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/logging.h"
  11 #include "third_party/icu/public/common/unicode/ucnv.h"
  12 #include "third_party/icu/public/common/unicode/ucnv_cb.h"
  13 #include "third_party/icu/public/common/unicode/uidna.h"
  14 #include "url/url_canon_icu.h"
  15 #include "url/url_canon_internal.h"  // for _itoa_s
  16
  17 namespace url_canon {
  18
  19 namespace {
  20
  21 // Called when converting a character that can not be represented, this will
  22 // append an escaped version of the numerical character reference for that code
  23 // point. It is of the form "&#1234;" and we will escape the non-digits to
  24 // "%26%231234%3B". Why? This is what Netscape did back in the olden days.
  25 void appendURLEscapedChar(const void* context,
  26                           UConverterFromUnicodeArgs* from_args,
  27                           const UChar* code_units,
  28                           int32_t length,
  29                           UChar32 code_point,
  30                           UConverterCallbackReason reason,
  31                           UErrorCode* err) {
  32   if (reason == UCNV_UNASSIGNED) {
  33     *err = U_ZERO_ERROR;
  34
  35     const static int prefix_len = 6;
  36     const static char prefix[prefix_len + 1] = "%26%23";  // "&#" percent-escaped
  37     ucnv_cbFromUWriteBytes(from_args, prefix, prefix_len, 0, err);
  38
  39     DCHECK(code_point < 0x110000);
  40     char number[8];  // Max Unicode code point is 7 digits.
  41     _itoa_s(code_point, number, 10);
  42     int number_len = static_cast<int>(strlen(number));
  43     ucnv_cbFromUWriteBytes(from_args, number, number_len, 0, err);
  44
  45     const static int postfix_len = 3;
  46     const static char postfix[postfix_len + 1] = "%3B";   // ";" percent-escaped
  47     ucnv_cbFromUWriteBytes(from_args, postfix, postfix_len, 0, err);
  48   }
  49 }
  50
  51 // A class for scoping the installation of the invalid character callback.
  52 class AppendHandlerInstaller {
  53  public:
  54   // The owner of this object must ensure that the converter is alive for the
  55   // duration of this object's lifetime.
  56   AppendHandlerInstaller(UConverter* converter) : converter_(converter) {
  57     UErrorCode err = U_ZERO_ERROR;
  58     ucnv_setFromUCallBack(converter_, appendURLEscapedChar, 0,
  59                           &old_callback_, &old_context_, &err);
  60   }
  61
  62   ~AppendHandlerInstaller() {
  63     UErrorCode err = U_ZERO_ERROR;
  64     ucnv_setFromUCallBack(converter_, old_callback_, old_context_, 0, 0, &err);
  65   }
  66
  67  private:
  68   UConverter* converter_;
  69
  70   UConverterFromUCallback old_callback_;
  71   const void* old_context_;
  72 };
  73
  74 }  // namespace
  75
  76 ICUCharsetConverter::ICUCharsetConverter(UConverter* converter)
  77     : converter_(converter) {
  78 }
  79
  80 ICUCharsetConverter::~ICUCharsetConverter() {
  81 }
  82
  83 void ICUCharsetConverter::ConvertFromUTF16(const base::char16* input,
  84                                            int input_len,
  85                                            CanonOutput* output) {
  86   // Install our error handler. It will be called for character that can not
  87   // be represented in the destination character set.
  88   AppendHandlerInstaller handler(converter_);
  89
  90   int begin_offset = output->length();
  91   int dest_capacity = output->capacity() - begin_offset;
  92   output->set_length(output->length());
  93
  94   do {
  95     UErrorCode err = U_ZERO_ERROR;
  96     char* dest = &output->data()[begin_offset];
  97     int required_capacity = ucnv_fromUChars(converter_, dest, dest_capacity,
  98                                             input, input_len, &err);
  99     if (err != U_BUFFER_OVERFLOW_ERROR) {
 100       output->set_length(begin_offset + required_capacity);
 101       return;
 102     }
 103
 104     // Output didn't fit, expand
 105     dest_capacity = required_capacity;
 106     output->Resize(begin_offset + dest_capacity);
 107   } while (true);
 108 }
 109
 110 // Converts the Unicode input representing a hostname to ASCII using IDN rules.
 111 // The output must be ASCII, but is represented as wide characters.
 112 //
 113 // On success, the output will be filled with the ASCII host name and it will
 114 // return true. Unlike most other canonicalization functions, this assumes that
 115 // the output is empty. The beginning of the host will be at offset 0, and
 116 // the length of the output will be set to the length of the new host name.
 117 //
 118 // On error, this will return false. The output in this case is undefined.
 119 bool IDNToASCII(const base::char16* src, int src_len, CanonOutputW* output) {
 120   DCHECK(output->length() == 0);  // Output buffer is assumed empty.
 121   while (true) {
 122     // Use ALLOW_UNASSIGNED to be more tolerant of hostnames that violate
 123     // the spec (which do exist). This does not present any risk and is a
 124     // little more future proof.
 125     UErrorCode err = U_ZERO_ERROR;
 126     int num_converted = uidna_IDNToASCII(src, src_len, output->data(),
 127                                          output->capacity(),
 128                                          UIDNA_ALLOW_UNASSIGNED, NULL, &err);
 129     if (err == U_ZERO_ERROR) {
 130       output->set_length(num_converted);
 131       return true;
 132     }
 133     if (err != U_BUFFER_OVERFLOW_ERROR)
 134       return false;  // Unknown error, give up.
 135
 136     // Not enough room in our buffer, expand.
 137     output->Resize(output->capacity() * 2);
 138   }
 139 }
 140
 141 bool ReadUTFChar(const char* str, int* begin, int length,
 142                  unsigned* code_point_out) {
 143   int code_point;  // Avoids warning when U8_NEXT writes -1 to it.
 144   U8_NEXT(str, *begin, length, code_point);
 145   *code_point_out = static_cast<unsigned>(code_point);
 146
 147   // The ICU macro above moves to the next char, we want to point to the last
 148   // char consumed.
 149   (*begin)--;
 150
 151   // Validate the decoded value.
 152   if (U_IS_UNICODE_CHAR(code_point))
 153     return true;
 154   *code_point_out = kUnicodeReplacementCharacter;
 155   return false;
 156 }
 157
 158 bool ReadUTFChar(const base::char16* str, int* begin, int length,
 159                  unsigned* code_point) {
 160   if (U16_IS_SURROGATE(str[*begin])) {
 161     if (!U16_IS_SURROGATE_LEAD(str[*begin]) || *begin + 1 >= length ||
 162         !U16_IS_TRAIL(str[*begin + 1])) {
 163       // Invalid surrogate pair.
 164       *code_point = kUnicodeReplacementCharacter;
 165       return false;
 166     } else {
 167       // Valid surrogate pair.
 168       *code_point = U16_GET_SUPPLEMENTARY(str[*begin], str[*begin + 1]);
 169       (*begin)++;
 170     }
 171   } else {
 172     // Not a surrogate, just one 16-bit word.
 173     *code_point = str[*begin];
 174   }
 175
 176   if (U_IS_UNICODE_CHAR(*code_point))
 177     return true;
 178
 179   // Invalid code point.
 180   *code_point = kUnicodeReplacementCharacter;
 181   return false;
 182 }
 183
 184 }  // namespace url_canon