builtins-string.cc | Explore in Territory

// Copyright 2016 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

#include <optional>

#include "src/builtins/builtins-utils-inl.h"
#include "src/builtins/builtins.h"
#include "src/heap/heap-inl.h"  // For ToBoolean. TODO(jkummerow): Drop.
#include "src/logging/counters.h"
#include "src/numbers/conversions.h"
#include "src/objects/objects-inl.h"
#ifdef V8_INTL_SUPPORT
#include "src/objects/intl-objects.h"
#endif
#include "src/base/strings.h"
#include "src/regexp/regexp-utils.h"
#include "src/strings/string-builder-inl.h"
#include "src/strings/string-case.h"
#include "src/strings/unicode-inl.h"
#include "src/strings/unicode.h"

namespace v8 {
namespace internal {

namespace {  // for String.fromCodePoint

bool IsValidCodePoint(Isolate* isolate, Handle<Object> value) { … }

static constexpr base::uc32 kInvalidCodePoint = …;

base::uc32 NextCodePoint(Isolate* isolate, BuiltinArguments args, int index) { … }

}  // namespace

// ES6 section 21.1.2.2 String.fromCodePoint ( ...codePoints )
BUILTIN(StringFromCodePoint) { … }

// ES6 section 21.1.3.9
// String.prototype.lastIndexOf ( searchString [ , position ] )
BUILTIN(StringPrototypeLastIndexOf) { … }

// ES6 section 21.1.3.10 String.prototype.localeCompare ( that )
//
// This function is implementation specific.  For now, we do not
// do anything locale specific.
BUILTIN(StringPrototypeLocaleCompare) { … }

#ifndef V8_INTL_SUPPORT
// ES6 section 21.1.3.12 String.prototype.normalize ( [form] )
//
// Simply checks the argument is valid and returns the string itself.
// If internationalization is enabled, then intl.js will override this function
// and provide the proper functionality, so this is just a fallback.
BUILTIN(StringPrototypeNormalize) {
  HandleScope handle_scope(isolate);
  TO_THIS_STRING(string, "String.prototype.normalize");

  Handle<Object> form_input = args.atOrUndefined(isolate, 1);
  if (IsUndefined(*form_input, isolate)) return *string;

  Handle<String> form;
  ASSIGN_RETURN_FAILURE_ON_EXCEPTION(isolate, form,
                                     Object::ToString(isolate, form_input));

  if (!(String::Equals(isolate, form, isolate->factory()->NFC_string()) ||
        String::Equals(isolate, form, isolate->factory()->NFD_string()) ||
        String::Equals(isolate, form, isolate->factory()->NFKC_string()) ||
        String::Equals(isolate, form, isolate->factory()->NFKD_string()))) {
    Handle<String> valid_forms =
        isolate->factory()->NewStringFromStaticChars("NFC, NFD, NFKC, NFKD");
    THROW_NEW_ERROR_RETURN_FAILURE(
        isolate,
        NewRangeError(MessageTemplate::kNormalizationForm, valid_forms));
  }

  return *string;
}
#endif  // !V8_INTL_SUPPORT


#ifndef V8_INTL_SUPPORT
namespace {

inline bool ToUpperOverflows(base::uc32 character) {
  // y with umlauts and the micro sign are the only characters that stop
  // fitting into one-byte when converting to uppercase.
  static const base::uc32 yuml_code = 0xFF;
  static const base::uc32 micro_code = 0xB5;
  return (character == yuml_code || character == micro_code);
}

template <class Converter>
V8_WARN_UNUSED_RESULT static Tagged<Object> ConvertCaseHelper(
    Isolate* isolate, Tagged<String> string, Tagged<SeqString> result,
    int result_length, unibrow::Mapping<Converter, 128>* mapping) {
  DisallowGarbageCollection no_gc;
  // We try this twice, once with the assumption that the result is no longer
  // than the input and, if that assumption breaks, again with the exact
  // length.  This may not be pretty, but it is nicer than what was here before
  // and I hereby claim my vaffel-is.
  //
  // NOTE: This assumes that the upper/lower case of an ASCII
  // character is also ASCII.  This is currently the case, but it
  // might break in the future if we implement more context and locale
  // dependent upper/lower conversions.
  bool has_changed_character = false;

  // Convert all characters to upper case, assuming that they will fit
  // in the buffer
  StringCharacterStream stream(string);
  unibrow::uchar chars[Converter::kMaxWidth];
  // We can assume that the string is not empty
  base::uc32 current = stream.GetNext();
  bool ignore_overflow = Converter::kIsToLower || IsSeqTwoByteString(result);
  for (int i = 0; i < result_length;) {
    bool has_next = stream.HasMore();
    base::uc32 next = has_next ? stream.GetNext() : 0;
    int char_length = mapping->get(current, next, chars);
    if (char_length == 0) {
      // The case conversion of this character is the character itself.
      result->Set(i, current);
      i++;
    } else if (char_length == 1 &&
               (ignore_overflow || !ToUpperOverflows(current))) {
      // Common case: converting the letter resulted in one character.
      DCHECK(static_cast<base::uc32>(chars[0]) != current);
      result->Set(i, chars[0]);
      has_changed_character = true;
      i++;
    } else if (result_length == string->length()) {
      bool overflows = ToUpperOverflows(current);
      // We've assumed that the result would be as long as the
      // input but here is a character that converts to several
      // characters.  No matter, we calculate the exact length
      // of the result and try the whole thing again.
      //
      // Note that this leaves room for optimization.  We could just
      // memcpy what we already have to the result string.  Also,
      // the result string is the last object allocated we could
      // "realloc" it and probably, in the vast majority of cases,
      // extend the existing string to be able to hold the full
      // result.
      int next_length = 0;
      if (has_next) {
        next_length = mapping->get(next, 0, chars);
        if (next_length == 0) next_length = 1;
      }
      int current_length = i + char_length + next_length;
      while (stream.HasMore()) {
        current = stream.GetNext();
        overflows |= ToUpperOverflows(current);
        // NOTE: we use 0 as the next character here because, while
        // the next character may affect what a character converts to,
        // it does not in any case affect the length of what it convert
        // to.
        int char_length = mapping->get(current, 0, chars);
        if (char_length == 0) char_length = 1;
        current_length += char_length;
        if (current_length > String::kMaxLength) {
          AllowGarbageCollection allocate_error_and_return;
          THROW_NEW_ERROR_RETURN_FAILURE(isolate,
                                         NewInvalidStringLengthError());
        }
      }
      // Try again with the real length.  Return signed if we need
      // to allocate a two-byte string for to uppercase.
      return (overflows && !ignore_overflow) ? Smi::FromInt(-current_length)
                                             : Smi::FromInt(current_length);
    } else {
      for (int j = 0; j < char_length; j++) {
        result->Set(i, chars[j]);
        i++;
      }
      has_changed_character = true;
    }
    current = next;
  }
  if (has_changed_character) {
    return result;
  } else {
    // If we didn't actually change anything in doing the conversion
    // we simple return the result and let the converted string
    // become garbage; there is no reason to keep two identical strings
    // alive.
    return string;
  }
}

template <class Converter>
V8_WARN_UNUSED_RESULT static Tagged<Object> ConvertCase(
    Handle<String> s, Isolate* isolate,
    unibrow::Mapping<Converter, 128>* mapping) {
  s = String::Flatten(isolate, s);
  int length = s->length();
  // Assume that the string is not empty; we need this assumption later
  if (length == 0) return *s;

  // Simpler handling of ASCII strings.
  //
  // NOTE: This assumes that the upper/lower case of an ASCII
  // character is also ASCII.  This is currently the case, but it
  // might break in the future if we implement more context and locale
  // dependent upper/lower conversions.
  if (String::IsOneByteRepresentationUnderneath(*s)) {
    // Same length as input.
    Handle<SeqOneByteString> result =
        isolate->factory()->NewRawOneByteString(length).ToHandleChecked();
    DisallowGarbageCollection no_gc;
    String::FlatContent flat_content = s->GetFlatContent(no_gc);
    DCHECK(flat_content.IsFlat());
    bool has_changed_character = false;
    int index_to_first_unprocessed = FastAsciiConvert<Converter::kIsToLower>(
        reinterpret_cast<char*>(result->GetChars(no_gc)),
        reinterpret_cast<const char*>(flat_content.ToOneByteVector().begin()),
        length, &has_changed_character);
    // If not ASCII, we discard the result and take the 2 byte path.
    if (index_to_first_unprocessed == length)
      return has_changed_character ? *result : *s;
  }

  Handle<SeqString> result;  // Same length as input.
  if (s->IsOneByteRepresentation()) {
    result = isolate->factory()->NewRawOneByteString(length).ToHandleChecked();
  } else {
    result = isolate->factory()->NewRawTwoByteString(length).ToHandleChecked();
  }

  Tagged<Object> answer =
      ConvertCaseHelper(isolate, *s, *result, length, mapping);
  if (IsException(answer, isolate) || IsString(answer)) return answer;

  DCHECK(IsSmi(answer));
  length = Smi::ToInt(answer);
  if (s->IsOneByteRepresentation() && length > 0) {
    ASSIGN_RETURN_FAILURE_ON_EXCEPTION(
        isolate, result, isolate->factory()->NewRawOneByteString(length));
  } else {
    if (length < 0) length = -length;
    ASSIGN_RETURN_FAILURE_ON_EXCEPTION(
        isolate, result, isolate->factory()->NewRawTwoByteString(length));
  }
  return ConvertCaseHelper(isolate, *s, *result, length, mapping);
}

}  // namespace

BUILTIN(StringPrototypeToLocaleLowerCase) {
  HandleScope scope(isolate);
  TO_THIS_STRING(string, "String.prototype.toLocaleLowerCase");
  return ConvertCase(string, isolate,
                     isolate->runtime_state()->to_lower_mapping());
}

BUILTIN(StringPrototypeToLocaleUpperCase) {
  HandleScope scope(isolate);
  TO_THIS_STRING(string, "String.prototype.toLocaleUpperCase");
  return ConvertCase(string, isolate,
                     isolate->runtime_state()->to_upper_mapping());
}

BUILTIN(StringPrototypeToLowerCase) {
  HandleScope scope(isolate);
  TO_THIS_STRING(string, "String.prototype.toLowerCase");
  return ConvertCase(string, isolate,
                     isolate->runtime_state()->to_lower_mapping());
}

BUILTIN(StringPrototypeToUpperCase) {
  HandleScope scope(isolate);
  TO_THIS_STRING(string, "String.prototype.toUpperCase");
  return ConvertCase(string, isolate,
                     isolate->runtime_state()->to_upper_mapping());
}
#endif  // !V8_INTL_SUPPORT

// ES6 #sec-string.prototype.raw
BUILTIN(StringRaw) { … }

}  // namespace internal
}  // namespace v8
chromium/v8/src/builtins/builtins-string.cc
