_ON_EXCEPTION( |
| 97 | isolate, cons1, isolate->factory()->NewConsString(first, replace), |
| 98 | String); |
| 99 | Handle<String> second = |
| 100 | isolate->factory()->NewSubString(subject, index + 1, subject->length()); |
| 101 | return isolate->factory()->NewConsString(cons1, second); |
| 102 | } |
| 103 | } |
| 104 | |
| 105 | |
| 106 | RUNTIME_FUNCTION(Runtime_StringReplaceOneCharWithString) { |
| 107 | HandleScope scope(isolate); |
| 108 | DCHECK(args.length() == 3); |
| 109 | CONVERT_ARG_HANDLE_CHECKED(String, subject, 0); |
| 110 | CONVERT_ARG_HANDLE_CHECKED(String, search, 1); |
| 111 | CONVERT_ARG_HANDLE_CHECKED(String, replace, 2); |
| 112 | |
| 113 | // If the cons string tree is too deep, we simply abort the recursion and |
| 114 | // retry with a flattened subject string. |
| 115 | const int kRecursionLimit = 0x1000; |
| 116 | bool found = false; |
| 117 | Handle<String> result; |
| 118 | if (StringReplaceOneCharWithString(isolate, subject, search, replace, &found, |
| 119 | kRecursionLimit).ToHandle(&result)) { |
| 120 | return *result; |
| 121 | } |
| 122 | if (isolate->has_pending_exception()) return isolate->heap()->exception(); |
| 123 | |
| 124 | subject = String::Flatten(subject); |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 125 | if (StringReplaceOneCharWithString(isolate, subject, search, replace, &found, |
| 126 | kRecursionLimit).ToHandle(&result)) { |
| 127 | return *result; |
| 128 | } |
| 129 | if (isolate->has_pending_exception()) return isolate->heap()->exception(); |
| 130 | // In case of empty handle and no pending exception we have stack overflow. |
| 131 | return isolate->StackOverflow(); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 132 | } |
| 133 | |
| 134 | |
| 135 | RUNTIME_FUNCTION(Runtime_StringIndexOf) { |
| 136 | HandleScope scope(isolate); |
| 137 | DCHECK(args.length() == 3); |
| 138 | |
| 139 | CONVERT_ARG_HANDLE_CHECKED(String, sub, 0); |
| 140 | CONVERT_ARG_HANDLE_CHECKED(String, pat, 1); |
| 141 | CONVERT_ARG_HANDLE_CHECKED(Object, index, 2); |
| 142 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 143 | uint32_t start_index = 0; |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 144 | if (!index->ToArrayIndex(&start_index)) return Smi::FromInt(-1); |
| 145 | |
| 146 | RUNTIME_ASSERT(start_index <= static_cast<uint32_t>(sub->length())); |
| 147 | int position = StringMatch(isolate, sub, pat, start_index); |
| 148 | return Smi::FromInt(position); |
| 149 | } |
| 150 | |
| 151 | |
| 152 | template <typename schar, typename pchar> |
| 153 | static int StringMatchBackwards(Vector<const schar> subject, |
| 154 | Vector<const pchar> pattern, int idx) { |
| 155 | int pattern_length = pattern.length(); |
| 156 | DCHECK(pattern_length >= 1); |
| 157 | DCHECK(idx + pattern_length <= subject.length()); |
| 158 | |
| 159 | if (sizeof(schar) == 1 && sizeof(pchar) > 1) { |
| 160 | for (int i = 0; i < pattern_length; i++) { |
| 161 | uc16 c = pattern[i]; |
| 162 | if (c > String::kMaxOneByteCharCode) { |
| 163 | return -1; |
| 164 | } |
| 165 | } |
| 166 | } |
| 167 | |
| 168 | pchar pattern_first_char = pattern[0]; |
| 169 | for (int i = idx; i >= 0; i--) { |
| 170 | if (subject[i] != pattern_first_char) continue; |
| 171 | int j = 1; |
| 172 | while (j < pattern_length) { |
| 173 | if (pattern[j] != subject[i + j]) { |
| 174 | break; |
| 175 | } |
| 176 | j++; |
| 177 | } |
| 178 | if (j == pattern_length) { |
| 179 | return i; |
| 180 | } |
| 181 | } |
| 182 | return -1; |
| 183 | } |
| 184 | |
| 185 | |
| 186 | RUNTIME_FUNCTION(Runtime_StringLastIndexOf) { |
| 187 | HandleScope scope(isolate); |
| 188 | DCHECK(args.length() == 3); |
| 189 | |
| 190 | CONVERT_ARG_HANDLE_CHECKED(String, sub, 0); |
| 191 | CONVERT_ARG_HANDLE_CHECKED(String, pat, 1); |
| 192 | CONVERT_ARG_HANDLE_CHECKED(Object, index, 2); |
| 193 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 194 | uint32_t start_index = 0; |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 195 | if (!index->ToArrayIndex(&start_index)) return Smi::FromInt(-1); |
| 196 | |
| 197 | uint32_t pat_length = pat->length(); |
| 198 | uint32_t sub_length = sub->length(); |
| 199 | |
| 200 | if (start_index + pat_length > sub_length) { |
| 201 | start_index = sub_length - pat_length; |
| 202 | } |
| 203 | |
| 204 | if (pat_length == 0) { |
| 205 | return Smi::FromInt(start_index); |
| 206 | } |
| 207 | |
| 208 | sub = String::Flatten(sub); |
| 209 | pat = String::Flatten(pat); |
| 210 | |
| 211 | int position = -1; |
| 212 | DisallowHeapAllocation no_gc; // ensure vectors stay valid |
| 213 | |
| 214 | String::FlatContent sub_content = sub->GetFlatContent(); |
| 215 | String::FlatContent pat_content = pat->GetFlatContent(); |
| 216 | |
| 217 | if (pat_content.IsOneByte()) { |
| 218 | Vector<const uint8_t> pat_vector = pat_content.ToOneByteVector(); |
| 219 | if (sub_content.IsOneByte()) { |
| 220 | position = StringMatchBackwards(sub_content.ToOneByteVector(), pat_vector, |
| 221 | start_index); |
| 222 | } else { |
| 223 | position = StringMatchBackwards(sub_content.ToUC16Vector(), pat_vector, |
| 224 | start_index); |
| 225 | } |
| 226 | } else { |
| 227 | Vector<const uc16> pat_vector = pat_content.ToUC16Vector(); |
| 228 | if (sub_content.IsOneByte()) { |
| 229 | position = StringMatchBackwards(sub_content.ToOneByteVector(), pat_vector, |
| 230 | start_index); |
| 231 | } else { |
| 232 | position = StringMatchBackwards(sub_content.ToUC16Vector(), pat_vector, |
| 233 | start_index); |
| 234 | } |
| 235 | } |
| 236 | |
| 237 | return Smi::FromInt(position); |
| 238 | } |
| 239 | |
| 240 | |
| 241 | RUNTIME_FUNCTION(Runtime_StringLocaleCompare) { |
| 242 | HandleScope handle_scope(isolate); |
| 243 | DCHECK(args.length() == 2); |
| 244 | |
| 245 | CONVERT_ARG_HANDLE_CHECKED(String, str1, 0); |
| 246 | CONVERT_ARG_HANDLE_CHECKED(String, str2, 1); |
| 247 | |
| 248 | if (str1.is_identical_to(str2)) return Smi::FromInt(0); // Equal. |
| 249 | int str1_length = str1->length(); |
| 250 | int str2_length = str2->length(); |
| 251 | |
| 252 | // Decide trivial cases without flattening. |
| 253 | if (str1_length == 0) { |
| 254 | if (str2_length == 0) return Smi::FromInt(0); // Equal. |
| 255 | return Smi::FromInt(-str2_length); |
| 256 | } else { |
| 257 | if (str2_length == 0) return Smi::FromInt(str1_length); |
| 258 | } |
| 259 | |
| 260 | int end = str1_length < str2_length ? str1_length : str2_length; |
| 261 | |
| 262 | // No need to flatten if we are going to find the answer on the first |
| 263 | // character. At this point we know there is at least one character |
| 264 | // in each string, due to the trivial case handling above. |
| 265 | int d = str1->Get(0) - str2->Get(0); |
| 266 | if (d != 0) return Smi::FromInt(d); |
| 267 | |
| 268 | str1 = String::Flatten(str1); |
| 269 | str2 = String::Flatten(str2); |
| 270 | |
| 271 | DisallowHeapAllocation no_gc; |
| 272 | String::FlatContent flat1 = str1->GetFlatContent(); |
| 273 | String::FlatContent flat2 = str2->GetFlatContent(); |
| 274 | |
| 275 | for (int i = 0; i < end; i++) { |
| 276 | if (flat1.Get(i) != flat2.Get(i)) { |
| 277 | return Smi::FromInt(flat1.Get(i) - flat2.Get(i)); |
| 278 | } |
| 279 | } |
| 280 | |
| 281 | return Smi::FromInt(str1_length - str2_length); |
| 282 | } |
| 283 | |
| 284 | |
| 285 | RUNTIME_FUNCTION(Runtime_SubString) { |
| 286 | HandleScope scope(isolate); |
| 287 | DCHECK(args.length() == 3); |
| 288 | |
| 289 | CONVERT_ARG_HANDLE_CHECKED(String, string, 0); |
| 290 | int start, end; |
| 291 | // We have a fast integer-only case here to avoid a conversion to double in |
| 292 | // the common case where from and to are Smis. |
| 293 | if (args[1]->IsSmi() && args[2]->IsSmi()) { |
| 294 | CONVERT_SMI_ARG_CHECKED(from_number, 1); |
| 295 | CONVERT_SMI_ARG_CHECKED(to_number, 2); |
| 296 | start = from_number; |
| 297 | end = to_number; |
| 298 | } else { |
| 299 | CONVERT_DOUBLE_ARG_CHECKED(from_number, 1); |
| 300 | CONVERT_DOUBLE_ARG_CHECKED(to_number, 2); |
| 301 | start = FastD2IChecked(from_number); |
| 302 | end = FastD2IChecked(to_number); |
| 303 | } |
| 304 | RUNTIME_ASSERT(end >= start); |
| 305 | RUNTIME_ASSERT(start >= 0); |
| 306 | RUNTIME_ASSERT(end <= string->length()); |
| 307 | isolate->counters()->sub_string_runtime()->Increment(); |
| 308 | |
| 309 | return *isolate->factory()->NewSubString(string, start, end); |
| 310 | } |
| 311 | |
| 312 | |
| 313 | RUNTIME_FUNCTION(Runtime_StringAdd) { |
| 314 | HandleScope scope(isolate); |
| 315 | DCHECK(args.length() == 2); |
| 316 | CONVERT_ARG_HANDLE_CHECKED(String, str1, 0); |
| 317 | CONVERT_ARG_HANDLE_CHECKED(String, str2, 1); |
| 318 | isolate->counters()->string_add_runtime()->Increment(); |
| 319 | Handle<String> result; |
| 320 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 321 | isolate, result, isolate->factory()->NewConsString(str1, str2)); |
| 322 | return *result; |
| 323 | } |
| 324 | |
| 325 | |
| 326 | RUNTIME_FUNCTION(Runtime_InternalizeString) { |
| 327 | HandleScope handles(isolate); |
| 328 | RUNTIME_ASSERT(args.length() == 1); |
| 329 | CONVERT_ARG_HANDLE_CHECKED(String, string, 0); |
| 330 | return *isolate->factory()->InternalizeString(string); |
| 331 | } |
| 332 | |
| 333 | |
| 334 | RUNTIME_FUNCTION(Runtime_StringMatch) { |
| 335 | HandleScope handles(isolate); |
| 336 | DCHECK(args.length() == 3); |
| 337 | |
| 338 | CONVERT_ARG_HANDLE_CHECKED(String, subject, 0); |
| 339 | CONVERT_ARG_HANDLE_CHECKED(JSRegExp, regexp, 1); |
| 340 | CONVERT_ARG_HANDLE_CHECKED(JSArray, regexp_info, 2); |
| 341 | |
| 342 | RUNTIME_ASSERT(regexp_info->HasFastObjectElements()); |
| 343 | |
Ben Murdoch | 097c5b2 | 2016-05-18 11:27:45 +0100 | [diff] [blame^] | 344 | RegExpImpl::GlobalCache global_cache(regexp, subject, isolate); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 345 | if (global_cache.HasException()) return isolate->heap()->exception(); |
| 346 | |
| 347 | int capture_count = regexp->CaptureCount(); |
| 348 | |
| 349 | ZoneScope zone_scope(isolate->runtime_zone()); |
| 350 | ZoneList<int> offsets(8, zone_scope.zone()); |
| 351 | |
| 352 | while (true) { |
| 353 | int32_t* match = global_cache.FetchNext(); |
| 354 | if (match == NULL) break; |
| 355 | offsets.Add(match[0], zone_scope.zone()); // start |
| 356 | offsets.Add(match[1], zone_scope.zone()); // end |
| 357 | } |
| 358 | |
| 359 | if (global_cache.HasException()) return isolate->heap()->exception(); |
| 360 | |
| 361 | if (offsets.length() == 0) { |
| 362 | // Not a single match. |
| 363 | return isolate->heap()->null_value(); |
| 364 | } |
| 365 | |
| 366 | RegExpImpl::SetLastMatchInfo(regexp_info, subject, capture_count, |
| 367 | global_cache.LastSuccessfulMatch()); |
| 368 | |
| 369 | int matches = offsets.length() / 2; |
| 370 | Handle<FixedArray> elements = isolate->factory()->NewFixedArray(matches); |
| 371 | Handle<String> substring = |
| 372 | isolate->factory()->NewSubString(subject, offsets.at(0), offsets.at(1)); |
| 373 | elements->set(0, *substring); |
| 374 | for (int i = 1; i < matches; i++) { |
| 375 | HandleScope temp_scope(isolate); |
| 376 | int from = offsets.at(i * 2); |
| 377 | int to = offsets.at(i * 2 + 1); |
| 378 | Handle<String> substring = |
| 379 | isolate->factory()->NewProperSubString(subject, from, to); |
| 380 | elements->set(i, *substring); |
| 381 | } |
| 382 | Handle<JSArray> result = isolate->factory()->NewJSArrayWithElements(elements); |
| 383 | result->set_length(Smi::FromInt(matches)); |
| 384 | return *result; |
| 385 | } |
| 386 | |
| 387 | |
| 388 | RUNTIME_FUNCTION(Runtime_StringCharCodeAtRT) { |
| 389 | HandleScope handle_scope(isolate); |
| 390 | DCHECK(args.length() == 2); |
| 391 | |
| 392 | CONVERT_ARG_HANDLE_CHECKED(String, subject, 0); |
| 393 | CONVERT_NUMBER_CHECKED(uint32_t, i, Uint32, args[1]); |
| 394 | |
| 395 | // Flatten the string. If someone wants to get a char at an index |
| 396 | // in a cons string, it is likely that more indices will be |
| 397 | // accessed. |
| 398 | subject = String::Flatten(subject); |
| 399 | |
| 400 | if (i >= static_cast<uint32_t>(subject->length())) { |
| 401 | return isolate->heap()->nan_value(); |
| 402 | } |
| 403 | |
| 404 | return Smi::FromInt(subject->Get(i)); |
| 405 | } |
| 406 | |
| 407 | |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 408 | RUNTIME_FUNCTION(Runtime_StringCompare) { |
| 409 | HandleScope handle_scope(isolate); |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 410 | DCHECK_EQ(2, args.length()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 411 | CONVERT_ARG_HANDLE_CHECKED(String, x, 0); |
| 412 | CONVERT_ARG_HANDLE_CHECKED(String, y, 1); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 413 | isolate->counters()->string_compare_runtime()->Increment(); |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 414 | switch (String::Compare(x, y)) { |
| 415 | case ComparisonResult::kLessThan: |
| 416 | return Smi::FromInt(LESS); |
| 417 | case ComparisonResult::kEqual: |
| 418 | return Smi::FromInt(EQUAL); |
| 419 | case ComparisonResult::kGreaterThan: |
| 420 | return Smi::FromInt(GREATER); |
| 421 | case ComparisonResult::kUndefined: |
| 422 | break; |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 423 | } |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 424 | UNREACHABLE(); |
| 425 | return Smi::FromInt(0); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 426 | } |
| 427 | |
| 428 | |
| 429 | RUNTIME_FUNCTION(Runtime_StringBuilderConcat) { |
| 430 | HandleScope scope(isolate); |
| 431 | DCHECK(args.length() == 3); |
| 432 | CONVERT_ARG_HANDLE_CHECKED(JSArray, array, 0); |
| 433 | int32_t array_length; |
| 434 | if (!args[1]->ToInt32(&array_length)) { |
| 435 | THROW_NEW_ERROR_RETURN_FAILURE(isolate, NewInvalidStringLengthError()); |
| 436 | } |
| 437 | CONVERT_ARG_HANDLE_CHECKED(String, special, 2); |
| 438 | |
| 439 | size_t actual_array_length = 0; |
| 440 | RUNTIME_ASSERT( |
| 441 | TryNumberToSize(isolate, array->length(), &actual_array_length)); |
| 442 | RUNTIME_ASSERT(array_length >= 0); |
| 443 | RUNTIME_ASSERT(static_cast<size_t>(array_length) <= actual_array_length); |
| 444 | |
| 445 | // This assumption is used by the slice encoding in one or two smis. |
| 446 | DCHECK(Smi::kMaxValue >= String::kMaxLength); |
| 447 | |
| 448 | RUNTIME_ASSERT(array->HasFastElements()); |
| 449 | JSObject::EnsureCanContainHeapObjectElements(array); |
| 450 | |
| 451 | int special_length = special->length(); |
| 452 | if (!array->HasFastObjectElements()) { |
| 453 | return isolate->Throw(isolate->heap()->illegal_argument_string()); |
| 454 | } |
| 455 | |
| 456 | int length; |
| 457 | bool one_byte = special->HasOnlyOneByteChars(); |
| 458 | |
| 459 | { |
| 460 | DisallowHeapAllocation no_gc; |
| 461 | FixedArray* fixed_array = FixedArray::cast(array->elements()); |
| 462 | if (fixed_array->length() < array_length) { |
| 463 | array_length = fixed_array->length(); |
| 464 | } |
| 465 | |
| 466 | if (array_length == 0) { |
| 467 | return isolate->heap()->empty_string(); |
| 468 | } else if (array_length == 1) { |
| 469 | Object* first = fixed_array->get(0); |
| 470 | if (first->IsString()) return first; |
| 471 | } |
| 472 | length = StringBuilderConcatLength(special_length, fixed_array, |
| 473 | array_length, &one_byte); |
| 474 | } |
| 475 | |
| 476 | if (length == -1) { |
| 477 | return isolate->Throw(isolate->heap()->illegal_argument_string()); |
| 478 | } |
| 479 | |
| 480 | if (one_byte) { |
| 481 | Handle<SeqOneByteString> answer; |
| 482 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 483 | isolate, answer, isolate->factory()->NewRawOneByteString(length)); |
| 484 | StringBuilderConcatHelper(*special, answer->GetChars(), |
| 485 | FixedArray::cast(array->elements()), |
| 486 | array_length); |
| 487 | return *answer; |
| 488 | } else { |
| 489 | Handle<SeqTwoByteString> answer; |
| 490 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 491 | isolate, answer, isolate->factory()->NewRawTwoByteString(length)); |
| 492 | StringBuilderConcatHelper(*special, answer->GetChars(), |
| 493 | FixedArray::cast(array->elements()), |
| 494 | array_length); |
| 495 | return *answer; |
| 496 | } |
| 497 | } |
| 498 | |
| 499 | |
| 500 | RUNTIME_FUNCTION(Runtime_StringBuilderJoin) { |
| 501 | HandleScope scope(isolate); |
| 502 | DCHECK(args.length() == 3); |
| 503 | CONVERT_ARG_HANDLE_CHECKED(JSArray, array, 0); |
| 504 | int32_t array_length; |
| 505 | if (!args[1]->ToInt32(&array_length)) { |
| 506 | THROW_NEW_ERROR_RETURN_FAILURE(isolate, NewInvalidStringLengthError()); |
| 507 | } |
| 508 | CONVERT_ARG_HANDLE_CHECKED(String, separator, 2); |
| 509 | RUNTIME_ASSERT(array->HasFastObjectElements()); |
| 510 | RUNTIME_ASSERT(array_length >= 0); |
| 511 | |
| 512 | Handle<FixedArray> fixed_array(FixedArray::cast(array->elements())); |
| 513 | if (fixed_array->length() < array_length) { |
| 514 | array_length = fixed_array->length(); |
| 515 | } |
| 516 | |
| 517 | if (array_length == 0) { |
| 518 | return isolate->heap()->empty_string(); |
| 519 | } else if (array_length == 1) { |
| 520 | Object* first = fixed_array->get(0); |
| 521 | RUNTIME_ASSERT(first->IsString()); |
| 522 | return first; |
| 523 | } |
| 524 | |
| 525 | int separator_length = separator->length(); |
| 526 | RUNTIME_ASSERT(separator_length > 0); |
| 527 | int max_nof_separators = |
| 528 | (String::kMaxLength + separator_length - 1) / separator_length; |
| 529 | if (max_nof_separators < (array_length - 1)) { |
| 530 | THROW_NEW_ERROR_RETURN_FAILURE(isolate, NewInvalidStringLengthError()); |
| 531 | } |
| 532 | int length = (array_length - 1) * separator_length; |
| 533 | for (int i = 0; i < array_length; i++) { |
| 534 | Object* element_obj = fixed_array->get(i); |
| 535 | RUNTIME_ASSERT(element_obj->IsString()); |
| 536 | String* element = String::cast(element_obj); |
| 537 | int increment = element->length(); |
| 538 | if (increment > String::kMaxLength - length) { |
| 539 | STATIC_ASSERT(String::kMaxLength < kMaxInt); |
| 540 | length = kMaxInt; // Provoke exception; |
| 541 | break; |
| 542 | } |
| 543 | length += increment; |
| 544 | } |
| 545 | |
| 546 | Handle<SeqTwoByteString> answer; |
| 547 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 548 | isolate, answer, isolate->factory()->NewRawTwoByteString(length)); |
| 549 | |
| 550 | DisallowHeapAllocation no_gc; |
| 551 | |
| 552 | uc16* sink = answer->GetChars(); |
| 553 | #ifdef DEBUG |
| 554 | uc16* end = sink + length; |
| 555 | #endif |
| 556 | |
| 557 | RUNTIME_ASSERT(fixed_array->get(0)->IsString()); |
| 558 | String* first = String::cast(fixed_array->get(0)); |
| 559 | String* separator_raw = *separator; |
| 560 | int first_length = first->length(); |
| 561 | String::WriteToFlat(first, sink, 0, first_length); |
| 562 | sink += first_length; |
| 563 | |
| 564 | for (int i = 1; i < array_length; i++) { |
| 565 | DCHECK(sink + separator_length <= end); |
| 566 | String::WriteToFlat(separator_raw, sink, 0, separator_length); |
| 567 | sink += separator_length; |
| 568 | |
| 569 | RUNTIME_ASSERT(fixed_array->get(i)->IsString()); |
| 570 | String* element = String::cast(fixed_array->get(i)); |
| 571 | int element_length = element->length(); |
| 572 | DCHECK(sink + element_length <= end); |
| 573 | String::WriteToFlat(element, sink, 0, element_length); |
| 574 | sink += element_length; |
| 575 | } |
| 576 | DCHECK(sink == end); |
| 577 | |
| 578 | // Use %_FastOneByteArrayJoin instead. |
| 579 | DCHECK(!answer->IsOneByteRepresentation()); |
| 580 | return *answer; |
| 581 | } |
| 582 | |
| 583 | template <typename Char> |
| 584 | static void JoinSparseArrayWithSeparator(FixedArray* elements, |
| 585 | int elements_length, |
| 586 | uint32_t array_length, |
| 587 | String* separator, |
| 588 | Vector<Char> buffer) { |
| 589 | DisallowHeapAllocation no_gc; |
| 590 | int previous_separator_position = 0; |
| 591 | int separator_length = separator->length(); |
| 592 | int cursor = 0; |
| 593 | for (int i = 0; i < elements_length; i += 2) { |
| 594 | int position = NumberToInt32(elements->get(i)); |
| 595 | String* string = String::cast(elements->get(i + 1)); |
| 596 | int string_length = string->length(); |
| 597 | if (string->length() > 0) { |
| 598 | while (previous_separator_position < position) { |
| 599 | String::WriteToFlat<Char>(separator, &buffer[cursor], 0, |
| 600 | separator_length); |
| 601 | cursor += separator_length; |
| 602 | previous_separator_position++; |
| 603 | } |
| 604 | String::WriteToFlat<Char>(string, &buffer[cursor], 0, string_length); |
| 605 | cursor += string->length(); |
| 606 | } |
| 607 | } |
| 608 | if (separator_length > 0) { |
| 609 | // Array length must be representable as a signed 32-bit number, |
| 610 | // otherwise the total string length would have been too large. |
| 611 | DCHECK(array_length <= 0x7fffffff); // Is int32_t. |
| 612 | int last_array_index = static_cast<int>(array_length - 1); |
| 613 | while (previous_separator_position < last_array_index) { |
| 614 | String::WriteToFlat<Char>(separator, &buffer[cursor], 0, |
| 615 | separator_length); |
| 616 | cursor += separator_length; |
| 617 | previous_separator_position++; |
| 618 | } |
| 619 | } |
| 620 | DCHECK(cursor <= buffer.length()); |
| 621 | } |
| 622 | |
| 623 | |
| 624 | RUNTIME_FUNCTION(Runtime_SparseJoinWithSeparator) { |
| 625 | HandleScope scope(isolate); |
| 626 | DCHECK(args.length() == 3); |
| 627 | CONVERT_ARG_HANDLE_CHECKED(JSArray, elements_array, 0); |
| 628 | CONVERT_NUMBER_CHECKED(uint32_t, array_length, Uint32, args[1]); |
| 629 | CONVERT_ARG_HANDLE_CHECKED(String, separator, 2); |
| 630 | // elements_array is fast-mode JSarray of alternating positions |
| 631 | // (increasing order) and strings. |
| 632 | RUNTIME_ASSERT(elements_array->HasFastSmiOrObjectElements()); |
| 633 | // array_length is length of original array (used to add separators); |
| 634 | // separator is string to put between elements. Assumed to be non-empty. |
| 635 | RUNTIME_ASSERT(array_length > 0); |
| 636 | |
| 637 | // Find total length of join result. |
| 638 | int string_length = 0; |
| 639 | bool is_one_byte = separator->IsOneByteRepresentation(); |
| 640 | bool overflow = false; |
| 641 | CONVERT_NUMBER_CHECKED(int, elements_length, Int32, elements_array->length()); |
| 642 | RUNTIME_ASSERT(elements_length <= elements_array->elements()->length()); |
| 643 | RUNTIME_ASSERT((elements_length & 1) == 0); // Even length. |
| 644 | FixedArray* elements = FixedArray::cast(elements_array->elements()); |
| 645 | for (int i = 0; i < elements_length; i += 2) { |
| 646 | RUNTIME_ASSERT(elements->get(i)->IsNumber()); |
| 647 | CONVERT_NUMBER_CHECKED(uint32_t, position, Uint32, elements->get(i)); |
| 648 | RUNTIME_ASSERT(position < array_length); |
| 649 | RUNTIME_ASSERT(elements->get(i + 1)->IsString()); |
| 650 | } |
| 651 | |
| 652 | { |
| 653 | DisallowHeapAllocation no_gc; |
| 654 | for (int i = 0; i < elements_length; i += 2) { |
| 655 | String* string = String::cast(elements->get(i + 1)); |
| 656 | int length = string->length(); |
| 657 | if (is_one_byte && !string->IsOneByteRepresentation()) { |
| 658 | is_one_byte = false; |
| 659 | } |
| 660 | if (length > String::kMaxLength || |
| 661 | String::kMaxLength - length < string_length) { |
| 662 | overflow = true; |
| 663 | break; |
| 664 | } |
| 665 | string_length += length; |
| 666 | } |
| 667 | } |
| 668 | |
| 669 | int separator_length = separator->length(); |
| 670 | if (!overflow && separator_length > 0) { |
| 671 | if (array_length <= 0x7fffffffu) { |
| 672 | int separator_count = static_cast<int>(array_length) - 1; |
| 673 | int remaining_length = String::kMaxLength - string_length; |
| 674 | if ((remaining_length / separator_length) >= separator_count) { |
| 675 | string_length += separator_length * (array_length - 1); |
| 676 | } else { |
| 677 | // Not room for the separators within the maximal string length. |
| 678 | overflow = true; |
| 679 | } |
| 680 | } else { |
| 681 | // Nonempty separator and at least 2^31-1 separators necessary |
| 682 | // means that the string is too large to create. |
| 683 | STATIC_ASSERT(String::kMaxLength < 0x7fffffff); |
| 684 | overflow = true; |
| 685 | } |
| 686 | } |
| 687 | if (overflow) { |
| 688 | // Throw an exception if the resulting string is too large. See |
| 689 | // https://code.google.com/p/chromium/issues/detail?id=336820 |
| 690 | // for details. |
| 691 | THROW_NEW_ERROR_RETURN_FAILURE(isolate, NewInvalidStringLengthError()); |
| 692 | } |
| 693 | |
| 694 | if (is_one_byte) { |
| 695 | Handle<SeqOneByteString> result = isolate->factory() |
| 696 | ->NewRawOneByteString(string_length) |
| 697 | .ToHandleChecked(); |
| 698 | JoinSparseArrayWithSeparator<uint8_t>( |
| 699 | FixedArray::cast(elements_array->elements()), elements_length, |
| 700 | array_length, *separator, |
| 701 | Vector<uint8_t>(result->GetChars(), string_length)); |
| 702 | return *result; |
| 703 | } else { |
| 704 | Handle<SeqTwoByteString> result = isolate->factory() |
| 705 | ->NewRawTwoByteString(string_length) |
| 706 | .ToHandleChecked(); |
| 707 | JoinSparseArrayWithSeparator<uc16>( |
| 708 | FixedArray::cast(elements_array->elements()), elements_length, |
| 709 | array_length, *separator, |
| 710 | Vector<uc16>(result->GetChars(), string_length)); |
| 711 | return *result; |
| 712 | } |
| 713 | } |
| 714 | |
| 715 | |
| 716 | // Copies Latin1 characters to the given fixed array looking up |
| 717 | // one-char strings in the cache. Gives up on the first char that is |
| 718 | // not in the cache and fills the remainder with smi zeros. Returns |
| 719 | // the length of the successfully copied prefix. |
| 720 | static int CopyCachedOneByteCharsToArray(Heap* heap, const uint8_t* chars, |
| 721 | FixedArray* elements, int length) { |
| 722 | DisallowHeapAllocation no_gc; |
| 723 | FixedArray* one_byte_cache = heap->single_character_string_cache(); |
| 724 | Object* undefined = heap->undefined_value(); |
| 725 | int i; |
| 726 | WriteBarrierMode mode = elements->GetWriteBarrierMode(no_gc); |
| 727 | for (i = 0; i < length; ++i) { |
| 728 | Object* value = one_byte_cache->get(chars[i]); |
| 729 | if (value == undefined) break; |
| 730 | elements->set(i, value, mode); |
| 731 | } |
| 732 | if (i < length) { |
| 733 | DCHECK(Smi::FromInt(0) == 0); |
| 734 | memset(elements->data_start() + i, 0, kPointerSize * (length - i)); |
| 735 | } |
| 736 | #ifdef DEBUG |
| 737 | for (int j = 0; j < length; ++j) { |
| 738 | Object* element = elements->get(j); |
| 739 | DCHECK(element == Smi::FromInt(0) || |
| 740 | (element->IsString() && String::cast(element)->LooksValid())); |
| 741 | } |
| 742 | #endif |
| 743 | return i; |
| 744 | } |
| 745 | |
| 746 | |
| 747 | // Converts a String to JSArray. |
| 748 | // For example, "foo" => ["f", "o", "o"]. |
| 749 | RUNTIME_FUNCTION(Runtime_StringToArray) { |
| 750 | HandleScope scope(isolate); |
| 751 | DCHECK(args.length() == 2); |
| 752 | CONVERT_ARG_HANDLE_CHECKED(String, s, 0); |
| 753 | CONVERT_NUMBER_CHECKED(uint32_t, limit, Uint32, args[1]); |
| 754 | |
| 755 | s = String::Flatten(s); |
| 756 | const int length = static_cast<int>(Min<uint32_t>(s->length(), limit)); |
| 757 | |
| 758 | Handle<FixedArray> elements; |
| 759 | int position = 0; |
| 760 | if (s->IsFlat() && s->IsOneByteRepresentation()) { |
| 761 | // Try using cached chars where possible. |
| 762 | elements = isolate->factory()->NewUninitializedFixedArray(length); |
| 763 | |
| 764 | DisallowHeapAllocation no_gc; |
| 765 | String::FlatContent content = s->GetFlatContent(); |
| 766 | if (content.IsOneByte()) { |
| 767 | Vector<const uint8_t> chars = content.ToOneByteVector(); |
| 768 | // Note, this will initialize all elements (not only the prefix) |
| 769 | // to prevent GC from seeing partially initialized array. |
| 770 | position = CopyCachedOneByteCharsToArray(isolate->heap(), chars.start(), |
| 771 | *elements, length); |
| 772 | } else { |
| 773 | MemsetPointer(elements->data_start(), isolate->heap()->undefined_value(), |
| 774 | length); |
| 775 | } |
| 776 | } else { |
| 777 | elements = isolate->factory()->NewFixedArray(length); |
| 778 | } |
| 779 | for (int i = position; i < length; ++i) { |
| 780 | Handle<Object> str = |
| 781 | isolate->factory()->LookupSingleCharacterStringFromCode(s->Get(i)); |
| 782 | elements->set(i, *str); |
| 783 | } |
| 784 | |
| 785 | #ifdef DEBUG |
| 786 | for (int i = 0; i < length; ++i) { |
| 787 | DCHECK(String::cast(elements->get(i))->length() == 1); |
| 788 | } |
| 789 | #endif |
| 790 | |
| 791 | return *isolate->factory()->NewJSArrayWithElements(elements); |
| 792 | } |
| 793 | |
| 794 | |
| 795 | static inline bool ToUpperOverflows(uc32 character) { |
| 796 | // y with umlauts and the micro sign are the only characters that stop |
| 797 | // fitting into one-byte when converting to uppercase. |
| 798 | static const uc32 yuml_code = 0xff; |
| 799 | static const uc32 micro_code = 0xb5; |
| 800 | return (character == yuml_code || character == micro_code); |
| 801 | } |
| 802 | |
| 803 | |
| 804 | template <class Converter> |
| 805 | MUST_USE_RESULT static Object* ConvertCaseHelper( |
| 806 | Isolate* isolate, String* string, SeqString* result, int result_length, |
| 807 | unibrow::Mapping<Converter, 128>* mapping) { |
| 808 | DisallowHeapAllocation no_gc; |
| 809 | // We try this twice, once with the assumption that the result is no longer |
| 810 | // than the input and, if that assumption breaks, again with the exact |
| 811 | // length. This may not be pretty, but it is nicer than what was here before |
| 812 | // and I hereby claim my vaffel-is. |
| 813 | // |
| 814 | // NOTE: This assumes that the upper/lower case of an ASCII |
| 815 | // character is also ASCII. This is currently the case, but it |
| 816 | // might break in the future if we implement more context and locale |
| 817 | // dependent upper/lower conversions. |
| 818 | bool has_changed_character = false; |
| 819 | |
| 820 | // Convert all characters to upper case, assuming that they will fit |
| 821 | // in the buffer |
| 822 | StringCharacterStream stream(string); |
| 823 | unibrow::uchar chars[Converter::kMaxWidth]; |
| 824 | // We can assume that the string is not empty |
| 825 | uc32 current = stream.GetNext(); |
| 826 | bool ignore_overflow = Converter::kIsToLower || result->IsSeqTwoByteString(); |
| 827 | for (int i = 0; i < result_length;) { |
| 828 | bool has_next = stream.HasMore(); |
| 829 | uc32 next = has_next ? stream.GetNext() : 0; |
| 830 | int char_length = mapping->get(current, next, chars); |
| 831 | if (char_length == 0) { |
| 832 | // The case conversion of this character is the character itself. |
| 833 | result->Set(i, current); |
| 834 | i++; |
| 835 | } else if (char_length == 1 && |
| 836 | (ignore_overflow || !ToUpperOverflows(current))) { |
| 837 | // Common case: converting the letter resulted in one character. |
| 838 | DCHECK(static_cast<uc32>(chars[0]) != current); |
| 839 | result->Set(i, chars[0]); |
| 840 | has_changed_character = true; |
| 841 | i++; |
| 842 | } else if (result_length == string->length()) { |
| 843 | bool overflows = ToUpperOverflows(current); |
| 844 | // We've assumed that the result would be as long as the |
| 845 | // input but here is a character that converts to several |
| 846 | // characters. No matter, we calculate the exact length |
| 847 | // of the result and try the whole thing again. |
| 848 | // |
| 849 | // Note that this leaves room for optimization. We could just |
| 850 | // memcpy what we already have to the result string. Also, |
| 851 | // the result string is the last object allocated we could |
| 852 | // "realloc" it and probably, in the vast majority of cases, |
| 853 | // extend the existing string to be able to hold the full |
| 854 | // result. |
| 855 | int next_length = 0; |
| 856 | if (has_next) { |
| 857 | next_length = mapping->get(next, 0, chars); |
| 858 | if (next_length == 0) next_length = 1; |
| 859 | } |
| 860 | int current_length = i + char_length + next_length; |
| 861 | while (stream.HasMore()) { |
| 862 | current = stream.GetNext(); |
| 863 | overflows |= ToUpperOverflows(current); |
| 864 | // NOTE: we use 0 as the next character here because, while |
| 865 | // the next character may affect what a character converts to, |
| 866 | // it does not in any case affect the length of what it convert |
| 867 | // to. |
| 868 | int char_length = mapping->get(current, 0, chars); |
| 869 | if (char_length == 0) char_length = 1; |
| 870 | current_length += char_length; |
| 871 | if (current_length > String::kMaxLength) { |
| 872 | AllowHeapAllocation allocate_error_and_return; |
| 873 | THROW_NEW_ERROR_RETURN_FAILURE(isolate, |
| 874 | NewInvalidStringLengthError()); |
| 875 | } |
| 876 | } |
| 877 | // Try again with the real length. Return signed if we need |
| 878 | // to allocate a two-byte string for to uppercase. |
| 879 | return (overflows && !ignore_overflow) ? Smi::FromInt(-current_length) |
| 880 | : Smi::FromInt(current_length); |
| 881 | } else { |
| 882 | for (int j = 0; j < char_length; j++) { |
| 883 | result->Set(i, chars[j]); |
| 884 | i++; |
| 885 | } |
| 886 | has_changed_character = true; |
| 887 | } |
| 888 | current = next; |
| 889 | } |
| 890 | if (has_changed_character) { |
| 891 | return result; |
| 892 | } else { |
| 893 | // If we didn't actually change anything in doing the conversion |
| 894 | // we simple return the result and let the converted string |
| 895 | // become garbage; there is no reason to keep two identical strings |
| 896 | // alive. |
| 897 | return string; |
| 898 | } |
| 899 | } |
| 900 | |
| 901 | |
| 902 | static const uintptr_t kOneInEveryByte = kUintptrAllBitsSet / 0xFF; |
| 903 | static const uintptr_t kAsciiMask = kOneInEveryByte << 7; |
| 904 | |
| 905 | // Given a word and two range boundaries returns a word with high bit |
| 906 | // set in every byte iff the corresponding input byte was strictly in |
| 907 | // the range (m, n). All the other bits in the result are cleared. |
| 908 | // This function is only useful when it can be inlined and the |
| 909 | // boundaries are statically known. |
| 910 | // Requires: all bytes in the input word and the boundaries must be |
| 911 | // ASCII (less than 0x7F). |
| 912 | static inline uintptr_t AsciiRangeMask(uintptr_t w, char m, char n) { |
| 913 | // Use strict inequalities since in edge cases the function could be |
| 914 | // further simplified. |
| 915 | DCHECK(0 < m && m < n); |
| 916 | // Has high bit set in every w byte less than n. |
| 917 | uintptr_t tmp1 = kOneInEveryByte * (0x7F + n) - w; |
| 918 | // Has high bit set in every w byte greater than m. |
| 919 | uintptr_t tmp2 = w + kOneInEveryByte * (0x7F - m); |
| 920 | return (tmp1 & tmp2 & (kOneInEveryByte * 0x80)); |
| 921 | } |
| 922 | |
| 923 | |
| 924 | #ifdef DEBUG |
| 925 | static bool CheckFastAsciiConvert(char* dst, const char* src, int length, |
| 926 | bool changed, bool is_to_lower) { |
| 927 | bool expected_changed = false; |
| 928 | for (int i = 0; i < length; i++) { |
| 929 | if (dst[i] == src[i]) continue; |
| 930 | expected_changed = true; |
| 931 | if (is_to_lower) { |
| 932 | DCHECK('A' <= src[i] && src[i] <= 'Z'); |
| 933 | DCHECK(dst[i] == src[i] + ('a' - 'A')); |
| 934 | } else { |
| 935 | DCHECK('a' <= src[i] && src[i] <= 'z'); |
| 936 | DCHECK(dst[i] == src[i] - ('a' - 'A')); |
| 937 | } |
| 938 | } |
| 939 | return (expected_changed == changed); |
| 940 | } |
| 941 | #endif |
| 942 | |
| 943 | |
| 944 | template <class Converter> |
| 945 | static bool FastAsciiConvert(char* dst, const char* src, int length, |
| 946 | bool* changed_out) { |
| 947 | #ifdef DEBUG |
| 948 | char* saved_dst = dst; |
| 949 | const char* saved_src = src; |
| 950 | #endif |
| 951 | DisallowHeapAllocation no_gc; |
| 952 | // We rely on the distance between upper and lower case letters |
| 953 | // being a known power of 2. |
| 954 | DCHECK('a' - 'A' == (1 << 5)); |
| 955 | // Boundaries for the range of input characters than require conversion. |
| 956 | static const char lo = Converter::kIsToLower ? 'A' - 1 : 'a' - 1; |
| 957 | static const char hi = Converter::kIsToLower ? 'Z' + 1 : 'z' + 1; |
| 958 | bool changed = false; |
| 959 | uintptr_t or_acc = 0; |
| 960 | const char* const limit = src + length; |
| 961 | |
| 962 | // dst is newly allocated and always aligned. |
| 963 | DCHECK(IsAligned(reinterpret_cast<intptr_t>(dst), sizeof(uintptr_t))); |
| 964 | // Only attempt processing one word at a time if src is also aligned. |
| 965 | if (IsAligned(reinterpret_cast<intptr_t>(src), sizeof(uintptr_t))) { |
| 966 | // Process the prefix of the input that requires no conversion one aligned |
| 967 | // (machine) word at a time. |
| 968 | while (src <= limit - sizeof(uintptr_t)) { |
| 969 | const uintptr_t w = *reinterpret_cast<const uintptr_t*>(src); |
| 970 | or_acc |= w; |
| 971 | if (AsciiRangeMask(w, lo, hi) != 0) { |
| 972 | changed = true; |
| 973 | break; |
| 974 | } |
| 975 | *reinterpret_cast<uintptr_t*>(dst) = w; |
| 976 | src += sizeof(uintptr_t); |
| 977 | dst += sizeof(uintptr_t); |
| 978 | } |
| 979 | // Process the remainder of the input performing conversion when |
| 980 | // required one word at a time. |
| 981 | while (src <= limit - sizeof(uintptr_t)) { |
| 982 | const uintptr_t w = *reinterpret_cast<const uintptr_t*>(src); |
| 983 | or_acc |= w; |
| 984 | uintptr_t m = AsciiRangeMask(w, lo, hi); |
| 985 | // The mask has high (7th) bit set in every byte that needs |
| 986 | // conversion and we know that the distance between cases is |
| 987 | // 1 << 5. |
| 988 | *reinterpret_cast<uintptr_t*>(dst) = w ^ (m >> 2); |
| 989 | src += sizeof(uintptr_t); |
| 990 | dst += sizeof(uintptr_t); |
| 991 | } |
| 992 | } |
| 993 | // Process the last few bytes of the input (or the whole input if |
| 994 | // unaligned access is not supported). |
| 995 | while (src < limit) { |
| 996 | char c = *src; |
| 997 | or_acc |= c; |
| 998 | if (lo < c && c < hi) { |
| 999 | c ^= (1 << 5); |
| 1000 | changed = true; |
| 1001 | } |
| 1002 | *dst = c; |
| 1003 | ++src; |
| 1004 | ++dst; |
| 1005 | } |
| 1006 | |
| 1007 | if ((or_acc & kAsciiMask) != 0) return false; |
| 1008 | |
| 1009 | DCHECK(CheckFastAsciiConvert(saved_dst, saved_src, length, changed, |
| 1010 | Converter::kIsToLower)); |
| 1011 | |
| 1012 | *changed_out = changed; |
| 1013 | return true; |
| 1014 | } |
| 1015 | |
| 1016 | |
| 1017 | template <class Converter> |
| 1018 | MUST_USE_RESULT static Object* ConvertCase( |
| 1019 | Handle<String> s, Isolate* isolate, |
| 1020 | unibrow::Mapping<Converter, 128>* mapping) { |
| 1021 | s = String::Flatten(s); |
| 1022 | int length = s->length(); |
| 1023 | // Assume that the string is not empty; we need this assumption later |
| 1024 | if (length == 0) return *s; |
| 1025 | |
| 1026 | // Simpler handling of ASCII strings. |
| 1027 | // |
| 1028 | // NOTE: This assumes that the upper/lower case of an ASCII |
| 1029 | // character is also ASCII. This is currently the case, but it |
| 1030 | // might break in the future if we implement more context and locale |
| 1031 | // dependent upper/lower conversions. |
| 1032 | if (s->IsOneByteRepresentationUnderneath()) { |
| 1033 | // Same length as input. |
| 1034 | Handle<SeqOneByteString> result = |
| 1035 | isolate->factory()->NewRawOneByteString(length).ToHandleChecked(); |
| 1036 | DisallowHeapAllocation no_gc; |
| 1037 | String::FlatContent flat_content = s->GetFlatContent(); |
| 1038 | DCHECK(flat_content.IsFlat()); |
| 1039 | bool has_changed_character = false; |
| 1040 | bool is_ascii = FastAsciiConvert<Converter>( |
| 1041 | reinterpret_cast<char*>(result->GetChars()), |
| 1042 | reinterpret_cast<const char*>(flat_content.ToOneByteVector().start()), |
| 1043 | length, &has_changed_character); |
| 1044 | // If not ASCII, we discard the result and take the 2 byte path. |
| 1045 | if (is_ascii) return has_changed_character ? *result : *s; |
| 1046 | } |
| 1047 | |
| 1048 | Handle<SeqString> result; // Same length as input. |
| 1049 | if (s->IsOneByteRepresentation()) { |
| 1050 | result = isolate->factory()->NewRawOneByteString(length).ToHandleChecked(); |
| 1051 | } else { |
| 1052 | result = isolate->factory()->NewRawTwoByteString(length).ToHandleChecked(); |
| 1053 | } |
| 1054 | |
| 1055 | Object* answer = ConvertCaseHelper(isolate, *s, *result, length, mapping); |
| 1056 | if (answer->IsException() || answer->IsString()) return answer; |
| 1057 | |
| 1058 | DCHECK(answer->IsSmi()); |
| 1059 | length = Smi::cast(answer)->value(); |
| 1060 | if (s->IsOneByteRepresentation() && length > 0) { |
| 1061 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 1062 | isolate, result, isolate->factory()->NewRawOneByteString(length)); |
| 1063 | } else { |
| 1064 | if (length < 0) length = -length; |
| 1065 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 1066 | isolate, result, isolate->factory()->NewRawTwoByteString(length)); |
| 1067 | } |
| 1068 | return ConvertCaseHelper(isolate, *s, *result, length, mapping); |
| 1069 | } |
| 1070 | |
| 1071 | |
| 1072 | RUNTIME_FUNCTION(Runtime_StringToLowerCase) { |
| 1073 | HandleScope scope(isolate); |
| 1074 | DCHECK(args.length() == 1); |
| 1075 | CONVERT_ARG_HANDLE_CHECKED(String, s, 0); |
| 1076 | return ConvertCase(s, isolate, isolate->runtime_state()->to_lower_mapping()); |
| 1077 | } |
| 1078 | |
| 1079 | |
| 1080 | RUNTIME_FUNCTION(Runtime_StringToUpperCase) { |
| 1081 | HandleScope scope(isolate); |
| 1082 | DCHECK(args.length() == 1); |
| 1083 | CONVERT_ARG_HANDLE_CHECKED(String, s, 0); |
| 1084 | return ConvertCase(s, isolate, isolate->runtime_state()->to_upper_mapping()); |
| 1085 | } |
| 1086 | |
| 1087 | |
| 1088 | RUNTIME_FUNCTION(Runtime_StringTrim) { |
| 1089 | HandleScope scope(isolate); |
| 1090 | DCHECK(args.length() == 3); |
| 1091 | |
| 1092 | CONVERT_ARG_HANDLE_CHECKED(String, string, 0); |
| 1093 | CONVERT_BOOLEAN_ARG_CHECKED(trimLeft, 1); |
| 1094 | CONVERT_BOOLEAN_ARG_CHECKED(trimRight, 2); |
| 1095 | |
| 1096 | string = String::Flatten(string); |
| 1097 | int length = string->length(); |
| 1098 | |
| 1099 | int left = 0; |
| 1100 | UnicodeCache* unicode_cache = isolate->unicode_cache(); |
| 1101 | if (trimLeft) { |
| 1102 | while (left < length && |
| 1103 | unicode_cache->IsWhiteSpaceOrLineTerminator(string->Get(left))) { |
| 1104 | left++; |
| 1105 | } |
| 1106 | } |
| 1107 | |
| 1108 | int right = length; |
| 1109 | if (trimRight) { |
| 1110 | while ( |
| 1111 | right > left && |
| 1112 | unicode_cache->IsWhiteSpaceOrLineTerminator(string->Get(right - 1))) { |
| 1113 | right--; |
| 1114 | } |
| 1115 | } |
| 1116 | |
| 1117 | return *isolate->factory()->NewSubString(string, left, right); |
| 1118 | } |
| 1119 | |
| 1120 | |
| 1121 | RUNTIME_FUNCTION(Runtime_TruncateString) { |
| 1122 | HandleScope scope(isolate); |
| 1123 | DCHECK(args.length() == 2); |
| 1124 | CONVERT_ARG_HANDLE_CHECKED(SeqString, string, 0); |
| 1125 | CONVERT_INT32_ARG_CHECKED(new_length, 1); |
| 1126 | RUNTIME_ASSERT(new_length >= 0); |
| 1127 | return *SeqString::Truncate(string, new_length); |
| 1128 | } |
| 1129 | |
| 1130 | |
| 1131 | RUNTIME_FUNCTION(Runtime_NewString) { |
| 1132 | HandleScope scope(isolate); |
| 1133 | DCHECK(args.length() == 2); |
| 1134 | CONVERT_INT32_ARG_CHECKED(length, 0); |
| 1135 | CONVERT_BOOLEAN_ARG_CHECKED(is_one_byte, 1); |
| 1136 | if (length == 0) return isolate->heap()->empty_string(); |
| 1137 | Handle<String> result; |
| 1138 | if (is_one_byte) { |
| 1139 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 1140 | isolate, result, isolate->factory()->NewRawOneByteString(length)); |
| 1141 | } else { |
| 1142 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 1143 | isolate, result, isolate->factory()->NewRawTwoByteString(length)); |
| 1144 | } |
| 1145 | return *result; |
| 1146 | } |
| 1147 | |
| 1148 | |
| 1149 | RUNTIME_FUNCTION(Runtime_StringEquals) { |
| 1150 | HandleScope handle_scope(isolate); |
| 1151 | DCHECK(args.length() == 2); |
| 1152 | |
| 1153 | CONVERT_ARG_HANDLE_CHECKED(String, x, 0); |
| 1154 | CONVERT_ARG_HANDLE_CHECKED(String, y, 1); |
| 1155 | |
| 1156 | bool not_equal = !String::Equals(x, y); |
| 1157 | // This is slightly convoluted because the value that signifies |
| 1158 | // equality is 0 and inequality is 1 so we have to negate the result |
| 1159 | // from String::Equals. |
| 1160 | DCHECK(not_equal == 0 || not_equal == 1); |
| 1161 | STATIC_ASSERT(EQUAL == 0); |
| 1162 | STATIC_ASSERT(NOT_EQUAL == 1); |
| 1163 | return Smi::FromInt(not_equal); |
| 1164 | } |
| 1165 | |
| 1166 | |
| 1167 | RUNTIME_FUNCTION(Runtime_FlattenString) { |
| 1168 | HandleScope scope(isolate); |
| 1169 | DCHECK(args.length() == 1); |
| 1170 | CONVERT_ARG_HANDLE_CHECKED(String, str, 0); |
| 1171 | return *String::Flatten(str); |
| 1172 | } |
| 1173 | |
| 1174 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1175 | RUNTIME_FUNCTION(Runtime_StringCharFromCode) { |
| 1176 | HandleScope handlescope(isolate); |
| 1177 | DCHECK_EQ(1, args.length()); |
| 1178 | if (args[0]->IsNumber()) { |
| 1179 | CONVERT_NUMBER_CHECKED(uint32_t, code, Uint32, args[0]); |
| 1180 | code &= 0xffff; |
| 1181 | return *isolate->factory()->LookupSingleCharacterStringFromCode(code); |
| 1182 | } |
| 1183 | return isolate->heap()->empty_string(); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1184 | } |
| 1185 | |
| 1186 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1187 | RUNTIME_FUNCTION(Runtime_StringCharAt) { |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1188 | SealHandleScope shs(isolate); |
| 1189 | DCHECK(args.length() == 2); |
| 1190 | if (!args[0]->IsString()) return Smi::FromInt(0); |
| 1191 | if (!args[1]->IsNumber()) return Smi::FromInt(0); |
| 1192 | if (std::isinf(args.number_at(1))) return isolate->heap()->empty_string(); |
| 1193 | Object* code = __RT_impl_Runtime_StringCharCodeAtRT(args, isolate); |
| 1194 | if (code->IsNaN()) return isolate->heap()->empty_string(); |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1195 | return __RT_impl_Runtime_StringCharFromCode(Arguments(1, &code), isolate); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1196 | } |
| 1197 | |
| 1198 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1199 | RUNTIME_FUNCTION(Runtime_OneByteSeqStringGetChar) { |
| 1200 | SealHandleScope shs(isolate); |
| 1201 | DCHECK(args.length() == 2); |
| 1202 | CONVERT_ARG_CHECKED(SeqOneByteString, string, 0); |
| 1203 | CONVERT_INT32_ARG_CHECKED(index, 1); |
| 1204 | return Smi::FromInt(string->SeqOneByteStringGet(index)); |
| 1205 | } |
| 1206 | |
| 1207 | |
| 1208 | RUNTIME_FUNCTION(Runtime_OneByteSeqStringSetChar) { |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1209 | SealHandleScope shs(isolate); |
| 1210 | DCHECK(args.length() == 3); |
| 1211 | CONVERT_INT32_ARG_CHECKED(index, 0); |
| 1212 | CONVERT_INT32_ARG_CHECKED(value, 1); |
| 1213 | CONVERT_ARG_CHECKED(SeqOneByteString, string, 2); |
| 1214 | string->SeqOneByteStringSet(index, value); |
| 1215 | return string; |
| 1216 | } |
| 1217 | |
| 1218 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1219 | RUNTIME_FUNCTION(Runtime_TwoByteSeqStringGetChar) { |
| 1220 | SealHandleScope shs(isolate); |
| 1221 | DCHECK(args.length() == 2); |
| 1222 | CONVERT_ARG_CHECKED(SeqTwoByteString, string, 0); |
| 1223 | CONVERT_INT32_ARG_CHECKED(index, 1); |
| 1224 | return Smi::FromInt(string->SeqTwoByteStringGet(index)); |
| 1225 | } |
| 1226 | |
| 1227 | |
| 1228 | RUNTIME_FUNCTION(Runtime_TwoByteSeqStringSetChar) { |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1229 | SealHandleScope shs(isolate); |
| 1230 | DCHECK(args.length() == 3); |
| 1231 | CONVERT_INT32_ARG_CHECKED(index, 0); |
| 1232 | CONVERT_INT32_ARG_CHECKED(value, 1); |
| 1233 | CONVERT_ARG_CHECKED(SeqTwoByteString, string, 2); |
| 1234 | string->SeqTwoByteStringSet(index, value); |
| 1235 | return string; |
| 1236 | } |
| 1237 | |
| 1238 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1239 | RUNTIME_FUNCTION(Runtime_StringCharCodeAt) { |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1240 | SealHandleScope shs(isolate); |
| 1241 | DCHECK(args.length() == 2); |
| 1242 | if (!args[0]->IsString()) return isolate->heap()->undefined_value(); |
| 1243 | if (!args[1]->IsNumber()) return isolate->heap()->undefined_value(); |
| 1244 | if (std::isinf(args.number_at(1))) return isolate->heap()->nan_value(); |
| 1245 | return __RT_impl_Runtime_StringCharCodeAtRT(args, isolate); |
| 1246 | } |
| 1247 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1248 | } // namespace internal |
| 1249 | } // namespace v8 |