| /* |
| * Copyright (C) 2006, 2007 Apple Inc. All rights reserved. |
| * Copyright (C) 2008 Nokia Corporation and/or its subsidiary(-ies) |
| * |
| * Redistribution and use in source and binary forms, with or without |
| * modification, are permitted provided that the following conditions |
| * are met: |
| * 1. Redistributions of source code must retain the above copyright |
| * notice, this list of conditions and the following disclaimer. |
| * 2. Redistributions in binary form must reproduce the above copyright |
| * notice, this list of conditions and the following disclaimer in the |
| * documentation and/or other materials provided with the distribution. |
| * |
| * THIS SOFTWARE IS PROVIDED BY APPLE COMPUTER, INC. ``AS IS'' AND ANY |
| * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
| * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE COMPUTER, INC. OR |
| * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
| * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
| * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR |
| * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY |
| * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| */ |
| |
| #include "config.h" |
| #include "TextCheckingHelper.h" |
| |
| #include "DocumentMarkerController.h" |
| #include "Range.h" |
| #include "TextCheckerClient.h" |
| #include "TextIterator.h" |
| #include "VisiblePosition.h" |
| #include "visible_units.h" |
| |
| namespace WebCore { |
| |
| static PassRefPtr<Range> expandToParagraphBoundary(PassRefPtr<Range> range) |
| { |
| ExceptionCode ec = 0; |
| RefPtr<Range> paragraphRange = range->cloneRange(ec); |
| setStart(paragraphRange.get(), startOfParagraph(range->startPosition())); |
| setEnd(paragraphRange.get(), endOfParagraph(range->endPosition())); |
| return paragraphRange; |
| } |
| |
| TextCheckingParagraph::TextCheckingParagraph(PassRefPtr<Range> checkingRange) |
| : m_checkingRange(checkingRange) |
| , m_checkingStart(-1) |
| , m_checkingEnd(-1) |
| , m_checkingLength(-1) |
| { |
| } |
| |
| TextCheckingParagraph::~TextCheckingParagraph() |
| { |
| } |
| |
| void TextCheckingParagraph::expandRangeToNextEnd() |
| { |
| ASSERT(m_checkingRange); |
| setEnd(paragraphRange().get(), endOfParagraph(startOfNextParagraph(paragraphRange()->startPosition()))); |
| invalidateParagraphRangeValues(); |
| } |
| |
| void TextCheckingParagraph::invalidateParagraphRangeValues() |
| { |
| m_checkingStart = m_checkingEnd = -1; |
| m_offsetAsRange = 0; |
| m_text = String(); |
| } |
| |
| int TextCheckingParagraph::rangeLength() const |
| { |
| ASSERT(m_checkingRange); |
| return TextIterator::rangeLength(paragraphRange().get()); |
| } |
| |
| PassRefPtr<Range> TextCheckingParagraph::paragraphRange() const |
| { |
| ASSERT(m_checkingRange); |
| if (!m_paragraphRange) |
| m_paragraphRange = expandToParagraphBoundary(checkingRange()); |
| return m_paragraphRange; |
| } |
| |
| PassRefPtr<Range> TextCheckingParagraph::subrange(int characterOffset, int characterCount) const |
| { |
| ASSERT(m_checkingRange); |
| return TextIterator::subrange(paragraphRange().get(), characterOffset, characterCount); |
| } |
| |
| int TextCheckingParagraph::offsetTo(const Position& position, ExceptionCode& ec) const |
| { |
| ASSERT(m_checkingRange); |
| RefPtr<Range> range = offsetAsRange(); |
| range->setEnd(position.containerNode(), position.computeOffsetInContainerNode(), ec); |
| if (ec) |
| return 0; |
| return TextIterator::rangeLength(range.get()); |
| } |
| |
| bool TextCheckingParagraph::isEmpty() const |
| { |
| // Both predicates should have same result, but we check both just for sure. |
| // We need to investigate to remove this redundancy. |
| return isRangeEmpty() || isTextEmpty(); |
| } |
| |
| PassRefPtr<Range> TextCheckingParagraph::offsetAsRange() const |
| { |
| ASSERT(m_checkingRange); |
| if (!m_offsetAsRange) { |
| ExceptionCode ec = 0; |
| m_offsetAsRange = Range::create(paragraphRange()->startContainer(ec)->document(), paragraphRange()->startPosition(), checkingRange()->startPosition()); |
| } |
| |
| return m_offsetAsRange; |
| } |
| |
| const String& TextCheckingParagraph::text() const |
| { |
| ASSERT(m_checkingRange); |
| if (m_text.isEmpty()) |
| m_text = plainText(paragraphRange().get()); |
| return m_text; |
| } |
| |
| int TextCheckingParagraph::checkingStart() const |
| { |
| ASSERT(m_checkingRange); |
| if (m_checkingStart == -1) |
| m_checkingStart = TextIterator::rangeLength(offsetAsRange().get()); |
| return m_checkingStart; |
| } |
| |
| int TextCheckingParagraph::checkingEnd() const |
| { |
| ASSERT(m_checkingRange); |
| if (m_checkingEnd == -1) |
| m_checkingEnd = checkingStart() + TextIterator::rangeLength(checkingRange().get()); |
| return m_checkingEnd; |
| } |
| |
| int TextCheckingParagraph::checkingLength() const |
| { |
| ASSERT(m_checkingRange); |
| if (-1 == m_checkingLength) |
| m_checkingLength = TextIterator::rangeLength(checkingRange().get()); |
| return m_checkingLength; |
| } |
| |
| TextCheckingHelper::TextCheckingHelper(EditorClient* client, PassRefPtr<Range> range) |
| : m_client(client) |
| , m_range(range) |
| { |
| ASSERT_ARG(m_client, m_client); |
| ASSERT_ARG(m_range, m_range); |
| } |
| |
| TextCheckingHelper::~TextCheckingHelper() |
| { |
| } |
| |
| String TextCheckingHelper::findFirstMisspelling(int& firstMisspellingOffset, bool markAll, RefPtr<Range>& firstMisspellingRange) |
| { |
| WordAwareIterator it(m_range.get()); |
| firstMisspellingOffset = 0; |
| |
| String firstMisspelling; |
| int currentChunkOffset = 0; |
| |
| while (!it.atEnd()) { |
| const UChar* chars = it.characters(); |
| int len = it.length(); |
| |
| // Skip some work for one-space-char hunks |
| if (!(len == 1 && chars[0] == ' ')) { |
| |
| int misspellingLocation = -1; |
| int misspellingLength = 0; |
| m_client->textChecker()->checkSpellingOfString(chars, len, &misspellingLocation, &misspellingLength); |
| |
| // 5490627 shows that there was some code path here where the String constructor below crashes. |
| // We don't know exactly what combination of bad input caused this, so we're making this much |
| // more robust against bad input on release builds. |
| ASSERT(misspellingLength >= 0); |
| ASSERT(misspellingLocation >= -1); |
| ASSERT(!misspellingLength || misspellingLocation >= 0); |
| ASSERT(misspellingLocation < len); |
| ASSERT(misspellingLength <= len); |
| ASSERT(misspellingLocation + misspellingLength <= len); |
| |
| if (misspellingLocation >= 0 && misspellingLength > 0 && misspellingLocation < len && misspellingLength <= len && misspellingLocation + misspellingLength <= len) { |
| |
| // Compute range of misspelled word |
| RefPtr<Range> misspellingRange = TextIterator::subrange(m_range.get(), currentChunkOffset + misspellingLocation, misspellingLength); |
| |
| // Remember first-encountered misspelling and its offset. |
| if (!firstMisspelling) { |
| firstMisspellingOffset = currentChunkOffset + misspellingLocation; |
| firstMisspelling = String(chars + misspellingLocation, misspellingLength); |
| firstMisspellingRange = misspellingRange; |
| } |
| |
| // Store marker for misspelled word. |
| ExceptionCode ec = 0; |
| misspellingRange->startContainer(ec)->document()->markers()->addMarker(misspellingRange.get(), DocumentMarker::Spelling); |
| ASSERT(!ec); |
| |
| // Bail out if we're marking only the first misspelling, and not all instances. |
| if (!markAll) |
| break; |
| } |
| } |
| |
| currentChunkOffset += len; |
| it.advance(); |
| } |
| |
| return firstMisspelling; |
| } |
| |
| String TextCheckingHelper::findFirstMisspellingOrBadGrammar(bool checkGrammar, bool& outIsSpelling, int& outFirstFoundOffset, GrammarDetail& outGrammarDetail) |
| { |
| #if USE(UNIFIED_TEXT_CHECKING) |
| String firstFoundItem; |
| String misspelledWord; |
| String badGrammarPhrase; |
| ExceptionCode ec = 0; |
| |
| // Initialize out parameters; these will be updated if we find something to return. |
| outIsSpelling = true; |
| outFirstFoundOffset = 0; |
| outGrammarDetail.location = -1; |
| outGrammarDetail.length = 0; |
| outGrammarDetail.guesses.clear(); |
| outGrammarDetail.userDescription = ""; |
| |
| // Expand the search range to encompass entire paragraphs, since text checking needs that much context. |
| // Determine the character offset from the start of the paragraph to the start of the original search range, |
| // since we will want to ignore results in this area. |
| RefPtr<Range> paragraphRange = m_range->cloneRange(ec); |
| setStart(paragraphRange.get(), startOfParagraph(m_range->startPosition())); |
| int totalRangeLength = TextIterator::rangeLength(paragraphRange.get()); |
| setEnd(paragraphRange.get(), endOfParagraph(m_range->startPosition())); |
| |
| RefPtr<Range> offsetAsRange = Range::create(paragraphRange->startContainer(ec)->document(), paragraphRange->startPosition(), m_range->startPosition()); |
| int rangeStartOffset = TextIterator::rangeLength(offsetAsRange.get()); |
| int totalLengthProcessed = 0; |
| |
| bool firstIteration = true; |
| bool lastIteration = false; |
| while (totalLengthProcessed < totalRangeLength) { |
| // Iterate through the search range by paragraphs, checking each one for spelling and grammar. |
| int currentLength = TextIterator::rangeLength(paragraphRange.get()); |
| int currentStartOffset = firstIteration ? rangeStartOffset : 0; |
| int currentEndOffset = currentLength; |
| if (inSameParagraph(paragraphRange->startPosition(), m_range->endPosition())) { |
| // Determine the character offset from the end of the original search range to the end of the paragraph, |
| // since we will want to ignore results in this area. |
| RefPtr<Range> endOffsetAsRange = Range::create(paragraphRange->startContainer(ec)->document(), paragraphRange->startPosition(), m_range->endPosition()); |
| currentEndOffset = TextIterator::rangeLength(endOffsetAsRange.get()); |
| lastIteration = true; |
| } |
| if (currentStartOffset < currentEndOffset) { |
| String paragraphString = plainText(paragraphRange.get()); |
| if (paragraphString.length() > 0) { |
| bool foundGrammar = false; |
| int spellingLocation = 0; |
| int grammarPhraseLocation = 0; |
| int grammarDetailLocation = 0; |
| unsigned grammarDetailIndex = 0; |
| |
| Vector<TextCheckingResult> results; |
| TextCheckingTypeMask checkingTypes = checkGrammar ? (TextCheckingTypeSpelling | TextCheckingTypeGrammar) : TextCheckingTypeSpelling; |
| m_client->textChecker()->checkTextOfParagraph(paragraphString.characters(), paragraphString.length(), checkingTypes, results); |
| |
| for (unsigned i = 0; i < results.size(); i++) { |
| const TextCheckingResult* result = &results[i]; |
| if (result->type == TextCheckingTypeSpelling && result->location >= currentStartOffset && result->location + result->length <= currentEndOffset) { |
| ASSERT(result->length > 0 && result->location >= 0); |
| spellingLocation = result->location; |
| misspelledWord = paragraphString.substring(result->location, result->length); |
| ASSERT(misspelledWord.length()); |
| break; |
| } |
| if (checkGrammar && result->type == TextCheckingTypeGrammar && result->location < currentEndOffset && result->location + result->length > currentStartOffset) { |
| ASSERT(result->length > 0 && result->location >= 0); |
| // We can't stop after the first grammar result, since there might still be a spelling result after |
| // it begins but before the first detail in it, but we can stop if we find a second grammar result. |
| if (foundGrammar) |
| break; |
| for (unsigned j = 0; j < result->details.size(); j++) { |
| const GrammarDetail* detail = &result->details[j]; |
| ASSERT(detail->length > 0 && detail->location >= 0); |
| if (result->location + detail->location >= currentStartOffset && result->location + detail->location + detail->length <= currentEndOffset && (!foundGrammar || result->location + detail->location < grammarDetailLocation)) { |
| grammarDetailIndex = j; |
| grammarDetailLocation = result->location + detail->location; |
| foundGrammar = true; |
| } |
| } |
| if (foundGrammar) { |
| grammarPhraseLocation = result->location; |
| outGrammarDetail = result->details[grammarDetailIndex]; |
| badGrammarPhrase = paragraphString.substring(result->location, result->length); |
| ASSERT(badGrammarPhrase.length()); |
| } |
| } |
| } |
| |
| if (!misspelledWord.isEmpty() && (!checkGrammar || badGrammarPhrase.isEmpty() || spellingLocation <= grammarDetailLocation)) { |
| int spellingOffset = spellingLocation - currentStartOffset; |
| if (!firstIteration) { |
| RefPtr<Range> paragraphOffsetAsRange = Range::create(paragraphRange->startContainer(ec)->document(), m_range->startPosition(), paragraphRange->startPosition()); |
| spellingOffset += TextIterator::rangeLength(paragraphOffsetAsRange.get()); |
| } |
| outIsSpelling = true; |
| outFirstFoundOffset = spellingOffset; |
| firstFoundItem = misspelledWord; |
| break; |
| } |
| if (checkGrammar && !badGrammarPhrase.isEmpty()) { |
| int grammarPhraseOffset = grammarPhraseLocation - currentStartOffset; |
| if (!firstIteration) { |
| RefPtr<Range> paragraphOffsetAsRange = Range::create(paragraphRange->startContainer(ec)->document(), m_range->startPosition(), paragraphRange->startPosition()); |
| grammarPhraseOffset += TextIterator::rangeLength(paragraphOffsetAsRange.get()); |
| } |
| outIsSpelling = false; |
| outFirstFoundOffset = grammarPhraseOffset; |
| firstFoundItem = badGrammarPhrase; |
| break; |
| } |
| } |
| } |
| if (lastIteration || totalLengthProcessed + currentLength >= totalRangeLength) |
| break; |
| VisiblePosition newParagraphStart = startOfNextParagraph(paragraphRange->endPosition()); |
| setStart(paragraphRange.get(), newParagraphStart); |
| setEnd(paragraphRange.get(), endOfParagraph(newParagraphStart)); |
| firstIteration = false; |
| totalLengthProcessed += currentLength; |
| } |
| return firstFoundItem; |
| #else |
| ASSERT_NOT_REACHED(); |
| UNUSED_PARAM(checkGrammar); |
| UNUSED_PARAM(outIsSpelling); |
| UNUSED_PARAM(outFirstFoundOffset); |
| UNUSED_PARAM(outGrammarDetail); |
| return ""; |
| #endif // USE(UNIFIED_TEXT_CHECKING) |
| } |
| |
| int TextCheckingHelper::findFirstGrammarDetail(const Vector<GrammarDetail>& grammarDetails, int badGrammarPhraseLocation, int /*badGrammarPhraseLength*/, int startOffset, int endOffset, bool markAll) |
| { |
| #if USE(GRAMMAR_CHECKING) |
| // Found some bad grammar. Find the earliest detail range that starts in our search range (if any). |
| // Optionally add a DocumentMarker for each detail in the range. |
| int earliestDetailLocationSoFar = -1; |
| int earliestDetailIndex = -1; |
| for (unsigned i = 0; i < grammarDetails.size(); i++) { |
| const GrammarDetail* detail = &grammarDetails[i]; |
| ASSERT(detail->length > 0 && detail->location >= 0); |
| |
| int detailStartOffsetInParagraph = badGrammarPhraseLocation + detail->location; |
| |
| // Skip this detail if it starts before the original search range |
| if (detailStartOffsetInParagraph < startOffset) |
| continue; |
| |
| // Skip this detail if it starts after the original search range |
| if (detailStartOffsetInParagraph >= endOffset) |
| continue; |
| |
| if (markAll) { |
| RefPtr<Range> badGrammarRange = TextIterator::subrange(m_range.get(), badGrammarPhraseLocation - startOffset + detail->location, detail->length); |
| ExceptionCode ec = 0; |
| badGrammarRange->startContainer(ec)->document()->markers()->addMarker(badGrammarRange.get(), DocumentMarker::Grammar, detail->userDescription); |
| ASSERT(!ec); |
| } |
| |
| // Remember this detail only if it's earlier than our current candidate (the details aren't in a guaranteed order) |
| if (earliestDetailIndex < 0 || earliestDetailLocationSoFar > detail->location) { |
| earliestDetailIndex = i; |
| earliestDetailLocationSoFar = detail->location; |
| } |
| } |
| |
| return earliestDetailIndex; |
| #else |
| ASSERT_NOT_REACHED(); |
| UNUSED_PARAM(grammarDetails); |
| UNUSED_PARAM(badGrammarPhraseLocation); |
| UNUSED_PARAM(startOffset); |
| UNUSED_PARAM(endOffset); |
| UNUSED_PARAM(markAll); |
| return 0; |
| #endif |
| } |
| |
| String TextCheckingHelper::findFirstBadGrammar(GrammarDetail& outGrammarDetail, int& outGrammarPhraseOffset, bool markAll) |
| { |
| ASSERT(WTF_USE_GRAMMAR_CHECKING); |
| // Initialize out parameters; these will be updated if we find something to return. |
| outGrammarDetail.location = -1; |
| outGrammarDetail.length = 0; |
| outGrammarDetail.guesses.clear(); |
| outGrammarDetail.userDescription = ""; |
| outGrammarPhraseOffset = 0; |
| |
| String firstBadGrammarPhrase; |
| |
| // Expand the search range to encompass entire paragraphs, since grammar checking needs that much context. |
| // Determine the character offset from the start of the paragraph to the start of the original search range, |
| // since we will want to ignore results in this area. |
| TextCheckingParagraph paragraph(m_range); |
| |
| // Start checking from beginning of paragraph, but skip past results that occur before the start of the original search range. |
| int startOffset = 0; |
| while (startOffset < paragraph.checkingEnd()) { |
| Vector<GrammarDetail> grammarDetails; |
| int badGrammarPhraseLocation = -1; |
| int badGrammarPhraseLength = 0; |
| m_client->textChecker()->checkGrammarOfString(paragraph.textCharacters() + startOffset, paragraph.textLength() - startOffset, grammarDetails, &badGrammarPhraseLocation, &badGrammarPhraseLength); |
| |
| if (!badGrammarPhraseLength) { |
| ASSERT(badGrammarPhraseLocation == -1); |
| return String(); |
| } |
| |
| ASSERT(badGrammarPhraseLocation >= 0); |
| badGrammarPhraseLocation += startOffset; |
| |
| |
| // Found some bad grammar. Find the earliest detail range that starts in our search range (if any). |
| int badGrammarIndex = findFirstGrammarDetail(grammarDetails, badGrammarPhraseLocation, badGrammarPhraseLength, paragraph.checkingStart(), paragraph.checkingEnd(), markAll); |
| if (badGrammarIndex >= 0) { |
| ASSERT(static_cast<unsigned>(badGrammarIndex) < grammarDetails.size()); |
| outGrammarDetail = grammarDetails[badGrammarIndex]; |
| } |
| |
| // If we found a detail in range, then we have found the first bad phrase (unless we found one earlier but |
| // kept going so we could mark all instances). |
| if (badGrammarIndex >= 0 && firstBadGrammarPhrase.isEmpty()) { |
| outGrammarPhraseOffset = badGrammarPhraseLocation - paragraph.checkingStart(); |
| firstBadGrammarPhrase = paragraph.textSubstring(badGrammarPhraseLocation, badGrammarPhraseLength); |
| |
| // Found one. We're done now, unless we're marking each instance. |
| if (!markAll) |
| break; |
| } |
| |
| // These results were all between the start of the paragraph and the start of the search range; look |
| // beyond this phrase. |
| startOffset = badGrammarPhraseLocation + badGrammarPhraseLength; |
| } |
| |
| return firstBadGrammarPhrase; |
| } |
| |
| |
| bool TextCheckingHelper::isUngrammatical(Vector<String>& guessesVector) const |
| { |
| ASSERT(WTF_USE_GRAMMAR_CHECKING); |
| if (!m_client) |
| return false; |
| |
| ExceptionCode ec; |
| if (!m_range || m_range->collapsed(ec)) |
| return false; |
| |
| // Returns true only if the passed range exactly corresponds to a bad grammar detail range. This is analogous |
| // to isSelectionMisspelled. It's not good enough for there to be some bad grammar somewhere in the range, |
| // or overlapping the range; the ranges must exactly match. |
| guessesVector.clear(); |
| int grammarPhraseOffset; |
| |
| GrammarDetail grammarDetail; |
| String badGrammarPhrase = const_cast<TextCheckingHelper*>(this)->findFirstBadGrammar(grammarDetail, grammarPhraseOffset, false); |
| |
| // No bad grammar in these parts at all. |
| if (badGrammarPhrase.isEmpty()) |
| return false; |
| |
| // Bad grammar, but phrase (e.g. sentence) starts beyond start of range. |
| if (grammarPhraseOffset > 0) |
| return false; |
| |
| ASSERT(grammarDetail.location >= 0 && grammarDetail.length > 0); |
| |
| // Bad grammar, but start of detail (e.g. ungrammatical word) doesn't match start of range |
| if (grammarDetail.location + grammarPhraseOffset) |
| return false; |
| |
| // Bad grammar at start of range, but end of bad grammar is before or after end of range |
| if (grammarDetail.length != TextIterator::rangeLength(m_range.get())) |
| return false; |
| |
| // Update the spelling panel to be displaying this error (whether or not the spelling panel is on screen). |
| // This is necessary to make a subsequent call to [NSSpellChecker ignoreWord:inSpellDocumentWithTag:] work |
| // correctly; that call behaves differently based on whether the spelling panel is displaying a misspelling |
| // or a grammar error. |
| m_client->updateSpellingUIWithGrammarString(badGrammarPhrase, grammarDetail); |
| |
| return true; |
| } |
| |
| Vector<String> TextCheckingHelper::guessesForMisspelledOrUngrammaticalRange(bool checkGrammar, bool& misspelled, bool& ungrammatical) const |
| { |
| #if USE(UNIFIED_TEXT_CHECKING) |
| Vector<String> guesses; |
| ExceptionCode ec; |
| misspelled = false; |
| ungrammatical = false; |
| |
| if (!m_client || !m_range || m_range->collapsed(ec)) |
| return guesses; |
| |
| // Expand the range to encompass entire paragraphs, since text checking needs that much context. |
| TextCheckingParagraph paragraph(m_range); |
| if (paragraph.isEmpty()) |
| return guesses; |
| |
| Vector<TextCheckingResult> results; |
| TextCheckingTypeMask checkingTypes = checkGrammar ? (TextCheckingTypeSpelling | TextCheckingTypeGrammar) : TextCheckingTypeSpelling; |
| m_client->textChecker()->checkTextOfParagraph(paragraph.textCharacters(), paragraph.textLength(), checkingTypes, results); |
| |
| for (unsigned i = 0; i < results.size(); i++) { |
| const TextCheckingResult* result = &results[i]; |
| if (result->type == TextCheckingTypeSpelling && paragraph.checkingRangeMatches(result->location, result->length)) { |
| String misspelledWord = paragraph.checkingSubstring(); |
| ASSERT(misspelledWord.length()); |
| m_client->textChecker()->getGuessesForWord(misspelledWord, String(), guesses); |
| m_client->updateSpellingUIWithMisspelledWord(misspelledWord); |
| misspelled = true; |
| return guesses; |
| } |
| } |
| |
| if (!checkGrammar) |
| return guesses; |
| |
| for (unsigned i = 0; i < results.size(); i++) { |
| const TextCheckingResult* result = &results[i]; |
| if (result->type == TextCheckingTypeGrammar && paragraph.isCheckingRangeCoveredBy(result->location, result->length)) { |
| for (unsigned j = 0; j < result->details.size(); j++) { |
| const GrammarDetail* detail = &result->details[j]; |
| ASSERT(detail->length > 0 && detail->location >= 0); |
| if (paragraph.checkingRangeMatches(result->location + detail->location, detail->length)) { |
| String badGrammarPhrase = paragraph.textSubstring(result->location, result->length); |
| ASSERT(badGrammarPhrase.length()); |
| for (unsigned k = 0; k < detail->guesses.size(); k++) |
| guesses.append(detail->guesses[k]); |
| m_client->updateSpellingUIWithGrammarString(badGrammarPhrase, *detail); |
| ungrammatical = true; |
| return guesses; |
| } |
| } |
| } |
| } |
| return guesses; |
| #else |
| ASSERT_NOT_REACHED(); |
| UNUSED_PARAM(checkGrammar); |
| UNUSED_PARAM(misspelled); |
| UNUSED_PARAM(ungrammatical); |
| return Vector<String>(); |
| #endif // USE(UNIFIED_TEXT_CHECKING) |
| } |
| |
| |
| void TextCheckingHelper::markAllMisspellings(RefPtr<Range>& firstMisspellingRange) |
| { |
| // Use the "markAll" feature of findFirstMisspelling. Ignore the return value and the "out parameter"; |
| // all we need to do is mark every instance. |
| int ignoredOffset; |
| findFirstMisspelling(ignoredOffset, true, firstMisspellingRange); |
| } |
| |
| void TextCheckingHelper::markAllBadGrammar() |
| { |
| ASSERT(WTF_USE_GRAMMAR_CHECKING); |
| // Use the "markAll" feature of ofindFirstBadGrammar. Ignore the return value and "out parameters"; all we need to |
| // do is mark every instance. |
| GrammarDetail ignoredGrammarDetail; |
| int ignoredOffset; |
| findFirstBadGrammar(ignoredGrammarDetail, ignoredOffset, true); |
| } |
| |
| } |