url/url_canon_icu.cc

   1 // Copyright 2013 The Chromium 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
   5 // ICU integration functions.
   6
   7 #include <stdlib.h>
   8 #include <string.h>
   9
  10 #include "base/lazy_instance.h"
  11 #include "base/logging.h"
  12 #include "third_party/icu/source/common/unicode/ucnv.h"
  13 #include "third_party/icu/source/common/unicode/ucnv_cb.h"
  14 #include "third_party/icu/source/common/unicode/uidna.h"
  15 #include "url/url_canon_icu.h"
  16 #include "url/url_canon_internal.h"  // for _itoa_s
  17
  18 namespace url_canon {
  19
  20 namespace {
  21
  22 // Called when converting a character that can not be represented, this will
  23 // append an escaped version of the numerical character reference for that code
  24 // point. It is of the form "&#1234;" and we will escape the non-digits to
  25 // "%26%231234%3B". Why? This is what Netscape did back in the olden days.
  26 void appendURLEscapedChar(const void* context,
  27                           UConverterFromUnicodeArgs* from_args,
  28                           const UChar* code_units,
  29                           int32_t length,
  30                           UChar32 code_point,
  31                           UConverterCallbackReason reason,
  32                           UErrorCode* err) {
  33   if (reason == UCNV_UNASSIGNED) {
  34     *err = U_ZERO_ERROR;
  35
  36     const static int prefix_len = 6;
  37     const static char prefix[prefix_len + 1] = "%26%23";  // "&#" percent-escaped
  38     ucnv_cbFromUWriteBytes(from_args, prefix, prefix_len, 0, err);
  39
  40     DCHECK(code_point < 0x110000);
  41     char number[8];  // Max Unicode code point is 7 digits.
  42     _itoa_s(code_point, number, 10);
  43     int number_len = static_cast<int>(strlen(number));
  44     ucnv_cbFromUWriteBytes(from_args, number, number_len, 0, err);
  45
  46     const static int postfix_len = 3;
  47     const static char postfix[postfix_len + 1] = "%3B";   // ";" percent-escaped
  48     ucnv_cbFromUWriteBytes(from_args, postfix, postfix_len, 0, err);
  49   }
  50 }
  51
  52 // A class for scoping the installation of the invalid character callback.
  53 class AppendHandlerInstaller {
  54  public:
  55   // The owner of this object must ensure that the converter is alive for the
  56   // duration of this object's lifetime.
  57   AppendHandlerInstaller(UConverter* converter) : converter_(converter) {
  58     UErrorCode err = U_ZERO_ERROR;
  59     ucnv_setFromUCallBack(converter_, appendURLEscapedChar, 0,
  60                           &old_callback_, &old_context_, &err);
  61   }
  62
  63   ~AppendHandlerInstaller() {
  64     UErrorCode err = U_ZERO_ERROR;
  65     ucnv_setFromUCallBack(converter_, old_callback_, old_context_, 0, 0, &err);
  66   }
  67
  68  private:
  69   UConverter* converter_;
  70
  71   UConverterFromUCallback old_callback_;
  72   const void* old_context_;
  73 };
  74
  75 // A wrapper to use LazyInstance<>::Leaky with ICU's UIDNA, a C pointer to
  76 // a UTS46/IDNA 2008 handling object opened with uidna_openUTS46().
  77 //
  78 // We use UTS46 with BiDiCheck to migrate from IDNA 2003 (with unassigned
  79 // code points allowed) to IDNA 2008 with
  80 // the backward compatibility in mind. What it does:
  81 //
  82 // 1. Use the up-to-date Unicode data.
  83 // 2. Define a case folding/mapping with the up-to-date Unicode data as
  84 //    in IDNA 2003.
  85 // 3. Use transitional mechanism for 4 deviation characters (sharp-s,
  86 //    final sigma, ZWJ and ZWNJ) for now.
  87 // 4. Continue to allow symbols and punctuations.
  88 // 5. Apply new BiDi check rules more permissive than the IDNA 2003 BiDI rules.
  89 // 6. Do not apply STD3 rules
  90 // 7. Do not allow unassigned code points.
  91 //
  92 // It also closely matches what IE 10 does except for the BiDi check (
  93 // http://goo.gl/3XBhqw ).
  94 // See http://http://unicode.org/reports/tr46/ and references therein
  95 // for more details.
  96 struct UIDNAWrapper {
  97   UIDNAWrapper() {
  98     UErrorCode err = U_ZERO_ERROR;
  99     // TODO(jungshik): Change options as different parties (browsers,
 100     // registrars, search engines) converge toward a consensus.
 101     value = uidna_openUTS46(UIDNA_CHECK_BIDI, &err);
 102     if (U_FAILURE(err))
 103       value = NULL;
 104   }
 105
 106   UIDNA* value;
 107 };
 108
 109 }  // namespace
 110
 111 ICUCharsetConverter::ICUCharsetConverter(UConverter* converter)
 112     : converter_(converter) {
 113 }
 114
 115 ICUCharsetConverter::~ICUCharsetConverter() {
 116 }
 117
 118 void ICUCharsetConverter::ConvertFromUTF16(const base::char16* input,
 119                                            int input_len,
 120                                            CanonOutput* output) {
 121   // Install our error handler. It will be called for character that can not
 122   // be represented in the destination character set.
 123   AppendHandlerInstaller handler(converter_);
 124
 125   int begin_offset = output->length();
 126   int dest_capacity = output->capacity() - begin_offset;
 127   output->set_length(output->length());
 128
 129   do {
 130     UErrorCode err = U_ZERO_ERROR;
 131     char* dest = &output->data()[begin_offset];
 132     int required_capacity = ucnv_fromUChars(converter_, dest, dest_capacity,
 133                                             input, input_len, &err);
 134     if (err != U_BUFFER_OVERFLOW_ERROR) {
 135       output->set_length(begin_offset + required_capacity);
 136       return;
 137     }
 138
 139     // Output didn't fit, expand
 140     dest_capacity = required_capacity;
 141     output->Resize(begin_offset + dest_capacity);
 142   } while (true);
 143 }
 144
 145 static base::LazyInstance<UIDNAWrapper>::Leaky
 146     g_uidna = LAZY_INSTANCE_INITIALIZER;
 147
 148 // Converts the Unicode input representing a hostname to ASCII using IDN rules.
 149 // The output must be ASCII, but is represented as wide characters.
 150 //
 151 // On success, the output will be filled with the ASCII host name and it will
 152 // return true. Unlike most other canonicalization functions, this assumes that
 153 // the output is empty. The beginning of the host will be at offset 0, and
 154 // the length of the output will be set to the length of the new host name.
 155 //
 156 // On error, this will return false. The output in this case is undefined.
 157 // TODO(jungshik): use UTF-8/ASCII version of nameToASCII.
 158 // Change the function signature and callers accordingly to avoid unnecessary
 159 // conversions in our code. In addition, consider using icu::IDNA's UTF-8/ASCII
 160 // version with StringByteSink. That way, we can avoid C wrappers and additional
 161 // string conversion.
 162 bool IDNToASCII(const base::char16* src, int src_len, CanonOutputW* output) {
 163   DCHECK(output->length() == 0);  // Output buffer is assumed empty.
 164
 165   UIDNA* uidna = g_uidna.Get().value;
 166   DCHECK(uidna != NULL);
 167   while (true) {
 168     UErrorCode err = U_ZERO_ERROR;
 169     UIDNAInfo info = UIDNA_INFO_INITIALIZER;
 170     int output_length = uidna_nameToASCII(uidna, src, src_len, output->data(),
 171                                           output->capacity(), &info, &err);
 172     if (U_SUCCESS(err) && info.errors == 0) {
 173       output->set_length(output_length);
 174       return true;
 175     }
 176
 177     // TODO(jungshik): Look at info.errors to handle them case-by-case basis
 178     // if necessary.
 179     if (err != U_BUFFER_OVERFLOW_ERROR || info.errors != 0)
 180       return false;  // Unknown error, give up.
 181
 182     // Not enough room in our buffer, expand.
 183     output->Resize(output_length);
 184   }
 185 }
 186
 187 bool ReadUTFChar(const char* str, int* begin, int length,
 188                  unsigned* code_point_out) {
 189   int code_point;  // Avoids warning when U8_NEXT writes -1 to it.
 190   U8_NEXT(str, *begin, length, code_point);
 191   *code_point_out = static_cast<unsigned>(code_point);
 192
 193   // The ICU macro above moves to the next char, we want to point to the last
 194   // char consumed.
 195   (*begin)--;
 196
 197   // Validate the decoded value.
 198   if (U_IS_UNICODE_CHAR(code_point))
 199     return true;
 200   *code_point_out = kUnicodeReplacementCharacter;
 201   return false;
 202 }
 203
 204 bool ReadUTFChar(const base::char16* str, int* begin, int length,
 205                  unsigned* code_point) {
 206   if (U16_IS_SURROGATE(str[*begin])) {
 207     if (!U16_IS_SURROGATE_LEAD(str[*begin]) || *begin + 1 >= length ||
 208         !U16_IS_TRAIL(str[*begin + 1])) {
 209       // Invalid surrogate pair.
 210       *code_point = kUnicodeReplacementCharacter;
 211       return false;
 212     } else {
 213       // Valid surrogate pair.
 214       *code_point = U16_GET_SUPPLEMENTARY(str[*begin], str[*begin + 1]);
 215       (*begin)++;
 216     }
 217   } else {
 218     // Not a surrogate, just one 16-bit word.
 219     *code_point = str[*begin];
 220   }
 221
 222   if (U_IS_UNICODE_CHAR(*code_point))
 223     return true;
 224
 225   // Invalid code point.
 226   *code_point = kUnicodeReplacementCharacter;
 227   return false;
 228 }
 229
 230 }  // namespace url_canon