blob: d9501c1a4883d525cc4477be7749d4b8b7b752ea [file] [log] [blame]
// Copyright 2013 The Chromium 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 "base/strings/string_util.h"
#include <ctype.h>
#include <errno.h>
#include <math.h>
#include <stdarg.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <wchar.h>
#include <wctype.h>
#include <algorithm>
#include <limits>
#include <vector>
#include "base/logging.h"
#include "base/no_destructor.h"
#include "base/stl_util.h"
#include "base/strings/utf_string_conversion_utils.h"
#include "base/strings/utf_string_conversions.h"
#include "base/third_party/icu/icu_utf.h"
#include "build/build_config.h"
namespace base {
namespace {
// Used by ReplaceStringPlaceholders to track the position in the string of
// replaced parameters.
struct ReplacementOffset {
ReplacementOffset(uintptr_t parameter, size_t offset)
: parameter(parameter),
offset(offset) {}
// Index of the parameter.
uintptr_t parameter;
// Starting position in the string.
size_t offset;
};
static bool CompareParameter(const ReplacementOffset& elem1,
const ReplacementOffset& elem2) {
return elem1.parameter < elem2.parameter;
}
// Overloaded function to append one string onto the end of another. Having a
// separate overload for |source| as both string and StringPiece allows for more
// efficient usage from functions templated to work with either type (avoiding a
// redundant call to the BasicStringPiece constructor in both cases).
template <typename string_type>
inline void AppendToString(string_type* target, const string_type& source) {
target->append(source);
}
template <typename string_type>
inline void AppendToString(string_type* target,
const BasicStringPiece<string_type>& source) {
source.AppendToString(target);
}
// Assuming that a pointer is the size of a "machine word", then
// uintptr_t is an integer type that is also a machine word.
using MachineWord = uintptr_t;
inline bool IsMachineWordAligned(const void* pointer) {
return !(reinterpret_cast<MachineWord>(pointer) & (sizeof(MachineWord) - 1));
}
template <typename CharacterType>
struct NonASCIIMask;
template <>
struct NonASCIIMask<char> {
static constexpr MachineWord value() {
return static_cast<MachineWord>(0x8080808080808080ULL);
}
};
template <>
struct NonASCIIMask<char16> {
static constexpr MachineWord value() {
return static_cast<MachineWord>(0xFF80FF80FF80FF80ULL);
}
};
#if defined(WCHAR_T_IS_UTF32)
template <>
struct NonASCIIMask<wchar_t> {
static constexpr MachineWord value() {
return static_cast<MachineWord>(0xFFFFFF80FFFFFF80ULL);
}
};
#endif // WCHAR_T_IS_UTF32
} // namespace
bool IsWprintfFormatPortable(const wchar_t* format) {
for (const wchar_t* position = format; *position != '\0'; ++position) {
if (*position == '%') {
bool in_specification = true;
bool modifier_l = false;
while (in_specification) {
// Eat up characters until reaching a known specifier.
if (*++position == '\0') {
// The format string ended in the middle of a specification. Call
// it portable because no unportable specifications were found. The
// string is equally broken on all platforms.
return true;
}
if (*position == 'l') {
// 'l' is the only thing that can save the 's' and 'c' specifiers.
modifier_l = true;
} else if (((*position == 's' || *position == 'c') && !modifier_l) ||
*position == 'S' || *position == 'C' || *position == 'F' ||
*position == 'D' || *position == 'O' || *position == 'U') {
// Not portable.
return false;
}
if (wcschr(L"diouxXeEfgGaAcspn%", *position)) {
// Portable, keep scanning the rest of the format string.
in_specification = false;
}
}
}
}
return true;
}
namespace {
template<typename StringType>
StringType ToLowerASCIIImpl(BasicStringPiece<StringType> str) {
StringType ret;
ret.reserve(str.size());
for (size_t i = 0; i < str.size(); i++)
ret.push_back(ToLowerASCII(str[i]));
return ret;
}
template<typename StringType>
StringType ToUpperASCIIImpl(BasicStringPiece<StringType> str) {
StringType ret;
ret.reserve(str.size());
for (size_t i = 0; i < str.size(); i++)
ret.push_back(ToUpperASCII(str[i]));
return ret;
}
} // namespace
std::string ToLowerASCII(StringPiece str) {
return ToLowerASCIIImpl<std::string>(str);
}
string16 ToLowerASCII(StringPiece16 str) {
return ToLowerASCIIImpl<string16>(str);
}
std::string ToUpperASCII(StringPiece str) {
return ToUpperASCIIImpl<std::string>(str);
}
string16 ToUpperASCII(StringPiece16 str) {
return ToUpperASCIIImpl<string16>(str);
}
template<class StringType>
int CompareCaseInsensitiveASCIIT(BasicStringPiece<StringType> a,
BasicStringPiece<StringType> b) {
// Find the first characters that aren't equal and compare them. If the end
// of one of the strings is found before a nonequal character, the lengths
// of the strings are compared.
size_t i = 0;
while (i < a.length() && i < b.length()) {
typename StringType::value_type lower_a = ToLowerASCII(a[i]);
typename StringType::value_type lower_b = ToLowerASCII(b[i]);
if (lower_a < lower_b)
return -1;
if (lower_a > lower_b)
return 1;
i++;
}
// End of one string hit before finding a different character. Expect the
// common case to be "strings equal" at this point so check that first.
if (a.length() == b.length())
return 0;
if (a.length() < b.length())
return -1;
return 1;
}
int CompareCaseInsensitiveASCII(StringPiece a, StringPiece b) {
return CompareCaseInsensitiveASCIIT<std::string>(a, b);
}
int CompareCaseInsensitiveASCII(StringPiece16 a, StringPiece16 b) {
return CompareCaseInsensitiveASCIIT<string16>(a, b);
}
bool EqualsCaseInsensitiveASCII(StringPiece a, StringPiece b) {
if (a.length() != b.length())
return false;
return CompareCaseInsensitiveASCIIT<std::string>(a, b) == 0;
}
bool EqualsCaseInsensitiveASCII(StringPiece16 a, StringPiece16 b) {
if (a.length() != b.length())
return false;
return CompareCaseInsensitiveASCIIT<string16>(a, b) == 0;
}
const std::string& EmptyString() {
static const base::NoDestructor<std::string> s;
return *s;
}
const string16& EmptyString16() {
static const base::NoDestructor<string16> s16;
return *s16;
}
template <class StringType>
bool ReplaceCharsT(const StringType& input,
BasicStringPiece<StringType> find_any_of_these,
BasicStringPiece<StringType> replace_with,
StringType* output);
bool ReplaceChars(const string16& input,
StringPiece16 replace_chars,
StringPiece16 replace_with,
string16* output) {
return ReplaceCharsT(input, replace_chars, replace_with, output);
}
bool ReplaceChars(const std::string& input,
StringPiece replace_chars,
StringPiece replace_with,
std::string* output) {
return ReplaceCharsT(input, replace_chars, replace_with, output);
}
bool RemoveChars(const string16& input,
StringPiece16 remove_chars,
string16* output) {
return ReplaceCharsT(input, remove_chars, StringPiece16(), output);
}
bool RemoveChars(const std::string& input,
StringPiece remove_chars,
std::string* output) {
return ReplaceCharsT(input, remove_chars, StringPiece(), output);
}
template <typename Str>
TrimPositions TrimStringT(BasicStringPiece<Str> input,
BasicStringPiece<Str> trim_chars,
TrimPositions positions,
Str* output) {
// Find the edges of leading/trailing whitespace as desired. Need to use
// a StringPiece version of input to be able to call find* on it with the
// StringPiece version of trim_chars (normally the trim_chars will be a
// constant so avoid making a copy).
const size_t last_char = input.length() - 1;
const size_t first_good_char =
(positions & TRIM_LEADING) ? input.find_first_not_of(trim_chars) : 0;
const size_t last_good_char = (positions & TRIM_TRAILING)
? input.find_last_not_of(trim_chars)
: last_char;
// When the string was all trimmed, report that we stripped off characters
// from whichever position the caller was interested in. For empty input, we
// stripped no characters, but we still need to clear |output|.
if (input.empty() || first_good_char == Str::npos ||
last_good_char == Str::npos) {
bool input_was_empty = input.empty(); // in case output == &input
output->clear();
return input_was_empty ? TRIM_NONE : positions;
}
// Trim.
output->assign(input.data() + first_good_char,
last_good_char - first_good_char + 1);
// Return where we trimmed from.
return static_cast<TrimPositions>(
(first_good_char == 0 ? TRIM_NONE : TRIM_LEADING) |
(last_good_char == last_char ? TRIM_NONE : TRIM_TRAILING));
}
bool TrimString(StringPiece16 input,
StringPiece16 trim_chars,
string16* output) {
return TrimStringT(input, trim_chars, TRIM_ALL, output) != TRIM_NONE;
}
bool TrimString(StringPiece input,
StringPiece trim_chars,
std::string* output) {
return TrimStringT(input, trim_chars, TRIM_ALL, output) != TRIM_NONE;
}
template<typename Str>
BasicStringPiece<Str> TrimStringPieceT(BasicStringPiece<Str> input,
BasicStringPiece<Str> trim_chars,
TrimPositions positions) {
size_t begin = (positions & TRIM_LEADING) ?
input.find_first_not_of(trim_chars) : 0;
size_t end = (positions & TRIM_TRAILING) ?
input.find_last_not_of(trim_chars) + 1 : input.size();
return input.substr(begin, end - begin);
}
StringPiece16 TrimString(StringPiece16 input,
StringPiece16 trim_chars,
TrimPositions positions) {
return TrimStringPieceT(input, trim_chars, positions);
}
StringPiece TrimString(StringPiece input,
StringPiece trim_chars,
TrimPositions positions) {
return TrimStringPieceT(input, trim_chars, positions);
}
void TruncateUTF8ToByteSize(const std::string& input,
const size_t byte_size,
std::string* output) {
DCHECK(output);
if (byte_size > input.length()) {
*output = input;
return;
}
DCHECK_LE(byte_size,
static_cast<uint32_t>(std::numeric_limits<int32_t>::max()));
// Note: This cast is necessary because CBU8_NEXT uses int32_ts.
int32_t truncation_length = static_cast<int32_t>(byte_size);
int32_t char_index = truncation_length - 1;
const char* data = input.data();
// Using CBU8, we will move backwards from the truncation point
// to the beginning of the string looking for a valid UTF8
// character. Once a full UTF8 character is found, we will
// truncate the string to the end of that character.
while (char_index >= 0) {
int32_t prev = char_index;
base_icu::UChar32 code_point = 0;
CBU8_NEXT(data, char_index, truncation_length, code_point);
if (!IsValidCharacter(code_point) ||
!IsValidCodepoint(code_point)) {
char_index = prev - 1;
} else {
break;
}
}
if (char_index >= 0 )
*output = input.substr(0, char_index);
else
output->clear();
}
TrimPositions TrimWhitespace(StringPiece16 input,
TrimPositions positions,
string16* output) {
return TrimStringT(input, StringPiece16(kWhitespaceUTF16), positions, output);
}
StringPiece16 TrimWhitespace(StringPiece16 input,
TrimPositions positions) {
return TrimStringPieceT(input, StringPiece16(kWhitespaceUTF16), positions);
}
TrimPositions TrimWhitespaceASCII(StringPiece input,
TrimPositions positions,
std::string* output) {
return TrimStringT(input, StringPiece(kWhitespaceASCII), positions, output);
}
StringPiece TrimWhitespaceASCII(StringPiece input, TrimPositions positions) {
return TrimStringPieceT(input, StringPiece(kWhitespaceASCII), positions);
}
template<typename STR>
STR CollapseWhitespaceT(const STR& text,
bool trim_sequences_with_line_breaks) {
STR result;
result.resize(text.size());
// Set flags to pretend we're already in a trimmed whitespace sequence, so we
// will trim any leading whitespace.
bool in_whitespace = true;
bool already_trimmed = true;
int chars_written = 0;
for (typename STR::const_iterator i(text.begin()); i != text.end(); ++i) {
if (IsUnicodeWhitespace(*i)) {
if (!in_whitespace) {
// Reduce all whitespace sequences to a single space.
in_whitespace = true;
result[chars_written++] = L' ';
}
if (trim_sequences_with_line_breaks && !already_trimmed &&
((*i == '\n') || (*i == '\r'))) {
// Whitespace sequences containing CR or LF are eliminated entirely.
already_trimmed = true;
--chars_written;
}
} else {
// Non-whitespace chracters are copied straight across.
in_whitespace = false;
already_trimmed = false;
result[chars_written++] = *i;
}
}
if (in_whitespace && !already_trimmed) {
// Any trailing whitespace is eliminated.
--chars_written;
}
result.resize(chars_written);
return result;
}
string16 CollapseWhitespace(const string16& text,
bool trim_sequences_with_line_breaks) {
return CollapseWhitespaceT(text, trim_sequences_with_line_breaks);
}
std::string CollapseWhitespaceASCII(const std::string& text,
bool trim_sequences_with_line_breaks) {
return CollapseWhitespaceT(text, trim_sequences_with_line_breaks);
}
bool ContainsOnlyChars(StringPiece input, StringPiece characters) {
return input.find_first_not_of(characters) == StringPiece::npos;
}
bool ContainsOnlyChars(StringPiece16 input, StringPiece16 characters) {
return input.find_first_not_of(characters) == StringPiece16::npos;
}
template <class Char>
inline bool DoIsStringASCII(const Char* characters, size_t length) {
if (!length)
return true;
constexpr MachineWord non_ascii_bit_mask = NonASCIIMask<Char>::value();
MachineWord all_char_bits = 0;
const Char* end = characters + length;
// Prologue: align the input.
while (!IsMachineWordAligned(characters) && characters < end)
all_char_bits |= *characters++;
if (all_char_bits & non_ascii_bit_mask)
return false;
// Compare the values of CPU word size.
constexpr size_t chars_per_word = sizeof(MachineWord) / sizeof(Char);
constexpr int batch_count = 16;
while (characters <= end - batch_count * chars_per_word) {
all_char_bits = 0;
for (int i = 0; i < batch_count; ++i) {
all_char_bits |= *(reinterpret_cast<const MachineWord*>(characters));
characters += chars_per_word;
}
if (all_char_bits & non_ascii_bit_mask)
return false;
}
// Process the remaining words.
all_char_bits = 0;
while (characters <= end - chars_per_word) {
all_char_bits |= *(reinterpret_cast<const MachineWord*>(characters));
characters += chars_per_word;
}
// Process the remaining bytes.
while (characters < end)
all_char_bits |= *characters++;
return !(all_char_bits & non_ascii_bit_mask);
}
bool IsStringASCII(StringPiece str) {
return DoIsStringASCII(str.data(), str.length());
}
bool IsStringASCII(StringPiece16 str) {
return DoIsStringASCII(str.data(), str.length());
}
#if defined(WCHAR_T_IS_UTF32)
bool IsStringASCII(WStringPiece str) {
return DoIsStringASCII(str.data(), str.length());
}
#endif
bool IsStringUTF8(StringPiece str) {
const char *src = str.data();
int32_t src_len = static_cast<int32_t>(str.length());
int32_t char_index = 0;
while (char_index < src_len) {
int32_t code_point;
CBU8_NEXT(src, char_index, src_len, code_point);
if (!IsValidCharacter(code_point))
return false;
}
return true;
}
// Implementation note: Normally this function will be called with a hardcoded
// constant for the lowercase_ascii parameter. Constructing a StringPiece from
// a C constant requires running strlen, so the result will be two passes
// through the buffers, one to file the length of lowercase_ascii, and one to
// compare each letter.
//
// This function could have taken a const char* to avoid this and only do one
// pass through the string. But the strlen is faster than the case-insensitive
// compares and lets us early-exit in the case that the strings are different
// lengths (will often be the case for non-matches). So whether one approach or
// the other will be faster depends on the case.
//
// The hardcoded strings are typically very short so it doesn't matter, and the
// string piece gives additional flexibility for the caller (doesn't have to be
// null terminated) so we choose the StringPiece route.
template<typename Str>
static inline bool DoLowerCaseEqualsASCII(BasicStringPiece<Str> str,
StringPiece lowercase_ascii) {
if (str.size() != lowercase_ascii.size())
return false;
for (size_t i = 0; i < str.size(); i++) {
if (ToLowerASCII(str[i]) != lowercase_ascii[i])
return false;
}
return true;
}
bool LowerCaseEqualsASCII(StringPiece str, StringPiece lowercase_ascii) {
return DoLowerCaseEqualsASCII<std::string>(str, lowercase_ascii);
}
bool LowerCaseEqualsASCII(StringPiece16 str, StringPiece lowercase_ascii) {
return DoLowerCaseEqualsASCII<string16>(str, lowercase_ascii);
}
bool EqualsASCII(StringPiece16 str, StringPiece ascii) {
if (str.length() != ascii.length())
return false;
return std::equal(ascii.begin(), ascii.end(), str.begin());
}
template<typename Str>
bool StartsWithT(BasicStringPiece<Str> str,
BasicStringPiece<Str> search_for,
CompareCase case_sensitivity) {
if (search_for.size() > str.size())
return false;
BasicStringPiece<Str> source = str.substr(0, search_for.size());
switch (case_sensitivity) {
case CompareCase::SENSITIVE:
return source == search_for;
case CompareCase::INSENSITIVE_ASCII:
return std::equal(
search_for.begin(), search_for.end(),
source.begin(),
CaseInsensitiveCompareASCII<typename Str::value_type>());
default:
NOTREACHED();
return false;
}
}
bool StartsWith(StringPiece str,
StringPiece search_for,
CompareCase case_sensitivity) {
return StartsWithT<std::string>(str, search_for, case_sensitivity);
}
bool StartsWith(StringPiece16 str,
StringPiece16 search_for,
CompareCase case_sensitivity) {
return StartsWithT<string16>(str, search_for, case_sensitivity);
}
template <typename Str>
bool EndsWithT(BasicStringPiece<Str> str,
BasicStringPiece<Str> search_for,
CompareCase case_sensitivity) {
if (search_for.size() > str.size())
return false;
BasicStringPiece<Str> source = str.substr(str.size() - search_for.size(),
search_for.size());
switch (case_sensitivity) {
case CompareCase::SENSITIVE:
return source == search_for;
case CompareCase::INSENSITIVE_ASCII:
return std::equal(
source.begin(), source.end(),
search_for.begin(),
CaseInsensitiveCompareASCII<typename Str::value_type>());
default:
NOTREACHED();
return false;
}
}
bool EndsWith(StringPiece str,
StringPiece search_for,
CompareCase case_sensitivity) {
return EndsWithT<std::string>(str, search_for, case_sensitivity);
}
bool EndsWith(StringPiece16 str,
StringPiece16 search_for,
CompareCase case_sensitivity) {
return EndsWithT<string16>(str, search_for, case_sensitivity);
}
char HexDigitToInt(wchar_t c) {
DCHECK(IsHexDigit(c));
if (c >= '0' && c <= '9')
return static_cast<char>(c - '0');
if (c >= 'A' && c <= 'F')
return static_cast<char>(c - 'A' + 10);
if (c >= 'a' && c <= 'f')
return static_cast<char>(c - 'a' + 10);
return 0;
}
bool IsUnicodeWhitespace(wchar_t c) {
// kWhitespaceWide is a NULL-terminated string
for (const wchar_t* cur = kWhitespaceWide; *cur; ++cur) {
if (*cur == c)
return true;
}
return false;
}
static const char* const kByteStringsUnlocalized[] = {
" B",
" kB",
" MB",
" GB",
" TB",
" PB"
};
string16 FormatBytesUnlocalized(int64_t bytes) {
double unit_amount = static_cast<double>(bytes);
size_t dimension = 0;
const int kKilo = 1024;
while (unit_amount >= kKilo &&
dimension < base::size(kByteStringsUnlocalized) - 1) {
unit_amount /= kKilo;
dimension++;
}
char buf[64];
if (bytes != 0 && dimension > 0 && unit_amount < 100) {
base::snprintf(buf, base::size(buf), "%.1lf%s", unit_amount,
kByteStringsUnlocalized[dimension]);
} else {
base::snprintf(buf, base::size(buf), "%.0lf%s", unit_amount,
kByteStringsUnlocalized[dimension]);
}
return ASCIIToUTF16(buf);
}
// A Matcher for DoReplaceMatchesAfterOffset() that matches substrings.
template <class StringType>
struct SubstringMatcher {
BasicStringPiece<StringType> find_this;
size_t Find(const StringType& input, size_t pos) {
return input.find(find_this.data(), pos, find_this.length());
}
size_t MatchSize() { return find_this.length(); }
};
// A Matcher for DoReplaceMatchesAfterOffset() that matches single characters.
template <class StringType>
struct CharacterMatcher {
BasicStringPiece<StringType> find_any_of_these;
size_t Find(const StringType& input, size_t pos) {
return input.find_first_of(find_any_of_these.data(), pos,
find_any_of_these.length());
}
constexpr size_t MatchSize() { return 1; }
};
enum class ReplaceType { REPLACE_ALL, REPLACE_FIRST };
// Runs in O(n) time in the length of |str|, and transforms the string without
// reallocating when possible. Returns |true| if any matches were found.
//
// This is parameterized on a |Matcher| traits type, so that it can be the
// implementation for both ReplaceChars() and ReplaceSubstringsAfterOffset().
template <class StringType, class Matcher>
bool DoReplaceMatchesAfterOffset(StringType* str,
size_t initial_offset,
Matcher matcher,
BasicStringPiece<StringType> replace_with,
ReplaceType replace_type) {
using CharTraits = typename StringType::traits_type;
const size_t find_length = matcher.MatchSize();
if (!find_length)
return false;
// If the find string doesn't appear, there's nothing to do.
size_t first_match = matcher.Find(*str, initial_offset);
if (first_match == StringType::npos)
return false;
// If we're only replacing one instance, there's no need to do anything
// complicated.
const size_t replace_length = replace_with.length();
if (replace_type == ReplaceType::REPLACE_FIRST) {
str->replace(first_match, find_length, replace_with.data(), replace_length);
return true;
}
// If the find and replace strings are the same length, we can simply use
// replace() on each instance, and finish the entire operation in O(n) time.
if (find_length == replace_length) {
auto* buffer = &((*str)[0]);
for (size_t offset = first_match; offset != StringType::npos;
offset = matcher.Find(*str, offset + replace_length)) {
CharTraits::copy(buffer + offset, replace_with.data(), replace_length);
}
return true;
}
// Since the find and replace strings aren't the same length, a loop like the
// one above would be O(n^2) in the worst case, as replace() will shift the
// entire remaining string each time. We need to be more clever to keep things
// O(n).
//
// When the string is being shortened, it's possible to just shift the matches
// down in one pass while finding, and truncate the length at the end of the
// search.
//
// If the string is being lengthened, more work is required. The strategy used
// here is to make two find() passes through the string. The first pass counts
// the number of matches to determine the new size. The second pass will
// either construct the new string into a new buffer (if the existing buffer
// lacked capacity), or else -- if there is room -- create a region of scratch
// space after |first_match| by shifting the tail of the string to a higher
// index, and doing in-place moves from the tail to lower indices thereafter.
size_t str_length = str->length();
size_t expansion = 0;
if (replace_length > find_length) {
// This operation lengthens the string; determine the new length by counting
// matches.
const size_t expansion_per_match = (replace_length - find_length);
size_t num_matches = 0;
for (size_t match = first_match; match != StringType::npos;
match = matcher.Find(*str, match + find_length)) {
expansion += expansion_per_match;
++num_matches;
}
const size_t final_length = str_length + expansion;
if (str->capacity() < final_length) {
// If we'd have to allocate a new buffer to grow the string, build the
// result directly into the new allocation via append().
StringType src(str->get_allocator());
str->swap(src);
str->reserve(final_length);
size_t pos = 0;
for (size_t match = first_match;; match = matcher.Find(src, pos)) {
str->append(src, pos, match - pos);
str->append(replace_with.data(), replace_length);
pos = match + find_length;
// A mid-loop test/break enables skipping the final Find() call; the
// number of matches is known, so don't search past the last one.
if (!--num_matches)
break;
}
// Handle substring after the final match.
str->append(src, pos, str_length - pos);
return true;
}
// Prepare for the copy/move loop below -- expand the string to its final
// size by shifting the data after the first match to the end of the resized
// string.
size_t shift_src = first_match + find_length;
size_t shift_dst = shift_src + expansion;
// Big |expansion| factors (relative to |str_length|) require padding up to
// |shift_dst|.
if (shift_dst > str_length)
str->resize(shift_dst);
str->replace(shift_dst, str_length - shift_src, *str, shift_src,
str_length - shift_src);
str_length = final_length;
}
// We can alternate replacement and move operations. This won't overwrite the
// unsearched region of the string so long as |write_offset| <= |read_offset|;
// that condition is always satisfied because:
//
// (a) If the string is being shortened, |expansion| is zero and
// |write_offset| grows slower than |read_offset|.
//
// (b) If the string is being lengthened, |write_offset| grows faster than
// |read_offset|, but |expansion| is big enough so that |write_offset|
// will only catch up to |read_offset| at the point of the last match.
auto* buffer = &((*str)[0]);
size_t write_offset = first_match;
size_t read_offset = first_match + expansion;
do {
if (replace_length) {
CharTraits::copy(buffer + write_offset, replace_with.data(),
replace_length);
write_offset += replace_length;
}
read_offset += find_length;
// min() clamps StringType::npos (the largest unsigned value) to str_length.
size_t match = std::min(matcher.Find(*str, read_offset), str_length);
size_t length = match - read_offset;
if (length) {
CharTraits::move(buffer + write_offset, buffer + read_offset, length);
write_offset += length;
read_offset += length;
}
} while (read_offset < str_length);
// If we're shortening the string, truncate it now.
str->resize(write_offset);
return true;
}
template <class StringType>
bool ReplaceCharsT(const StringType& input,
BasicStringPiece<StringType> find_any_of_these,
BasicStringPiece<StringType> replace_with,
StringType* output) {
// Commonly, this is called with output and input being the same string; in
// that case, this assignment is inexpensive.
*output = input;
return DoReplaceMatchesAfterOffset(
output, 0, CharacterMatcher<StringType>{find_any_of_these}, replace_with,
ReplaceType::REPLACE_ALL);
}
void ReplaceFirstSubstringAfterOffset(string16* str,
size_t start_offset,
StringPiece16 find_this,
StringPiece16 replace_with) {
DoReplaceMatchesAfterOffset(str, start_offset,
SubstringMatcher<string16>{find_this},
replace_with, ReplaceType::REPLACE_FIRST);
}
void ReplaceFirstSubstringAfterOffset(std::string* str,
size_t start_offset,
StringPiece find_this,
StringPiece replace_with) {
DoReplaceMatchesAfterOffset(str, start_offset,
SubstringMatcher<std::string>{find_this},
replace_with, ReplaceType::REPLACE_FIRST);
}
void ReplaceSubstringsAfterOffset(string16* str,
size_t start_offset,
StringPiece16 find_this,
StringPiece16 replace_with) {
DoReplaceMatchesAfterOffset(str, start_offset,
SubstringMatcher<string16>{find_this},
replace_with, ReplaceType::REPLACE_ALL);
}
void ReplaceSubstringsAfterOffset(std::string* str,
size_t start_offset,
StringPiece find_this,
StringPiece replace_with) {
DoReplaceMatchesAfterOffset(str, start_offset,
SubstringMatcher<std::string>{find_this},
replace_with, ReplaceType::REPLACE_ALL);
}
template <class string_type>
inline typename string_type::value_type* WriteIntoT(string_type* str,
size_t length_with_null) {
DCHECK_GE(length_with_null, 1u);
str->reserve(length_with_null);
str->resize(length_with_null - 1);
return &((*str)[0]);
}
char* WriteInto(std::string* str, size_t length_with_null) {
return WriteIntoT(str, length_with_null);
}
char16* WriteInto(string16* str, size_t length_with_null) {
return WriteIntoT(str, length_with_null);
}
#if defined(_MSC_VER) && !defined(__clang__)
// Work around VC++ code-gen bug. https://crbug.com/804884
#pragma optimize("", off)
#endif
// Generic version for all JoinString overloads. |list_type| must be a sequence
// (std::vector or std::initializer_list) of strings/StringPieces (std::string,
// string16, StringPiece or StringPiece16). |string_type| is either std::string
// or string16.
template <typename list_type, typename string_type>
static string_type JoinStringT(const list_type& parts,
BasicStringPiece<string_type> sep) {
if (parts.size() == 0)
return string_type();
// Pre-allocate the eventual size of the string. Start with the size of all of
// the separators (note that this *assumes* parts.size() > 0).
size_t total_size = (parts.size() - 1) * sep.size();
for (const auto& part : parts)
total_size += part.size();
string_type result;
result.reserve(total_size);
auto iter = parts.begin();
DCHECK(iter != parts.end());
AppendToString(&result, *iter);
++iter;
for (; iter != parts.end(); ++iter) {
sep.AppendToString(&result);
// Using the overloaded AppendToString allows this template function to work
// on both strings and StringPieces without creating an intermediate
// StringPiece object.
AppendToString(&result, *iter);
}
// Sanity-check that we pre-allocated correctly.
DCHECK_EQ(total_size, result.size());
return result;
}
std::string JoinString(const std::vector<std::string>& parts,
StringPiece separator) {
return JoinStringT(parts, separator);
}
string16 JoinString(const std::vector<string16>& parts,
StringPiece16 separator) {
return JoinStringT(parts, separator);
}
#if defined(_MSC_VER) && !defined(__clang__)
// Work around VC++ code-gen bug. https://crbug.com/804884
#pragma optimize("", on)
#endif
std::string JoinString(const std::vector<StringPiece>& parts,
StringPiece separator) {
return JoinStringT(parts, separator);
}
string16 JoinString(const std::vector<StringPiece16>& parts,
StringPiece16 separator) {
return JoinStringT(parts, separator);
}
std::string JoinString(std::initializer_list<StringPiece> parts,
StringPiece separator) {
return JoinStringT(parts, separator);
}
string16 JoinString(std::initializer_list<StringPiece16> parts,
StringPiece16 separator) {
return JoinStringT(parts, separator);
}
template<class FormatStringType, class OutStringType>
OutStringType DoReplaceStringPlaceholders(
const FormatStringType& format_string,
const std::vector<OutStringType>& subst,
std::vector<size_t>* offsets) {
size_t substitutions = subst.size();
DCHECK_LT(substitutions, 10U);
size_t sub_length = 0;
for (const auto& cur : subst)
sub_length += cur.length();
OutStringType formatted;
formatted.reserve(format_string.length() + sub_length);
std::vector<ReplacementOffset> r_offsets;
for (auto i = format_string.begin(); i != format_string.end(); ++i) {
if ('$' == *i) {
if (i + 1 != format_string.end()) {
++i;
if ('$' == *i) {
while (i != format_string.end() && '$' == *i) {
formatted.push_back('$');
++i;
}
--i;
} else {
if (*i < '1' || *i > '9') {
DLOG(ERROR) << "Invalid placeholder: $" << *i;
continue;
}
uintptr_t index = *i - '1';
if (offsets) {
ReplacementOffset r_offset(index,
static_cast<int>(formatted.size()));
r_offsets.insert(
std::upper_bound(r_offsets.begin(), r_offsets.end(), r_offset,
&CompareParameter),
r_offset);
}
if (index < substitutions)
formatted.append(subst.at(index));
}
}
} else {
formatted.push_back(*i);
}
}
if (offsets) {
for (const auto& cur : r_offsets)
offsets->push_back(cur.offset);
}
return formatted;
}
string16 ReplaceStringPlaceholders(const string16& format_string,
const std::vector<string16>& subst,
std::vector<size_t>* offsets) {
return DoReplaceStringPlaceholders(format_string, subst, offsets);
}
std::string ReplaceStringPlaceholders(StringPiece format_string,
const std::vector<std::string>& subst,
std::vector<size_t>* offsets) {
return DoReplaceStringPlaceholders(format_string, subst, offsets);
}
string16 ReplaceStringPlaceholders(const string16& format_string,
const string16& a,
size_t* offset) {
std::vector<size_t> offsets;
std::vector<string16> subst;
subst.push_back(a);
string16 result = ReplaceStringPlaceholders(format_string, subst, &offsets);
DCHECK_EQ(1U, offsets.size());
if (offset)
*offset = offsets[0];
return result;
}
#if defined(OS_WIN) && defined(BASE_STRING16_IS_STD_U16STRING)
TrimPositions TrimWhitespace(WStringPiece input,
TrimPositions positions,
std::wstring* output) {
return TrimStringT(input, WStringPiece(kWhitespaceWide), positions, output);
}
WStringPiece TrimWhitespace(WStringPiece input, TrimPositions positions) {
return TrimStringPieceT(input, WStringPiece(kWhitespaceWide), positions);
}
bool TrimString(WStringPiece input,
WStringPiece trim_chars,
std::wstring* output) {
return TrimStringT(input, trim_chars, TRIM_ALL, output) != TRIM_NONE;
}
WStringPiece TrimString(WStringPiece input,
WStringPiece trim_chars,
TrimPositions positions) {
return TrimStringPieceT(input, trim_chars, positions);
}
#endif
// The following code is compatible with the OpenBSD lcpy interface. See:
// http://www.gratisoft.us/todd/papers/strlcpy.html
// ftp://ftp.openbsd.org/pub/OpenBSD/src/lib/libc/string/{wcs,str}lcpy.c
namespace {
template <typename CHAR>
size_t lcpyT(CHAR* dst, const CHAR* src, size_t dst_size) {
for (size_t i = 0; i < dst_size; ++i) {
if ((dst[i] = src[i]) == 0) // We hit and copied the terminating NULL.
return i;
}
// We were left off at dst_size. We over copied 1 byte. Null terminate.
if (dst_size != 0)
dst[dst_size - 1] = 0;
// Count the rest of the |src|, and return it's length in characters.
while (src[dst_size]) ++dst_size;
return dst_size;
}
} // namespace
size_t strlcpy(char* dst, const char* src, size_t dst_size) {
return lcpyT<char>(dst, src, dst_size);
}
size_t wcslcpy(wchar_t* dst, const wchar_t* src, size_t dst_size) {
return lcpyT<wchar_t>(dst, src, dst_size);
}
} // namespace base