| // Copyright 2012 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 "ui/gfx/text_utils.h" |
| |
| #include "base/i18n/char_iterator.h" |
| #include "base/logging.h" |
| #include "base/numerics/safe_conversions.h" |
| #include "third_party/icu/source/common/unicode/uchar.h" |
| #include "third_party/icu/source/common/unicode/utf16.h" |
| |
| namespace gfx { |
| |
| namespace { |
| |
| // Returns true if the code point |c| is a combining mark character in Unicode. |
| bool CharIsMark(UChar32 c) { |
| int8_t char_type = u_charType(c); |
| return char_type == U_NON_SPACING_MARK || char_type == U_ENCLOSING_MARK || |
| char_type == U_COMBINING_SPACING_MARK; |
| } |
| |
| // Gets the code point of |str| at the given code unit position |index|. If |
| // |index| is a surrogate code unit, returns the whole code point (unless the |
| // code unit is unpaired, in which case it just returns the surrogate value). |
| UChar32 GetCodePointAt(const base::string16& str, size_t index) { |
| UChar32 c; |
| U16_GET(str.data(), 0, index, str.size(), c); |
| return c; |
| } |
| |
| } // namespace |
| |
| base::string16 RemoveAcceleratorChar(const base::string16& s, |
| base::char16 accelerator_char, |
| int* accelerated_char_pos, |
| int* accelerated_char_span) { |
| bool escaped = false; |
| ptrdiff_t last_char_pos = -1; |
| int last_char_span = 0; |
| base::i18n::UTF16CharIterator chars(&s); |
| base::string16 accelerator_removed; |
| |
| accelerator_removed.reserve(s.size()); |
| while (!chars.end()) { |
| int32 c = chars.get(); |
| int array_pos = chars.array_pos(); |
| chars.Advance(); |
| |
| if (c != accelerator_char || escaped) { |
| int span = chars.array_pos() - array_pos; |
| if (escaped && c != accelerator_char) { |
| last_char_pos = accelerator_removed.size(); |
| last_char_span = span; |
| } |
| for (int i = 0; i < span; i++) |
| accelerator_removed.push_back(s[array_pos + i]); |
| escaped = false; |
| } else { |
| escaped = true; |
| } |
| } |
| |
| if (accelerated_char_pos) |
| *accelerated_char_pos = last_char_pos; |
| if (accelerated_char_span) |
| *accelerated_char_span = last_char_span; |
| |
| return accelerator_removed; |
| } |
| |
| size_t FindValidBoundaryBefore(const base::string16& text, size_t index) { |
| size_t length = text.length(); |
| DCHECK_LE(index, length); |
| if (index == length) |
| return index; |
| |
| // If |index| straddles a combining character sequence, go back until we find |
| // a base character. |
| while (index > 0 && CharIsMark(GetCodePointAt(text, index))) |
| --index; |
| |
| // If |index| straddles a UTF-16 surrogate pair, go back. |
| U16_SET_CP_START(text.data(), 0, index); |
| return index; |
| } |
| |
| size_t FindValidBoundaryAfter(const base::string16& text, size_t index) { |
| DCHECK_LE(index, text.length()); |
| if (index == text.length()) |
| return index; |
| |
| int32_t text_index = base::checked_cast<int32_t>(index); |
| int32_t text_length = base::checked_cast<int32_t>(text.length()); |
| |
| // If |index| straddles a combining character sequence, go forward until we |
| // find a base character. |
| while (text_index < text_length && |
| CharIsMark(GetCodePointAt(text, text_index))) { |
| ++text_index; |
| } |
| |
| // If |index| straddles a UTF-16 surrogate pair, go forward. |
| U16_SET_CP_LIMIT(text.data(), 0, text_index, text_length); |
| return static_cast<size_t>(text_index); |
| } |
| |
| } // namespace gfx |