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/regexp/jsregexp-inl.h" |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 9 | #include "src/string-builder.h" |
| 10 | #include "src/string-search.h" |
| 11 | |
| 12 | namespace v8 { |
| 13 | namespace internal { |
| 14 | |
| 15 | |
| 16 | // Perform string match of pattern on subject, starting at start index. |
| 17 | // Caller must ensure that 0 <= start_index <= sub->length(), |
| 18 | // and should check that pat->length() + start_index <= sub->length(). |
| 19 | int StringMatch(Isolate* isolate, Handle<String> sub, Handle<String> pat, |
| 20 | int start_index) { |
| 21 | DCHECK(0 <= start_index); |
| 22 | DCHECK(start_index <= sub->length()); |
| 23 | |
| 24 | int pattern_length = pat->length(); |
| 25 | if (pattern_length == 0) return start_index; |
| 26 | |
| 27 | int subject_length = sub->length(); |
| 28 | if (start_index + pattern_length > subject_length) return -1; |
| 29 | |
| 30 | sub = String::Flatten(sub); |
| 31 | pat = String::Flatten(pat); |
| 32 | |
| 33 | DisallowHeapAllocation no_gc; // ensure vectors stay valid |
| 34 | // Extract flattened substrings of cons strings before getting encoding. |
| 35 | String::FlatContent seq_sub = sub->GetFlatContent(); |
| 36 | String::FlatContent seq_pat = pat->GetFlatContent(); |
| 37 | |
| 38 | // dispatch on type of strings |
| 39 | if (seq_pat.IsOneByte()) { |
| 40 | Vector<const uint8_t> pat_vector = seq_pat.ToOneByteVector(); |
| 41 | if (seq_sub.IsOneByte()) { |
| 42 | return SearchString(isolate, seq_sub.ToOneByteVector(), pat_vector, |
| 43 | start_index); |
| 44 | } |
| 45 | return SearchString(isolate, seq_sub.ToUC16Vector(), pat_vector, |
| 46 | start_index); |
| 47 | } |
| 48 | Vector<const uc16> pat_vector = seq_pat.ToUC16Vector(); |
| 49 | if (seq_sub.IsOneByte()) { |
| 50 | return SearchString(isolate, seq_sub.ToOneByteVector(), pat_vector, |
| 51 | start_index); |
| 52 | } |
| 53 | return SearchString(isolate, seq_sub.ToUC16Vector(), pat_vector, start_index); |
| 54 | } |
| 55 | |
| 56 | |
| 57 | // This may return an empty MaybeHandle if an exception is thrown or |
| 58 | // we abort due to reaching the recursion limit. |
| 59 | MaybeHandle<String> StringReplaceOneCharWithString( |
| 60 | Isolate* isolate, Handle<String> subject, Handle<String> search, |
| 61 | Handle<String> replace, bool* found, int recursion_limit) { |
| 62 | StackLimitCheck stackLimitCheck(isolate); |
| 63 | if (stackLimitCheck.HasOverflowed() || (recursion_limit == 0)) { |
| 64 | return MaybeHandle<String>(); |
| 65 | } |
| 66 | recursion_limit--; |
| 67 | if (subject->IsConsString()) { |
| 68 | ConsString* cons = ConsString::cast(*subject); |
| 69 | Handle<String> first = Handle<String>(cons->first()); |
| 70 | Handle<String> second = Handle<String>(cons->second()); |
| 71 | Handle<String> new_first; |
| 72 | if (!StringReplaceOneCharWithString(isolate, first, search, replace, found, |
| 73 | recursion_limit).ToHandle(&new_first)) { |
| 74 | return MaybeHandle<String>(); |
| 75 | } |
| 76 | if (*found) return isolate->factory()->NewConsString(new_first, second); |
| 77 | |
| 78 | Handle<String> new_second; |
| 79 | if (!StringReplaceOneCharWithString(isolate, second, search, replace, found, |
| 80 | recursion_limit) |
| 81 | .ToHandle(&new_second)) { |
| 82 | return MaybeHandle<String>(); |
| 83 | } |
| 84 | if (*found) return isolate->factory()->NewConsString(first, new_second); |
| 85 | |
| 86 | return subject; |
| 87 | } else { |
| 88 | int index = StringMatch(isolate, subject, search, 0); |
| 89 | if (index == -1) return subject; |
| 90 | *found = true; |
| 91 | Handle<String> first = isolate->factory()->NewSubString(subject, 0, index); |
| 92 | Handle<String> cons1; |
| 93 | ASSIGN_RETURN_ON_EXCEPTION( |
| 94 | isolate, cons1, isolate->factory()->NewConsString(first, replace), |
| 95 | String); |
| 96 | Handle<String> second = |
| 97 | isolate->factory()->NewSubString(subject, index + 1, subject->length()); |
| 98 | return isolate->factory()->NewConsString(cons1, second); |
| 99 | } |
| 100 | } |
| 101 | |
| 102 | |
| 103 | RUNTIME_FUNCTION(Runtime_StringReplaceOneCharWithString) { |
| 104 | HandleScope scope(isolate); |
| 105 | DCHECK(args.length() == 3); |
| 106 | CONVERT_ARG_HANDLE_CHECKED(String, subject, 0); |
| 107 | CONVERT_ARG_HANDLE_CHECKED(String, search, 1); |
| 108 | CONVERT_ARG_HANDLE_CHECKED(String, replace, 2); |
| 109 | |
| 110 | // If the cons string tree is too deep, we simply abort the recursion and |
| 111 | // retry with a flattened subject string. |
| 112 | const int kRecursionLimit = 0x1000; |
| 113 | bool found = false; |
| 114 | Handle<String> result; |
| 115 | if (StringReplaceOneCharWithString(isolate, subject, search, replace, &found, |
| 116 | kRecursionLimit).ToHandle(&result)) { |
| 117 | return *result; |
| 118 | } |
| 119 | if (isolate->has_pending_exception()) return isolate->heap()->exception(); |
| 120 | |
| 121 | subject = String::Flatten(subject); |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 122 | if (StringReplaceOneCharWithString(isolate, subject, search, replace, &found, |
| 123 | kRecursionLimit).ToHandle(&result)) { |
| 124 | return *result; |
| 125 | } |
| 126 | if (isolate->has_pending_exception()) return isolate->heap()->exception(); |
| 127 | // In case of empty handle and no pending exception we have stack overflow. |
| 128 | return isolate->StackOverflow(); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 129 | } |
| 130 | |
| 131 | |
| 132 | RUNTIME_FUNCTION(Runtime_StringIndexOf) { |
| 133 | HandleScope scope(isolate); |
| 134 | DCHECK(args.length() == 3); |
| 135 | |
| 136 | CONVERT_ARG_HANDLE_CHECKED(String, sub, 0); |
| 137 | CONVERT_ARG_HANDLE_CHECKED(String, pat, 1); |
| 138 | CONVERT_ARG_HANDLE_CHECKED(Object, index, 2); |
| 139 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 140 | uint32_t start_index = 0; |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 141 | if (!index->ToArrayIndex(&start_index)) return Smi::FromInt(-1); |
| 142 | |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 143 | CHECK(start_index <= static_cast<uint32_t>(sub->length())); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 144 | int position = StringMatch(isolate, sub, pat, start_index); |
| 145 | return Smi::FromInt(position); |
| 146 | } |
| 147 | |
| 148 | |
| 149 | template <typename schar, typename pchar> |
| 150 | static int StringMatchBackwards(Vector<const schar> subject, |
| 151 | Vector<const pchar> pattern, int idx) { |
| 152 | int pattern_length = pattern.length(); |
| 153 | DCHECK(pattern_length >= 1); |
| 154 | DCHECK(idx + pattern_length <= subject.length()); |
| 155 | |
| 156 | if (sizeof(schar) == 1 && sizeof(pchar) > 1) { |
| 157 | for (int i = 0; i < pattern_length; i++) { |
| 158 | uc16 c = pattern[i]; |
| 159 | if (c > String::kMaxOneByteCharCode) { |
| 160 | return -1; |
| 161 | } |
| 162 | } |
| 163 | } |
| 164 | |
| 165 | pchar pattern_first_char = pattern[0]; |
| 166 | for (int i = idx; i >= 0; i--) { |
| 167 | if (subject[i] != pattern_first_char) continue; |
| 168 | int j = 1; |
| 169 | while (j < pattern_length) { |
| 170 | if (pattern[j] != subject[i + j]) { |
| 171 | break; |
| 172 | } |
| 173 | j++; |
| 174 | } |
| 175 | if (j == pattern_length) { |
| 176 | return i; |
| 177 | } |
| 178 | } |
| 179 | return -1; |
| 180 | } |
| 181 | |
| 182 | |
| 183 | RUNTIME_FUNCTION(Runtime_StringLastIndexOf) { |
| 184 | HandleScope scope(isolate); |
| 185 | DCHECK(args.length() == 3); |
| 186 | |
| 187 | CONVERT_ARG_HANDLE_CHECKED(String, sub, 0); |
| 188 | CONVERT_ARG_HANDLE_CHECKED(String, pat, 1); |
| 189 | CONVERT_ARG_HANDLE_CHECKED(Object, index, 2); |
| 190 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 191 | uint32_t start_index = 0; |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 192 | if (!index->ToArrayIndex(&start_index)) return Smi::FromInt(-1); |
| 193 | |
| 194 | uint32_t pat_length = pat->length(); |
| 195 | uint32_t sub_length = sub->length(); |
| 196 | |
| 197 | if (start_index + pat_length > sub_length) { |
| 198 | start_index = sub_length - pat_length; |
| 199 | } |
| 200 | |
| 201 | if (pat_length == 0) { |
| 202 | return Smi::FromInt(start_index); |
| 203 | } |
| 204 | |
| 205 | sub = String::Flatten(sub); |
| 206 | pat = String::Flatten(pat); |
| 207 | |
| 208 | int position = -1; |
| 209 | DisallowHeapAllocation no_gc; // ensure vectors stay valid |
| 210 | |
| 211 | String::FlatContent sub_content = sub->GetFlatContent(); |
| 212 | String::FlatContent pat_content = pat->GetFlatContent(); |
| 213 | |
| 214 | if (pat_content.IsOneByte()) { |
| 215 | Vector<const uint8_t> pat_vector = pat_content.ToOneByteVector(); |
| 216 | if (sub_content.IsOneByte()) { |
| 217 | position = StringMatchBackwards(sub_content.ToOneByteVector(), pat_vector, |
| 218 | start_index); |
| 219 | } else { |
| 220 | position = StringMatchBackwards(sub_content.ToUC16Vector(), pat_vector, |
| 221 | start_index); |
| 222 | } |
| 223 | } else { |
| 224 | Vector<const uc16> pat_vector = pat_content.ToUC16Vector(); |
| 225 | if (sub_content.IsOneByte()) { |
| 226 | position = StringMatchBackwards(sub_content.ToOneByteVector(), pat_vector, |
| 227 | start_index); |
| 228 | } else { |
| 229 | position = StringMatchBackwards(sub_content.ToUC16Vector(), pat_vector, |
| 230 | start_index); |
| 231 | } |
| 232 | } |
| 233 | |
| 234 | return Smi::FromInt(position); |
| 235 | } |
| 236 | |
| 237 | |
| 238 | RUNTIME_FUNCTION(Runtime_StringLocaleCompare) { |
| 239 | HandleScope handle_scope(isolate); |
| 240 | DCHECK(args.length() == 2); |
| 241 | |
| 242 | CONVERT_ARG_HANDLE_CHECKED(String, str1, 0); |
| 243 | CONVERT_ARG_HANDLE_CHECKED(String, str2, 1); |
| 244 | |
| 245 | if (str1.is_identical_to(str2)) return Smi::FromInt(0); // Equal. |
| 246 | int str1_length = str1->length(); |
| 247 | int str2_length = str2->length(); |
| 248 | |
| 249 | // Decide trivial cases without flattening. |
| 250 | if (str1_length == 0) { |
| 251 | if (str2_length == 0) return Smi::FromInt(0); // Equal. |
| 252 | return Smi::FromInt(-str2_length); |
| 253 | } else { |
| 254 | if (str2_length == 0) return Smi::FromInt(str1_length); |
| 255 | } |
| 256 | |
| 257 | int end = str1_length < str2_length ? str1_length : str2_length; |
| 258 | |
| 259 | // No need to flatten if we are going to find the answer on the first |
| 260 | // character. At this point we know there is at least one character |
| 261 | // in each string, due to the trivial case handling above. |
| 262 | int d = str1->Get(0) - str2->Get(0); |
| 263 | if (d != 0) return Smi::FromInt(d); |
| 264 | |
| 265 | str1 = String::Flatten(str1); |
| 266 | str2 = String::Flatten(str2); |
| 267 | |
| 268 | DisallowHeapAllocation no_gc; |
| 269 | String::FlatContent flat1 = str1->GetFlatContent(); |
| 270 | String::FlatContent flat2 = str2->GetFlatContent(); |
| 271 | |
| 272 | for (int i = 0; i < end; i++) { |
| 273 | if (flat1.Get(i) != flat2.Get(i)) { |
| 274 | return Smi::FromInt(flat1.Get(i) - flat2.Get(i)); |
| 275 | } |
| 276 | } |
| 277 | |
| 278 | return Smi::FromInt(str1_length - str2_length); |
| 279 | } |
| 280 | |
| 281 | |
| 282 | RUNTIME_FUNCTION(Runtime_SubString) { |
| 283 | HandleScope scope(isolate); |
| 284 | DCHECK(args.length() == 3); |
| 285 | |
| 286 | CONVERT_ARG_HANDLE_CHECKED(String, string, 0); |
| 287 | int start, end; |
| 288 | // We have a fast integer-only case here to avoid a conversion to double in |
| 289 | // the common case where from and to are Smis. |
| 290 | if (args[1]->IsSmi() && args[2]->IsSmi()) { |
| 291 | CONVERT_SMI_ARG_CHECKED(from_number, 1); |
| 292 | CONVERT_SMI_ARG_CHECKED(to_number, 2); |
| 293 | start = from_number; |
| 294 | end = to_number; |
| 295 | } else { |
| 296 | CONVERT_DOUBLE_ARG_CHECKED(from_number, 1); |
| 297 | CONVERT_DOUBLE_ARG_CHECKED(to_number, 2); |
| 298 | start = FastD2IChecked(from_number); |
| 299 | end = FastD2IChecked(to_number); |
| 300 | } |
| 301 | RUNTIME_ASSERT(end >= start); |
| 302 | RUNTIME_ASSERT(start >= 0); |
| 303 | RUNTIME_ASSERT(end <= string->length()); |
| 304 | isolate->counters()->sub_string_runtime()->Increment(); |
| 305 | |
| 306 | return *isolate->factory()->NewSubString(string, start, end); |
| 307 | } |
| 308 | |
| 309 | |
| 310 | RUNTIME_FUNCTION(Runtime_StringAdd) { |
| 311 | HandleScope scope(isolate); |
| 312 | DCHECK(args.length() == 2); |
| 313 | CONVERT_ARG_HANDLE_CHECKED(String, str1, 0); |
| 314 | CONVERT_ARG_HANDLE_CHECKED(String, str2, 1); |
| 315 | isolate->counters()->string_add_runtime()->Increment(); |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 316 | RETURN_RESULT_OR_FAILURE(isolate, |
| 317 | isolate->factory()->NewConsString(str1, str2)); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 318 | } |
| 319 | |
| 320 | |
| 321 | RUNTIME_FUNCTION(Runtime_InternalizeString) { |
| 322 | HandleScope handles(isolate); |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 323 | DCHECK(args.length() == 1); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 324 | CONVERT_ARG_HANDLE_CHECKED(String, string, 0); |
| 325 | return *isolate->factory()->InternalizeString(string); |
| 326 | } |
| 327 | |
| 328 | |
| 329 | RUNTIME_FUNCTION(Runtime_StringMatch) { |
| 330 | HandleScope handles(isolate); |
| 331 | DCHECK(args.length() == 3); |
| 332 | |
| 333 | CONVERT_ARG_HANDLE_CHECKED(String, subject, 0); |
| 334 | CONVERT_ARG_HANDLE_CHECKED(JSRegExp, regexp, 1); |
| 335 | CONVERT_ARG_HANDLE_CHECKED(JSArray, regexp_info, 2); |
| 336 | |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 337 | CHECK(regexp_info->HasFastObjectElements()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 338 | |
Ben Murdoch | 097c5b2 | 2016-05-18 11:27:45 +0100 | [diff] [blame] | 339 | RegExpImpl::GlobalCache global_cache(regexp, subject, isolate); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 340 | if (global_cache.HasException()) return isolate->heap()->exception(); |
| 341 | |
| 342 | int capture_count = regexp->CaptureCount(); |
| 343 | |
| 344 | ZoneScope zone_scope(isolate->runtime_zone()); |
| 345 | ZoneList<int> offsets(8, zone_scope.zone()); |
| 346 | |
| 347 | while (true) { |
| 348 | int32_t* match = global_cache.FetchNext(); |
| 349 | if (match == NULL) break; |
| 350 | offsets.Add(match[0], zone_scope.zone()); // start |
| 351 | offsets.Add(match[1], zone_scope.zone()); // end |
| 352 | } |
| 353 | |
| 354 | if (global_cache.HasException()) return isolate->heap()->exception(); |
| 355 | |
| 356 | if (offsets.length() == 0) { |
| 357 | // Not a single match. |
| 358 | return isolate->heap()->null_value(); |
| 359 | } |
| 360 | |
| 361 | RegExpImpl::SetLastMatchInfo(regexp_info, subject, capture_count, |
| 362 | global_cache.LastSuccessfulMatch()); |
| 363 | |
| 364 | int matches = offsets.length() / 2; |
| 365 | Handle<FixedArray> elements = isolate->factory()->NewFixedArray(matches); |
| 366 | Handle<String> substring = |
| 367 | isolate->factory()->NewSubString(subject, offsets.at(0), offsets.at(1)); |
| 368 | elements->set(0, *substring); |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 369 | FOR_WITH_HANDLE_SCOPE(isolate, int, i = 1, i, i < matches, i++, { |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 370 | int from = offsets.at(i * 2); |
| 371 | int to = offsets.at(i * 2 + 1); |
| 372 | Handle<String> substring = |
| 373 | isolate->factory()->NewProperSubString(subject, from, to); |
| 374 | elements->set(i, *substring); |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 375 | }); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 376 | Handle<JSArray> result = isolate->factory()->NewJSArrayWithElements(elements); |
| 377 | result->set_length(Smi::FromInt(matches)); |
| 378 | return *result; |
| 379 | } |
| 380 | |
| 381 | |
| 382 | RUNTIME_FUNCTION(Runtime_StringCharCodeAtRT) { |
| 383 | HandleScope handle_scope(isolate); |
| 384 | DCHECK(args.length() == 2); |
| 385 | |
| 386 | CONVERT_ARG_HANDLE_CHECKED(String, subject, 0); |
| 387 | CONVERT_NUMBER_CHECKED(uint32_t, i, Uint32, args[1]); |
| 388 | |
| 389 | // Flatten the string. If someone wants to get a char at an index |
| 390 | // in a cons string, it is likely that more indices will be |
| 391 | // accessed. |
| 392 | subject = String::Flatten(subject); |
| 393 | |
| 394 | if (i >= static_cast<uint32_t>(subject->length())) { |
| 395 | return isolate->heap()->nan_value(); |
| 396 | } |
| 397 | |
| 398 | return Smi::FromInt(subject->Get(i)); |
| 399 | } |
| 400 | |
| 401 | |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 402 | RUNTIME_FUNCTION(Runtime_StringCompare) { |
| 403 | HandleScope handle_scope(isolate); |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 404 | DCHECK_EQ(2, args.length()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 405 | CONVERT_ARG_HANDLE_CHECKED(String, x, 0); |
| 406 | CONVERT_ARG_HANDLE_CHECKED(String, y, 1); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 407 | isolate->counters()->string_compare_runtime()->Increment(); |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 408 | switch (String::Compare(x, y)) { |
| 409 | case ComparisonResult::kLessThan: |
| 410 | return Smi::FromInt(LESS); |
| 411 | case ComparisonResult::kEqual: |
| 412 | return Smi::FromInt(EQUAL); |
| 413 | case ComparisonResult::kGreaterThan: |
| 414 | return Smi::FromInt(GREATER); |
| 415 | case ComparisonResult::kUndefined: |
| 416 | break; |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 417 | } |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 418 | UNREACHABLE(); |
| 419 | return Smi::FromInt(0); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 420 | } |
| 421 | |
| 422 | |
| 423 | RUNTIME_FUNCTION(Runtime_StringBuilderConcat) { |
| 424 | HandleScope scope(isolate); |
| 425 | DCHECK(args.length() == 3); |
| 426 | CONVERT_ARG_HANDLE_CHECKED(JSArray, array, 0); |
| 427 | int32_t array_length; |
| 428 | if (!args[1]->ToInt32(&array_length)) { |
| 429 | THROW_NEW_ERROR_RETURN_FAILURE(isolate, NewInvalidStringLengthError()); |
| 430 | } |
| 431 | CONVERT_ARG_HANDLE_CHECKED(String, special, 2); |
| 432 | |
| 433 | size_t actual_array_length = 0; |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 434 | CHECK(TryNumberToSize(isolate, array->length(), &actual_array_length)); |
| 435 | CHECK(array_length >= 0); |
| 436 | CHECK(static_cast<size_t>(array_length) <= actual_array_length); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 437 | |
| 438 | // This assumption is used by the slice encoding in one or two smis. |
| 439 | DCHECK(Smi::kMaxValue >= String::kMaxLength); |
| 440 | |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 441 | CHECK(array->HasFastElements()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 442 | JSObject::EnsureCanContainHeapObjectElements(array); |
| 443 | |
| 444 | int special_length = special->length(); |
| 445 | if (!array->HasFastObjectElements()) { |
| 446 | return isolate->Throw(isolate->heap()->illegal_argument_string()); |
| 447 | } |
| 448 | |
| 449 | int length; |
| 450 | bool one_byte = special->HasOnlyOneByteChars(); |
| 451 | |
| 452 | { |
| 453 | DisallowHeapAllocation no_gc; |
| 454 | FixedArray* fixed_array = FixedArray::cast(array->elements()); |
| 455 | if (fixed_array->length() < array_length) { |
| 456 | array_length = fixed_array->length(); |
| 457 | } |
| 458 | |
| 459 | if (array_length == 0) { |
| 460 | return isolate->heap()->empty_string(); |
| 461 | } else if (array_length == 1) { |
| 462 | Object* first = fixed_array->get(0); |
| 463 | if (first->IsString()) return first; |
| 464 | } |
| 465 | length = StringBuilderConcatLength(special_length, fixed_array, |
| 466 | array_length, &one_byte); |
| 467 | } |
| 468 | |
| 469 | if (length == -1) { |
| 470 | return isolate->Throw(isolate->heap()->illegal_argument_string()); |
| 471 | } |
| 472 | |
| 473 | if (one_byte) { |
| 474 | Handle<SeqOneByteString> answer; |
| 475 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 476 | isolate, answer, isolate->factory()->NewRawOneByteString(length)); |
| 477 | StringBuilderConcatHelper(*special, answer->GetChars(), |
| 478 | FixedArray::cast(array->elements()), |
| 479 | array_length); |
| 480 | return *answer; |
| 481 | } else { |
| 482 | Handle<SeqTwoByteString> answer; |
| 483 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 484 | isolate, answer, isolate->factory()->NewRawTwoByteString(length)); |
| 485 | StringBuilderConcatHelper(*special, answer->GetChars(), |
| 486 | FixedArray::cast(array->elements()), |
| 487 | array_length); |
| 488 | return *answer; |
| 489 | } |
| 490 | } |
| 491 | |
| 492 | |
| 493 | RUNTIME_FUNCTION(Runtime_StringBuilderJoin) { |
| 494 | HandleScope scope(isolate); |
| 495 | DCHECK(args.length() == 3); |
| 496 | CONVERT_ARG_HANDLE_CHECKED(JSArray, array, 0); |
| 497 | int32_t array_length; |
| 498 | if (!args[1]->ToInt32(&array_length)) { |
| 499 | THROW_NEW_ERROR_RETURN_FAILURE(isolate, NewInvalidStringLengthError()); |
| 500 | } |
| 501 | CONVERT_ARG_HANDLE_CHECKED(String, separator, 2); |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 502 | CHECK(array->HasFastObjectElements()); |
| 503 | CHECK(array_length >= 0); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 504 | |
| 505 | Handle<FixedArray> fixed_array(FixedArray::cast(array->elements())); |
| 506 | if (fixed_array->length() < array_length) { |
| 507 | array_length = fixed_array->length(); |
| 508 | } |
| 509 | |
| 510 | if (array_length == 0) { |
| 511 | return isolate->heap()->empty_string(); |
| 512 | } else if (array_length == 1) { |
| 513 | Object* first = fixed_array->get(0); |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 514 | CHECK(first->IsString()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 515 | return first; |
| 516 | } |
| 517 | |
| 518 | int separator_length = separator->length(); |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 519 | CHECK(separator_length > 0); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 520 | int max_nof_separators = |
| 521 | (String::kMaxLength + separator_length - 1) / separator_length; |
| 522 | if (max_nof_separators < (array_length - 1)) { |
| 523 | THROW_NEW_ERROR_RETURN_FAILURE(isolate, NewInvalidStringLengthError()); |
| 524 | } |
| 525 | int length = (array_length - 1) * separator_length; |
| 526 | for (int i = 0; i < array_length; i++) { |
| 527 | Object* element_obj = fixed_array->get(i); |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 528 | CHECK(element_obj->IsString()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 529 | String* element = String::cast(element_obj); |
| 530 | int increment = element->length(); |
| 531 | if (increment > String::kMaxLength - length) { |
| 532 | STATIC_ASSERT(String::kMaxLength < kMaxInt); |
| 533 | length = kMaxInt; // Provoke exception; |
| 534 | break; |
| 535 | } |
| 536 | length += increment; |
| 537 | } |
| 538 | |
| 539 | Handle<SeqTwoByteString> answer; |
| 540 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 541 | isolate, answer, isolate->factory()->NewRawTwoByteString(length)); |
| 542 | |
| 543 | DisallowHeapAllocation no_gc; |
| 544 | |
| 545 | uc16* sink = answer->GetChars(); |
| 546 | #ifdef DEBUG |
| 547 | uc16* end = sink + length; |
| 548 | #endif |
| 549 | |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 550 | CHECK(fixed_array->get(0)->IsString()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 551 | String* first = String::cast(fixed_array->get(0)); |
| 552 | String* separator_raw = *separator; |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 553 | |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 554 | int first_length = first->length(); |
| 555 | String::WriteToFlat(first, sink, 0, first_length); |
| 556 | sink += first_length; |
| 557 | |
| 558 | for (int i = 1; i < array_length; i++) { |
| 559 | DCHECK(sink + separator_length <= end); |
| 560 | String::WriteToFlat(separator_raw, sink, 0, separator_length); |
| 561 | sink += separator_length; |
| 562 | |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 563 | CHECK(fixed_array->get(i)->IsString()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 564 | String* element = String::cast(fixed_array->get(i)); |
| 565 | int element_length = element->length(); |
| 566 | DCHECK(sink + element_length <= end); |
| 567 | String::WriteToFlat(element, sink, 0, element_length); |
| 568 | sink += element_length; |
| 569 | } |
| 570 | DCHECK(sink == end); |
| 571 | |
| 572 | // Use %_FastOneByteArrayJoin instead. |
| 573 | DCHECK(!answer->IsOneByteRepresentation()); |
| 574 | return *answer; |
| 575 | } |
| 576 | |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 577 | template <typename sinkchar> |
| 578 | static void WriteRepeatToFlat(String* src, Vector<sinkchar> buffer, int cursor, |
| 579 | int repeat, int length) { |
| 580 | if (repeat == 0) return; |
| 581 | |
| 582 | sinkchar* start = &buffer[cursor]; |
| 583 | String::WriteToFlat<sinkchar>(src, start, 0, length); |
| 584 | |
| 585 | int done = 1; |
| 586 | sinkchar* next = start + length; |
| 587 | |
| 588 | while (done < repeat) { |
| 589 | int block = Min(done, repeat - done); |
| 590 | int block_chars = block * length; |
| 591 | CopyChars(next, start, block_chars); |
| 592 | next += block_chars; |
| 593 | done += block; |
| 594 | } |
| 595 | } |
| 596 | |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 597 | template <typename Char> |
| 598 | static void JoinSparseArrayWithSeparator(FixedArray* elements, |
| 599 | int elements_length, |
| 600 | uint32_t array_length, |
| 601 | String* separator, |
| 602 | Vector<Char> buffer) { |
| 603 | DisallowHeapAllocation no_gc; |
| 604 | int previous_separator_position = 0; |
| 605 | int separator_length = separator->length(); |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 606 | DCHECK_LT(0, separator_length); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 607 | int cursor = 0; |
| 608 | for (int i = 0; i < elements_length; i += 2) { |
| 609 | int position = NumberToInt32(elements->get(i)); |
| 610 | String* string = String::cast(elements->get(i + 1)); |
| 611 | int string_length = string->length(); |
| 612 | if (string->length() > 0) { |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 613 | int repeat = position - previous_separator_position; |
| 614 | WriteRepeatToFlat<Char>(separator, buffer, cursor, repeat, |
| 615 | separator_length); |
| 616 | cursor += repeat * separator_length; |
| 617 | previous_separator_position = position; |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 618 | String::WriteToFlat<Char>(string, &buffer[cursor], 0, string_length); |
| 619 | cursor += string->length(); |
| 620 | } |
| 621 | } |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 622 | |
| 623 | int last_array_index = static_cast<int>(array_length - 1); |
| 624 | // Array length must be representable as a signed 32-bit number, |
| 625 | // otherwise the total string length would have been too large. |
| 626 | DCHECK(array_length <= 0x7fffffff); // Is int32_t. |
| 627 | int repeat = last_array_index - previous_separator_position; |
| 628 | WriteRepeatToFlat<Char>(separator, buffer, cursor, repeat, separator_length); |
| 629 | cursor += repeat * separator_length; |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 630 | DCHECK(cursor <= buffer.length()); |
| 631 | } |
| 632 | |
| 633 | |
| 634 | RUNTIME_FUNCTION(Runtime_SparseJoinWithSeparator) { |
| 635 | HandleScope scope(isolate); |
| 636 | DCHECK(args.length() == 3); |
| 637 | CONVERT_ARG_HANDLE_CHECKED(JSArray, elements_array, 0); |
| 638 | CONVERT_NUMBER_CHECKED(uint32_t, array_length, Uint32, args[1]); |
| 639 | CONVERT_ARG_HANDLE_CHECKED(String, separator, 2); |
| 640 | // elements_array is fast-mode JSarray of alternating positions |
| 641 | // (increasing order) and strings. |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 642 | CHECK(elements_array->HasFastSmiOrObjectElements()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 643 | // array_length is length of original array (used to add separators); |
| 644 | // separator is string to put between elements. Assumed to be non-empty. |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 645 | CHECK(array_length > 0); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 646 | |
| 647 | // Find total length of join result. |
| 648 | int string_length = 0; |
| 649 | bool is_one_byte = separator->IsOneByteRepresentation(); |
| 650 | bool overflow = false; |
| 651 | CONVERT_NUMBER_CHECKED(int, elements_length, Int32, elements_array->length()); |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 652 | CHECK(elements_length <= elements_array->elements()->length()); |
| 653 | CHECK((elements_length & 1) == 0); // Even length. |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 654 | FixedArray* elements = FixedArray::cast(elements_array->elements()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 655 | { |
| 656 | DisallowHeapAllocation no_gc; |
| 657 | for (int i = 0; i < elements_length; i += 2) { |
| 658 | String* string = String::cast(elements->get(i + 1)); |
| 659 | int length = string->length(); |
| 660 | if (is_one_byte && !string->IsOneByteRepresentation()) { |
| 661 | is_one_byte = false; |
| 662 | } |
| 663 | if (length > String::kMaxLength || |
| 664 | String::kMaxLength - length < string_length) { |
| 665 | overflow = true; |
| 666 | break; |
| 667 | } |
| 668 | string_length += length; |
| 669 | } |
| 670 | } |
| 671 | |
| 672 | int separator_length = separator->length(); |
| 673 | if (!overflow && separator_length > 0) { |
| 674 | if (array_length <= 0x7fffffffu) { |
| 675 | int separator_count = static_cast<int>(array_length) - 1; |
| 676 | int remaining_length = String::kMaxLength - string_length; |
| 677 | if ((remaining_length / separator_length) >= separator_count) { |
| 678 | string_length += separator_length * (array_length - 1); |
| 679 | } else { |
| 680 | // Not room for the separators within the maximal string length. |
| 681 | overflow = true; |
| 682 | } |
| 683 | } else { |
| 684 | // Nonempty separator and at least 2^31-1 separators necessary |
| 685 | // means that the string is too large to create. |
| 686 | STATIC_ASSERT(String::kMaxLength < 0x7fffffff); |
| 687 | overflow = true; |
| 688 | } |
| 689 | } |
| 690 | if (overflow) { |
| 691 | // Throw an exception if the resulting string is too large. See |
| 692 | // https://code.google.com/p/chromium/issues/detail?id=336820 |
| 693 | // for details. |
| 694 | THROW_NEW_ERROR_RETURN_FAILURE(isolate, NewInvalidStringLengthError()); |
| 695 | } |
| 696 | |
| 697 | if (is_one_byte) { |
| 698 | Handle<SeqOneByteString> result = isolate->factory() |
| 699 | ->NewRawOneByteString(string_length) |
| 700 | .ToHandleChecked(); |
| 701 | JoinSparseArrayWithSeparator<uint8_t>( |
| 702 | FixedArray::cast(elements_array->elements()), elements_length, |
| 703 | array_length, *separator, |
| 704 | Vector<uint8_t>(result->GetChars(), string_length)); |
| 705 | return *result; |
| 706 | } else { |
| 707 | Handle<SeqTwoByteString> result = isolate->factory() |
| 708 | ->NewRawTwoByteString(string_length) |
| 709 | .ToHandleChecked(); |
| 710 | JoinSparseArrayWithSeparator<uc16>( |
| 711 | FixedArray::cast(elements_array->elements()), elements_length, |
| 712 | array_length, *separator, |
| 713 | Vector<uc16>(result->GetChars(), string_length)); |
| 714 | return *result; |
| 715 | } |
| 716 | } |
| 717 | |
| 718 | |
| 719 | // Copies Latin1 characters to the given fixed array looking up |
| 720 | // one-char strings in the cache. Gives up on the first char that is |
| 721 | // not in the cache and fills the remainder with smi zeros. Returns |
| 722 | // the length of the successfully copied prefix. |
| 723 | static int CopyCachedOneByteCharsToArray(Heap* heap, const uint8_t* chars, |
| 724 | FixedArray* elements, int length) { |
| 725 | DisallowHeapAllocation no_gc; |
| 726 | FixedArray* one_byte_cache = heap->single_character_string_cache(); |
| 727 | Object* undefined = heap->undefined_value(); |
| 728 | int i; |
| 729 | WriteBarrierMode mode = elements->GetWriteBarrierMode(no_gc); |
| 730 | for (i = 0; i < length; ++i) { |
| 731 | Object* value = one_byte_cache->get(chars[i]); |
| 732 | if (value == undefined) break; |
| 733 | elements->set(i, value, mode); |
| 734 | } |
| 735 | if (i < length) { |
| 736 | DCHECK(Smi::FromInt(0) == 0); |
| 737 | memset(elements->data_start() + i, 0, kPointerSize * (length - i)); |
| 738 | } |
| 739 | #ifdef DEBUG |
| 740 | for (int j = 0; j < length; ++j) { |
| 741 | Object* element = elements->get(j); |
| 742 | DCHECK(element == Smi::FromInt(0) || |
| 743 | (element->IsString() && String::cast(element)->LooksValid())); |
| 744 | } |
| 745 | #endif |
| 746 | return i; |
| 747 | } |
| 748 | |
| 749 | |
| 750 | // Converts a String to JSArray. |
| 751 | // For example, "foo" => ["f", "o", "o"]. |
| 752 | RUNTIME_FUNCTION(Runtime_StringToArray) { |
| 753 | HandleScope scope(isolate); |
| 754 | DCHECK(args.length() == 2); |
| 755 | CONVERT_ARG_HANDLE_CHECKED(String, s, 0); |
| 756 | CONVERT_NUMBER_CHECKED(uint32_t, limit, Uint32, args[1]); |
| 757 | |
| 758 | s = String::Flatten(s); |
| 759 | const int length = static_cast<int>(Min<uint32_t>(s->length(), limit)); |
| 760 | |
| 761 | Handle<FixedArray> elements; |
| 762 | int position = 0; |
| 763 | if (s->IsFlat() && s->IsOneByteRepresentation()) { |
| 764 | // Try using cached chars where possible. |
| 765 | elements = isolate->factory()->NewUninitializedFixedArray(length); |
| 766 | |
| 767 | DisallowHeapAllocation no_gc; |
| 768 | String::FlatContent content = s->GetFlatContent(); |
| 769 | if (content.IsOneByte()) { |
| 770 | Vector<const uint8_t> chars = content.ToOneByteVector(); |
| 771 | // Note, this will initialize all elements (not only the prefix) |
| 772 | // to prevent GC from seeing partially initialized array. |
| 773 | position = CopyCachedOneByteCharsToArray(isolate->heap(), chars.start(), |
| 774 | *elements, length); |
| 775 | } else { |
| 776 | MemsetPointer(elements->data_start(), isolate->heap()->undefined_value(), |
| 777 | length); |
| 778 | } |
| 779 | } else { |
| 780 | elements = isolate->factory()->NewFixedArray(length); |
| 781 | } |
| 782 | for (int i = position; i < length; ++i) { |
| 783 | Handle<Object> str = |
| 784 | isolate->factory()->LookupSingleCharacterStringFromCode(s->Get(i)); |
| 785 | elements->set(i, *str); |
| 786 | } |
| 787 | |
| 788 | #ifdef DEBUG |
| 789 | for (int i = 0; i < length; ++i) { |
| 790 | DCHECK(String::cast(elements->get(i))->length() == 1); |
| 791 | } |
| 792 | #endif |
| 793 | |
| 794 | return *isolate->factory()->NewJSArrayWithElements(elements); |
| 795 | } |
| 796 | |
| 797 | |
| 798 | static inline bool ToUpperOverflows(uc32 character) { |
| 799 | // y with umlauts and the micro sign are the only characters that stop |
| 800 | // fitting into one-byte when converting to uppercase. |
| 801 | static const uc32 yuml_code = 0xff; |
| 802 | static const uc32 micro_code = 0xb5; |
| 803 | return (character == yuml_code || character == micro_code); |
| 804 | } |
| 805 | |
| 806 | |
| 807 | template <class Converter> |
| 808 | MUST_USE_RESULT static Object* ConvertCaseHelper( |
| 809 | Isolate* isolate, String* string, SeqString* result, int result_length, |
| 810 | unibrow::Mapping<Converter, 128>* mapping) { |
| 811 | DisallowHeapAllocation no_gc; |
| 812 | // We try this twice, once with the assumption that the result is no longer |
| 813 | // than the input and, if that assumption breaks, again with the exact |
| 814 | // length. This may not be pretty, but it is nicer than what was here before |
| 815 | // and I hereby claim my vaffel-is. |
| 816 | // |
| 817 | // NOTE: This assumes that the upper/lower case of an ASCII |
| 818 | // character is also ASCII. This is currently the case, but it |
| 819 | // might break in the future if we implement more context and locale |
| 820 | // dependent upper/lower conversions. |
| 821 | bool has_changed_character = false; |
| 822 | |
| 823 | // Convert all characters to upper case, assuming that they will fit |
| 824 | // in the buffer |
| 825 | StringCharacterStream stream(string); |
| 826 | unibrow::uchar chars[Converter::kMaxWidth]; |
| 827 | // We can assume that the string is not empty |
| 828 | uc32 current = stream.GetNext(); |
| 829 | bool ignore_overflow = Converter::kIsToLower || result->IsSeqTwoByteString(); |
| 830 | for (int i = 0; i < result_length;) { |
| 831 | bool has_next = stream.HasMore(); |
| 832 | uc32 next = has_next ? stream.GetNext() : 0; |
| 833 | int char_length = mapping->get(current, next, chars); |
| 834 | if (char_length == 0) { |
| 835 | // The case conversion of this character is the character itself. |
| 836 | result->Set(i, current); |
| 837 | i++; |
| 838 | } else if (char_length == 1 && |
| 839 | (ignore_overflow || !ToUpperOverflows(current))) { |
| 840 | // Common case: converting the letter resulted in one character. |
| 841 | DCHECK(static_cast<uc32>(chars[0]) != current); |
| 842 | result->Set(i, chars[0]); |
| 843 | has_changed_character = true; |
| 844 | i++; |
| 845 | } else if (result_length == string->length()) { |
| 846 | bool overflows = ToUpperOverflows(current); |
| 847 | // We've assumed that the result would be as long as the |
| 848 | // input but here is a character that converts to several |
| 849 | // characters. No matter, we calculate the exact length |
| 850 | // of the result and try the whole thing again. |
| 851 | // |
| 852 | // Note that this leaves room for optimization. We could just |
| 853 | // memcpy what we already have to the result string. Also, |
| 854 | // the result string is the last object allocated we could |
| 855 | // "realloc" it and probably, in the vast majority of cases, |
| 856 | // extend the existing string to be able to hold the full |
| 857 | // result. |
| 858 | int next_length = 0; |
| 859 | if (has_next) { |
| 860 | next_length = mapping->get(next, 0, chars); |
| 861 | if (next_length == 0) next_length = 1; |
| 862 | } |
| 863 | int current_length = i + char_length + next_length; |
| 864 | while (stream.HasMore()) { |
| 865 | current = stream.GetNext(); |
| 866 | overflows |= ToUpperOverflows(current); |
| 867 | // NOTE: we use 0 as the next character here because, while |
| 868 | // the next character may affect what a character converts to, |
| 869 | // it does not in any case affect the length of what it convert |
| 870 | // to. |
| 871 | int char_length = mapping->get(current, 0, chars); |
| 872 | if (char_length == 0) char_length = 1; |
| 873 | current_length += char_length; |
| 874 | if (current_length > String::kMaxLength) { |
| 875 | AllowHeapAllocation allocate_error_and_return; |
| 876 | THROW_NEW_ERROR_RETURN_FAILURE(isolate, |
| 877 | NewInvalidStringLengthError()); |
| 878 | } |
| 879 | } |
| 880 | // Try again with the real length. Return signed if we need |
| 881 | // to allocate a two-byte string for to uppercase. |
| 882 | return (overflows && !ignore_overflow) ? Smi::FromInt(-current_length) |
| 883 | : Smi::FromInt(current_length); |
| 884 | } else { |
| 885 | for (int j = 0; j < char_length; j++) { |
| 886 | result->Set(i, chars[j]); |
| 887 | i++; |
| 888 | } |
| 889 | has_changed_character = true; |
| 890 | } |
| 891 | current = next; |
| 892 | } |
| 893 | if (has_changed_character) { |
| 894 | return result; |
| 895 | } else { |
| 896 | // If we didn't actually change anything in doing the conversion |
| 897 | // we simple return the result and let the converted string |
| 898 | // become garbage; there is no reason to keep two identical strings |
| 899 | // alive. |
| 900 | return string; |
| 901 | } |
| 902 | } |
| 903 | |
| 904 | |
| 905 | static const uintptr_t kOneInEveryByte = kUintptrAllBitsSet / 0xFF; |
| 906 | static const uintptr_t kAsciiMask = kOneInEveryByte << 7; |
| 907 | |
| 908 | // Given a word and two range boundaries returns a word with high bit |
| 909 | // set in every byte iff the corresponding input byte was strictly in |
| 910 | // the range (m, n). All the other bits in the result are cleared. |
| 911 | // This function is only useful when it can be inlined and the |
| 912 | // boundaries are statically known. |
| 913 | // Requires: all bytes in the input word and the boundaries must be |
| 914 | // ASCII (less than 0x7F). |
| 915 | static inline uintptr_t AsciiRangeMask(uintptr_t w, char m, char n) { |
| 916 | // Use strict inequalities since in edge cases the function could be |
| 917 | // further simplified. |
| 918 | DCHECK(0 < m && m < n); |
| 919 | // Has high bit set in every w byte less than n. |
| 920 | uintptr_t tmp1 = kOneInEveryByte * (0x7F + n) - w; |
| 921 | // Has high bit set in every w byte greater than m. |
| 922 | uintptr_t tmp2 = w + kOneInEveryByte * (0x7F - m); |
| 923 | return (tmp1 & tmp2 & (kOneInEveryByte * 0x80)); |
| 924 | } |
| 925 | |
| 926 | |
| 927 | #ifdef DEBUG |
| 928 | static bool CheckFastAsciiConvert(char* dst, const char* src, int length, |
| 929 | bool changed, bool is_to_lower) { |
| 930 | bool expected_changed = false; |
| 931 | for (int i = 0; i < length; i++) { |
| 932 | if (dst[i] == src[i]) continue; |
| 933 | expected_changed = true; |
| 934 | if (is_to_lower) { |
| 935 | DCHECK('A' <= src[i] && src[i] <= 'Z'); |
| 936 | DCHECK(dst[i] == src[i] + ('a' - 'A')); |
| 937 | } else { |
| 938 | DCHECK('a' <= src[i] && src[i] <= 'z'); |
| 939 | DCHECK(dst[i] == src[i] - ('a' - 'A')); |
| 940 | } |
| 941 | } |
| 942 | return (expected_changed == changed); |
| 943 | } |
| 944 | #endif |
| 945 | |
| 946 | |
| 947 | template <class Converter> |
| 948 | static bool FastAsciiConvert(char* dst, const char* src, int length, |
| 949 | bool* changed_out) { |
| 950 | #ifdef DEBUG |
| 951 | char* saved_dst = dst; |
| 952 | const char* saved_src = src; |
| 953 | #endif |
| 954 | DisallowHeapAllocation no_gc; |
| 955 | // We rely on the distance between upper and lower case letters |
| 956 | // being a known power of 2. |
| 957 | DCHECK('a' - 'A' == (1 << 5)); |
| 958 | // Boundaries for the range of input characters than require conversion. |
| 959 | static const char lo = Converter::kIsToLower ? 'A' - 1 : 'a' - 1; |
| 960 | static const char hi = Converter::kIsToLower ? 'Z' + 1 : 'z' + 1; |
| 961 | bool changed = false; |
| 962 | uintptr_t or_acc = 0; |
| 963 | const char* const limit = src + length; |
| 964 | |
| 965 | // dst is newly allocated and always aligned. |
| 966 | DCHECK(IsAligned(reinterpret_cast<intptr_t>(dst), sizeof(uintptr_t))); |
| 967 | // Only attempt processing one word at a time if src is also aligned. |
| 968 | if (IsAligned(reinterpret_cast<intptr_t>(src), sizeof(uintptr_t))) { |
| 969 | // Process the prefix of the input that requires no conversion one aligned |
| 970 | // (machine) word at a time. |
| 971 | while (src <= limit - sizeof(uintptr_t)) { |
| 972 | const uintptr_t w = *reinterpret_cast<const uintptr_t*>(src); |
| 973 | or_acc |= w; |
| 974 | if (AsciiRangeMask(w, lo, hi) != 0) { |
| 975 | changed = true; |
| 976 | break; |
| 977 | } |
| 978 | *reinterpret_cast<uintptr_t*>(dst) = w; |
| 979 | src += sizeof(uintptr_t); |
| 980 | dst += sizeof(uintptr_t); |
| 981 | } |
| 982 | // Process the remainder of the input performing conversion when |
| 983 | // required one word at a time. |
| 984 | while (src <= limit - sizeof(uintptr_t)) { |
| 985 | const uintptr_t w = *reinterpret_cast<const uintptr_t*>(src); |
| 986 | or_acc |= w; |
| 987 | uintptr_t m = AsciiRangeMask(w, lo, hi); |
| 988 | // The mask has high (7th) bit set in every byte that needs |
| 989 | // conversion and we know that the distance between cases is |
| 990 | // 1 << 5. |
| 991 | *reinterpret_cast<uintptr_t*>(dst) = w ^ (m >> 2); |
| 992 | src += sizeof(uintptr_t); |
| 993 | dst += sizeof(uintptr_t); |
| 994 | } |
| 995 | } |
| 996 | // Process the last few bytes of the input (or the whole input if |
| 997 | // unaligned access is not supported). |
| 998 | while (src < limit) { |
| 999 | char c = *src; |
| 1000 | or_acc |= c; |
| 1001 | if (lo < c && c < hi) { |
| 1002 | c ^= (1 << 5); |
| 1003 | changed = true; |
| 1004 | } |
| 1005 | *dst = c; |
| 1006 | ++src; |
| 1007 | ++dst; |
| 1008 | } |
| 1009 | |
| 1010 | if ((or_acc & kAsciiMask) != 0) return false; |
| 1011 | |
| 1012 | DCHECK(CheckFastAsciiConvert(saved_dst, saved_src, length, changed, |
| 1013 | Converter::kIsToLower)); |
| 1014 | |
| 1015 | *changed_out = changed; |
| 1016 | return true; |
| 1017 | } |
| 1018 | |
| 1019 | |
| 1020 | template <class Converter> |
| 1021 | MUST_USE_RESULT static Object* ConvertCase( |
| 1022 | Handle<String> s, Isolate* isolate, |
| 1023 | unibrow::Mapping<Converter, 128>* mapping) { |
| 1024 | s = String::Flatten(s); |
| 1025 | int length = s->length(); |
| 1026 | // Assume that the string is not empty; we need this assumption later |
| 1027 | if (length == 0) return *s; |
| 1028 | |
| 1029 | // Simpler handling of ASCII strings. |
| 1030 | // |
| 1031 | // NOTE: This assumes that the upper/lower case of an ASCII |
| 1032 | // character is also ASCII. This is currently the case, but it |
| 1033 | // might break in the future if we implement more context and locale |
| 1034 | // dependent upper/lower conversions. |
| 1035 | if (s->IsOneByteRepresentationUnderneath()) { |
| 1036 | // Same length as input. |
| 1037 | Handle<SeqOneByteString> result = |
| 1038 | isolate->factory()->NewRawOneByteString(length).ToHandleChecked(); |
| 1039 | DisallowHeapAllocation no_gc; |
| 1040 | String::FlatContent flat_content = s->GetFlatContent(); |
| 1041 | DCHECK(flat_content.IsFlat()); |
| 1042 | bool has_changed_character = false; |
| 1043 | bool is_ascii = FastAsciiConvert<Converter>( |
| 1044 | reinterpret_cast<char*>(result->GetChars()), |
| 1045 | reinterpret_cast<const char*>(flat_content.ToOneByteVector().start()), |
| 1046 | length, &has_changed_character); |
| 1047 | // If not ASCII, we discard the result and take the 2 byte path. |
| 1048 | if (is_ascii) return has_changed_character ? *result : *s; |
| 1049 | } |
| 1050 | |
| 1051 | Handle<SeqString> result; // Same length as input. |
| 1052 | if (s->IsOneByteRepresentation()) { |
| 1053 | result = isolate->factory()->NewRawOneByteString(length).ToHandleChecked(); |
| 1054 | } else { |
| 1055 | result = isolate->factory()->NewRawTwoByteString(length).ToHandleChecked(); |
| 1056 | } |
| 1057 | |
| 1058 | Object* answer = ConvertCaseHelper(isolate, *s, *result, length, mapping); |
Ben Murdoch | 61f157c | 2016-09-16 13:49:30 +0100 | [diff] [blame^] | 1059 | if (answer->IsException(isolate) || answer->IsString()) return answer; |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1060 | |
| 1061 | DCHECK(answer->IsSmi()); |
| 1062 | length = Smi::cast(answer)->value(); |
| 1063 | if (s->IsOneByteRepresentation() && length > 0) { |
| 1064 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 1065 | isolate, result, isolate->factory()->NewRawOneByteString(length)); |
| 1066 | } else { |
| 1067 | if (length < 0) length = -length; |
| 1068 | ASSIGN_RETURN_FAILURE_ON_EXCEPTION( |
| 1069 | isolate, result, isolate->factory()->NewRawTwoByteString(length)); |
| 1070 | } |
| 1071 | return ConvertCaseHelper(isolate, *s, *result, length, mapping); |
| 1072 | } |
| 1073 | |
| 1074 | |
| 1075 | RUNTIME_FUNCTION(Runtime_StringToLowerCase) { |
| 1076 | HandleScope scope(isolate); |
Ben Murdoch | c561043 | 2016-08-08 18:44:38 +0100 | [diff] [blame] | 1077 | DCHECK_EQ(args.length(), 1); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1078 | CONVERT_ARG_HANDLE_CHECKED(String, s, 0); |
| 1079 | return ConvertCase(s, isolate, isolate->runtime_state()->to_lower_mapping()); |
| 1080 | } |
| 1081 | |
| 1082 | |
| 1083 | RUNTIME_FUNCTION(Runtime_StringToUpperCase) { |
| 1084 | HandleScope scope(isolate); |
Ben Murdoch | c561043 | 2016-08-08 18:44:38 +0100 | [diff] [blame] | 1085 | DCHECK_EQ(args.length(), 1); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1086 | CONVERT_ARG_HANDLE_CHECKED(String, s, 0); |
| 1087 | return ConvertCase(s, isolate, isolate->runtime_state()->to_upper_mapping()); |
| 1088 | } |
| 1089 | |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 1090 | RUNTIME_FUNCTION(Runtime_StringLessThan) { |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1091 | HandleScope handle_scope(isolate); |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 1092 | DCHECK_EQ(2, args.length()); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1093 | CONVERT_ARG_HANDLE_CHECKED(String, x, 0); |
| 1094 | CONVERT_ARG_HANDLE_CHECKED(String, y, 1); |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 1095 | switch (String::Compare(x, y)) { |
| 1096 | case ComparisonResult::kLessThan: |
| 1097 | return isolate->heap()->true_value(); |
| 1098 | case ComparisonResult::kEqual: |
| 1099 | case ComparisonResult::kGreaterThan: |
| 1100 | return isolate->heap()->false_value(); |
| 1101 | case ComparisonResult::kUndefined: |
| 1102 | break; |
| 1103 | } |
| 1104 | UNREACHABLE(); |
| 1105 | return Smi::FromInt(0); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1106 | } |
| 1107 | |
Ben Murdoch | da12d29 | 2016-06-02 14:46:10 +0100 | [diff] [blame] | 1108 | RUNTIME_FUNCTION(Runtime_StringLessThanOrEqual) { |
| 1109 | HandleScope handle_scope(isolate); |
| 1110 | DCHECK_EQ(2, args.length()); |
| 1111 | CONVERT_ARG_HANDLE_CHECKED(String, x, 0); |
| 1112 | CONVERT_ARG_HANDLE_CHECKED(String, y, 1); |
| 1113 | switch (String::Compare(x, y)) { |
| 1114 | case ComparisonResult::kEqual: |
| 1115 | case ComparisonResult::kLessThan: |
| 1116 | return isolate->heap()->true_value(); |
| 1117 | case ComparisonResult::kGreaterThan: |
| 1118 | return isolate->heap()->false_value(); |
| 1119 | case ComparisonResult::kUndefined: |
| 1120 | break; |
| 1121 | } |
| 1122 | UNREACHABLE(); |
| 1123 | return Smi::FromInt(0); |
| 1124 | } |
| 1125 | |
| 1126 | RUNTIME_FUNCTION(Runtime_StringGreaterThan) { |
| 1127 | HandleScope handle_scope(isolate); |
| 1128 | DCHECK_EQ(2, args.length()); |
| 1129 | CONVERT_ARG_HANDLE_CHECKED(String, x, 0); |
| 1130 | CONVERT_ARG_HANDLE_CHECKED(String, y, 1); |
| 1131 | switch (String::Compare(x, y)) { |
| 1132 | case ComparisonResult::kGreaterThan: |
| 1133 | return isolate->heap()->true_value(); |
| 1134 | case ComparisonResult::kEqual: |
| 1135 | case ComparisonResult::kLessThan: |
| 1136 | return isolate->heap()->false_value(); |
| 1137 | case ComparisonResult::kUndefined: |
| 1138 | break; |
| 1139 | } |
| 1140 | UNREACHABLE(); |
| 1141 | return Smi::FromInt(0); |
| 1142 | } |
| 1143 | |
| 1144 | RUNTIME_FUNCTION(Runtime_StringGreaterThanOrEqual) { |
| 1145 | HandleScope handle_scope(isolate); |
| 1146 | DCHECK_EQ(2, args.length()); |
| 1147 | CONVERT_ARG_HANDLE_CHECKED(String, x, 0); |
| 1148 | CONVERT_ARG_HANDLE_CHECKED(String, y, 1); |
| 1149 | switch (String::Compare(x, y)) { |
| 1150 | case ComparisonResult::kEqual: |
| 1151 | case ComparisonResult::kGreaterThan: |
| 1152 | return isolate->heap()->true_value(); |
| 1153 | case ComparisonResult::kLessThan: |
| 1154 | return isolate->heap()->false_value(); |
| 1155 | case ComparisonResult::kUndefined: |
| 1156 | break; |
| 1157 | } |
| 1158 | UNREACHABLE(); |
| 1159 | return Smi::FromInt(0); |
| 1160 | } |
| 1161 | |
| 1162 | RUNTIME_FUNCTION(Runtime_StringEqual) { |
| 1163 | HandleScope handle_scope(isolate); |
| 1164 | DCHECK_EQ(2, args.length()); |
| 1165 | CONVERT_ARG_HANDLE_CHECKED(String, x, 0); |
| 1166 | CONVERT_ARG_HANDLE_CHECKED(String, y, 1); |
| 1167 | return isolate->heap()->ToBoolean(String::Equals(x, y)); |
| 1168 | } |
| 1169 | |
| 1170 | RUNTIME_FUNCTION(Runtime_StringNotEqual) { |
| 1171 | HandleScope handle_scope(isolate); |
| 1172 | DCHECK_EQ(2, args.length()); |
| 1173 | CONVERT_ARG_HANDLE_CHECKED(String, x, 0); |
| 1174 | CONVERT_ARG_HANDLE_CHECKED(String, y, 1); |
| 1175 | return isolate->heap()->ToBoolean(!String::Equals(x, y)); |
| 1176 | } |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1177 | |
| 1178 | RUNTIME_FUNCTION(Runtime_FlattenString) { |
| 1179 | HandleScope scope(isolate); |
| 1180 | DCHECK(args.length() == 1); |
| 1181 | CONVERT_ARG_HANDLE_CHECKED(String, str, 0); |
| 1182 | return *String::Flatten(str); |
| 1183 | } |
| 1184 | |
| 1185 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1186 | RUNTIME_FUNCTION(Runtime_StringCharFromCode) { |
| 1187 | HandleScope handlescope(isolate); |
| 1188 | DCHECK_EQ(1, args.length()); |
| 1189 | if (args[0]->IsNumber()) { |
| 1190 | CONVERT_NUMBER_CHECKED(uint32_t, code, Uint32, args[0]); |
| 1191 | code &= 0xffff; |
| 1192 | return *isolate->factory()->LookupSingleCharacterStringFromCode(code); |
| 1193 | } |
| 1194 | return isolate->heap()->empty_string(); |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1195 | } |
| 1196 | |
Ben Murdoch | c561043 | 2016-08-08 18:44:38 +0100 | [diff] [blame] | 1197 | RUNTIME_FUNCTION(Runtime_ExternalStringGetChar) { |
| 1198 | SealHandleScope shs(isolate); |
| 1199 | DCHECK_EQ(2, args.length()); |
| 1200 | CONVERT_ARG_CHECKED(ExternalString, string, 0); |
| 1201 | CONVERT_INT32_ARG_CHECKED(index, 1); |
| 1202 | return Smi::FromInt(string->Get(index)); |
| 1203 | } |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1204 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1205 | RUNTIME_FUNCTION(Runtime_StringCharCodeAt) { |
Emily Bernier | d0a1eb7 | 2015-03-24 16:35:39 -0400 | [diff] [blame] | 1206 | SealHandleScope shs(isolate); |
| 1207 | DCHECK(args.length() == 2); |
| 1208 | if (!args[0]->IsString()) return isolate->heap()->undefined_value(); |
| 1209 | if (!args[1]->IsNumber()) return isolate->heap()->undefined_value(); |
| 1210 | if (std::isinf(args.number_at(1))) return isolate->heap()->nan_value(); |
| 1211 | return __RT_impl_Runtime_StringCharCodeAtRT(args, isolate); |
| 1212 | } |
| 1213 | |
Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame] | 1214 | } // namespace internal |
| 1215 | } // namespace v8 |