1 /*
2 * Copyright (C) 2004, 2005, 2006, 2007, 2008, 2009, 2010 Apple Inc. All rights reserved.
3 * Copyright (C) 2005 Alexey Proskuryakov.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY APPLE COMPUTER, INC. ``AS IS'' AND ANY
15 * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
17 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE COMPUTER, INC. OR
18 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
19 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
20 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
21 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
22 * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
24 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 */
26
27 #include "config.h"
28 #include "TextIterator.h"
29
30 #include "CharacterNames.h"
31 #include "Document.h"
32 #include "HTMLElement.h"
33 #include "HTMLNames.h"
34 #include "htmlediting.h"
35 #include "InlineTextBox.h"
36 #include "Range.h"
37 #include "RenderTableCell.h"
38 #include "RenderTableRow.h"
39 #include "RenderTextControl.h"
40 #include "VisiblePosition.h"
41 #include "visible_units.h"
42
43 #if USE(ICU_UNICODE) && !UCONFIG_NO_COLLATION
44 #include "TextBreakIteratorInternalICU.h"
45 #include <unicode/usearch.h>
46 #endif
47
48 using namespace WTF::Unicode;
49 using namespace std;
50
51 namespace WebCore {
52
53 using namespace HTMLNames;
54
55 // Buffer that knows how to compare with a search target.
56 // Keeps enough of the previous text to be able to search in the future, but no more.
57 // Non-breaking spaces are always equal to normal spaces.
58 // Case folding is also done if <isCaseSensitive> is false.
59 class SearchBuffer : public Noncopyable {
60 public:
61 SearchBuffer(const String& target, bool isCaseSensitive);
62 ~SearchBuffer();
63
64 // Returns number of characters appended; guaranteed to be in the range [1, length].
65 size_t append(const UChar*, size_t length);
66 void reachedBreak();
67
68 // Result is the size in characters of what was found.
69 // And <startOffset> is the number of characters back to the start of what was found.
70 size_t search(size_t& startOffset);
71 bool atBreak() const;
72
73 #if USE(ICU_UNICODE) && !UCONFIG_NO_COLLATION
74
75 private:
76 bool isBadMatch(const UChar*, size_t length) const;
77
78 String m_target;
79 Vector<UChar> m_buffer;
80 size_t m_overlap;
81 bool m_atBreak;
82
83 bool m_targetRequiresKanaWorkaround;
84 Vector<UChar> m_normalizedTarget;
85 mutable Vector<UChar> m_normalizedMatch;
86
87 #else
88
89 private:
90 void append(UChar, bool isCharacterStart);
91 size_t length() const;
92
93 String m_target;
94 bool m_isCaseSensitive;
95
96 Vector<UChar> m_buffer;
97 Vector<bool> m_isCharacterStartBuffer;
98 bool m_isBufferFull;
99 size_t m_cursor;
100
101 #endif
102 };
103
104 // --------
105
106 static const unsigned bitsInWord = sizeof(unsigned) * 8;
107 static const unsigned bitInWordMask = bitsInWord - 1;
108
BitStack()109 BitStack::BitStack()
110 : m_size(0)
111 {
112 }
113
push(bool bit)114 void BitStack::push(bool bit)
115 {
116 unsigned index = m_size / bitsInWord;
117 unsigned shift = m_size & bitInWordMask;
118 if (!shift && index == m_words.size()) {
119 m_words.grow(index + 1);
120 m_words[index] = 0;
121 }
122 unsigned& word = m_words[index];
123 unsigned mask = 1U << shift;
124 if (bit)
125 word |= mask;
126 else
127 word &= ~mask;
128 ++m_size;
129 }
130
pop()131 void BitStack::pop()
132 {
133 if (m_size)
134 --m_size;
135 }
136
top() const137 bool BitStack::top() const
138 {
139 if (!m_size)
140 return false;
141 unsigned shift = (m_size - 1) & bitInWordMask;
142 return m_words.last() & (1U << shift);
143 }
144
size() const145 unsigned BitStack::size() const
146 {
147 return m_size;
148 }
149
150 // --------
151
parentCrossingShadowBoundaries(Node * node)152 static inline Node* parentCrossingShadowBoundaries(Node* node)
153 {
154 if (Node* parent = node->parentNode())
155 return parent;
156 return node->shadowParentNode();
157 }
158
159 #if !ASSERT_DISABLED
160
depthCrossingShadowBoundaries(Node * node)161 static unsigned depthCrossingShadowBoundaries(Node* node)
162 {
163 unsigned depth = 0;
164 for (Node* parent = parentCrossingShadowBoundaries(node); parent; parent = parentCrossingShadowBoundaries(parent))
165 ++depth;
166 return depth;
167 }
168
169 #endif
170
171 // This function is like Range::pastLastNode, except for the fact that it can climb up out of shadow trees.
nextInPreOrderCrossingShadowBoundaries(Node * rangeEndContainer,int rangeEndOffset)172 static Node* nextInPreOrderCrossingShadowBoundaries(Node* rangeEndContainer, int rangeEndOffset)
173 {
174 if (!rangeEndContainer)
175 return 0;
176 if (rangeEndOffset >= 0 && !rangeEndContainer->offsetInCharacters()) {
177 if (Node* next = rangeEndContainer->childNode(rangeEndOffset))
178 return next;
179 }
180 for (Node* node = rangeEndContainer; node; node = parentCrossingShadowBoundaries(node)) {
181 if (Node* next = node->nextSibling())
182 return next;
183 }
184 return 0;
185 }
186
previousInPostOrderCrossingShadowBoundaries(Node * rangeStartContainer,int rangeStartOffset)187 static Node* previousInPostOrderCrossingShadowBoundaries(Node* rangeStartContainer, int rangeStartOffset)
188 {
189 if (!rangeStartContainer)
190 return 0;
191 if (rangeStartOffset > 0 && !rangeStartContainer->offsetInCharacters()) {
192 if (Node* previous = rangeStartContainer->childNode(rangeStartOffset - 1))
193 return previous;
194 }
195 for (Node* node = rangeStartContainer; node; node = parentCrossingShadowBoundaries(node)) {
196 if (Node* previous = node->previousSibling())
197 return previous;
198 }
199 return 0;
200 }
201
202 // --------
203
fullyClipsContents(Node * node)204 static inline bool fullyClipsContents(Node* node)
205 {
206 RenderObject* renderer = node->renderer();
207 if (!renderer || !renderer->isBox() || !renderer->hasOverflowClip())
208 return false;
209 return toRenderBox(renderer)->size().isEmpty();
210 }
211
ignoresContainerClip(Node * node)212 static inline bool ignoresContainerClip(Node* node)
213 {
214 RenderObject* renderer = node->renderer();
215 if (!renderer || renderer->isText())
216 return false;
217 EPosition position = renderer->style()->position();
218 return position == AbsolutePosition || position == FixedPosition;
219 }
220
pushFullyClippedState(BitStack & stack,Node * node)221 static void pushFullyClippedState(BitStack& stack, Node* node)
222 {
223 ASSERT(stack.size() == depthCrossingShadowBoundaries(node));
224
225 // Push true if this node full clips its contents, or if a parent already has fully
226 // clipped and this is not a node that ignores its container's clip.
227 stack.push(fullyClipsContents(node) || (stack.top() && !ignoresContainerClip(node)));
228 }
229
setUpFullyClippedStack(BitStack & stack,Node * node)230 static void setUpFullyClippedStack(BitStack& stack, Node* node)
231 {
232 // Put the nodes in a vector so we can iterate in reverse order.
233 Vector<Node*, 100> ancestry;
234 for (Node* parent = parentCrossingShadowBoundaries(node); parent; parent = parentCrossingShadowBoundaries(parent))
235 ancestry.append(parent);
236
237 // Call pushFullyClippedState on each node starting with the earliest ancestor.
238 size_t size = ancestry.size();
239 for (size_t i = 0; i < size; ++i)
240 pushFullyClippedState(stack, ancestry[size - i - 1]);
241 pushFullyClippedState(stack, node);
242
243 ASSERT(stack.size() == 1 + depthCrossingShadowBoundaries(node));
244 }
245
246 // --------
247
TextIterator()248 TextIterator::TextIterator()
249 : m_startContainer(0)
250 , m_startOffset(0)
251 , m_endContainer(0)
252 , m_endOffset(0)
253 , m_positionNode(0)
254 , m_textCharacters(0)
255 , m_textLength(0)
256 , m_lastCharacter(0)
257 , m_emitCharactersBetweenAllVisiblePositions(false)
258 , m_enterTextControls(false)
259 {
260 }
261
TextIterator(const Range * r,bool emitCharactersBetweenAllVisiblePositions,bool enterTextControls)262 TextIterator::TextIterator(const Range* r, bool emitCharactersBetweenAllVisiblePositions, bool enterTextControls)
263 : m_startContainer(0)
264 , m_startOffset(0)
265 , m_endContainer(0)
266 , m_endOffset(0)
267 , m_positionNode(0)
268 , m_textCharacters(0)
269 , m_textLength(0)
270 , m_emitCharactersBetweenAllVisiblePositions(emitCharactersBetweenAllVisiblePositions)
271 , m_enterTextControls(enterTextControls)
272 {
273 if (!r)
274 return;
275
276 // get and validate the range endpoints
277 Node* startContainer = r->startContainer();
278 if (!startContainer)
279 return;
280 int startOffset = r->startOffset();
281 Node* endContainer = r->endContainer();
282 int endOffset = r->endOffset();
283
284 // Callers should be handing us well-formed ranges. If we discover that this isn't
285 // the case, we could consider changing this assertion to an early return.
286 ASSERT(r->boundaryPointsValid());
287
288 // remember range - this does not change
289 m_startContainer = startContainer;
290 m_startOffset = startOffset;
291 m_endContainer = endContainer;
292 m_endOffset = endOffset;
293
294 // set up the current node for processing
295 m_node = r->firstNode();
296 if (!m_node)
297 return;
298 setUpFullyClippedStack(m_fullyClippedStack, m_node);
299 m_offset = m_node == m_startContainer ? m_startOffset : 0;
300 m_handledNode = false;
301 m_handledChildren = false;
302
303 // calculate first out of bounds node
304 m_pastEndNode = nextInPreOrderCrossingShadowBoundaries(endContainer, endOffset);
305
306 // initialize node processing state
307 m_needAnotherNewline = false;
308 m_textBox = 0;
309
310 // initialize record of previous node processing
311 m_haveEmitted = false;
312 m_lastTextNode = 0;
313 m_lastTextNodeEndedWithCollapsedSpace = false;
314 m_lastCharacter = 0;
315
316 #ifndef NDEBUG
317 // need this just because of the assert in advance()
318 m_positionNode = m_node;
319 #endif
320
321 // identify the first run
322 advance();
323 }
324
advance()325 void TextIterator::advance()
326 {
327 // reset the run information
328 m_positionNode = 0;
329 m_textLength = 0;
330
331 // handle remembered node that needed a newline after the text node's newline
332 if (m_needAnotherNewline) {
333 // Emit the extra newline, and position it *inside* m_node, after m_node's
334 // contents, in case it's a block, in the same way that we position the first
335 // newline. The range for the emitted newline should start where the line
336 // break begins.
337 // FIXME: It would be cleaner if we emitted two newlines during the last
338 // iteration, instead of using m_needAnotherNewline.
339 Node* baseNode = m_node->lastChild() ? m_node->lastChild() : m_node;
340 emitCharacter('\n', baseNode->parentNode(), baseNode, 1, 1);
341 m_needAnotherNewline = false;
342 return;
343 }
344
345 // handle remembered text box
346 if (m_textBox) {
347 handleTextBox();
348 if (m_positionNode)
349 return;
350 }
351
352 while (m_node && m_node != m_pastEndNode) {
353 // if the range ends at offset 0 of an element, represent the
354 // position, but not the content, of that element e.g. if the
355 // node is a blockflow element, emit a newline that
356 // precedes the element
357 if (m_node == m_endContainer && m_endOffset == 0) {
358 representNodeOffsetZero();
359 m_node = 0;
360 return;
361 }
362
363 RenderObject* renderer = m_node->renderer();
364 if (!renderer) {
365 m_handledNode = true;
366 m_handledChildren = true;
367 } else {
368 // handle current node according to its type
369 if (!m_handledNode) {
370 if (renderer->isText() && m_node->nodeType() == Node::TEXT_NODE) // FIXME: What about CDATA_SECTION_NODE?
371 m_handledNode = handleTextNode();
372 else if (renderer && (renderer->isImage() || renderer->isWidget() ||
373 (renderer->node() && renderer->node()->isElementNode() &&
374 static_cast<Element*>(renderer->node())->isFormControlElement())))
375 m_handledNode = handleReplacedElement();
376 else
377 m_handledNode = handleNonTextNode();
378 if (m_positionNode)
379 return;
380 }
381 }
382
383 // find a new current node to handle in depth-first manner,
384 // calling exitNode() as we come back thru a parent node
385 Node* next = m_handledChildren ? 0 : m_node->firstChild();
386 m_offset = 0;
387 if (!next) {
388 next = m_node->nextSibling();
389 if (!next) {
390 bool pastEnd = m_node->traverseNextNode() == m_pastEndNode;
391 Node* parentNode = parentCrossingShadowBoundaries(m_node);
392 while (!next && parentNode) {
393 if ((pastEnd && parentNode == m_endContainer) || m_endContainer->isDescendantOf(parentNode))
394 return;
395 bool haveRenderer = m_node->renderer();
396 m_node = parentNode;
397 m_fullyClippedStack.pop();
398 parentNode = parentCrossingShadowBoundaries(m_node);
399 if (haveRenderer)
400 exitNode();
401 if (m_positionNode) {
402 m_handledNode = true;
403 m_handledChildren = true;
404 return;
405 }
406 next = m_node->nextSibling();
407 }
408 }
409 m_fullyClippedStack.pop();
410 }
411
412 // set the new current node
413 m_node = next;
414 if (m_node)
415 pushFullyClippedState(m_fullyClippedStack, m_node);
416 m_handledNode = false;
417 m_handledChildren = false;
418
419 // how would this ever be?
420 if (m_positionNode)
421 return;
422 }
423 }
424
compareBoxStart(const InlineTextBox * first,const InlineTextBox * second)425 static inline bool compareBoxStart(const InlineTextBox* first, const InlineTextBox* second)
426 {
427 return first->start() < second->start();
428 }
429
handleTextNode()430 bool TextIterator::handleTextNode()
431 {
432 if (m_fullyClippedStack.top())
433 return false;
434
435 RenderText* renderer = toRenderText(m_node->renderer());
436 if (renderer->style()->visibility() != VISIBLE)
437 return false;
438
439 m_lastTextNode = m_node;
440 String str = renderer->text();
441
442 // handle pre-formatted text
443 if (!renderer->style()->collapseWhiteSpace()) {
444 int runStart = m_offset;
445 if (m_lastTextNodeEndedWithCollapsedSpace) {
446 emitCharacter(' ', m_node, 0, runStart, runStart);
447 return false;
448 }
449 int strLength = str.length();
450 int end = (m_node == m_endContainer) ? m_endOffset : INT_MAX;
451 int runEnd = min(strLength, end);
452
453 if (runStart >= runEnd)
454 return true;
455
456 emitText(m_node, runStart, runEnd);
457 return true;
458 }
459
460 if (!renderer->firstTextBox() && str.length() > 0) {
461 m_lastTextNodeEndedWithCollapsedSpace = true; // entire block is collapsed space
462 return true;
463 }
464
465 // Used when text boxes are out of order (Hebrew/Arabic w/ embeded LTR text)
466 if (renderer->containsReversedText()) {
467 m_sortedTextBoxes.clear();
468 for (InlineTextBox* textBox = renderer->firstTextBox(); textBox; textBox = textBox->nextTextBox()) {
469 m_sortedTextBoxes.append(textBox);
470 }
471 std::sort(m_sortedTextBoxes.begin(), m_sortedTextBoxes.end(), compareBoxStart);
472 m_sortedTextBoxesPosition = 0;
473 }
474
475 m_textBox = renderer->containsReversedText() ? m_sortedTextBoxes[0] : renderer->firstTextBox();
476 handleTextBox();
477 return true;
478 }
479
handleTextBox()480 void TextIterator::handleTextBox()
481 {
482 RenderText* renderer = toRenderText(m_node->renderer());
483 String str = renderer->text();
484 int start = m_offset;
485 int end = (m_node == m_endContainer) ? m_endOffset : INT_MAX;
486 while (m_textBox) {
487 int textBoxStart = m_textBox->start();
488 int runStart = max(textBoxStart, start);
489
490 // Check for collapsed space at the start of this run.
491 InlineTextBox* firstTextBox = renderer->containsReversedText() ? m_sortedTextBoxes[0] : renderer->firstTextBox();
492 bool needSpace = m_lastTextNodeEndedWithCollapsedSpace
493 || (m_textBox == firstTextBox && textBoxStart == runStart && runStart > 0);
494 if (needSpace && !isCollapsibleWhitespace(m_lastCharacter) && m_lastCharacter) {
495 if (m_lastTextNode == m_node && runStart > 0 && str[runStart - 1] == ' ') {
496 unsigned spaceRunStart = runStart - 1;
497 while (spaceRunStart > 0 && str[spaceRunStart - 1] == ' ')
498 --spaceRunStart;
499 emitText(m_node, spaceRunStart, spaceRunStart + 1);
500 } else
501 emitCharacter(' ', m_node, 0, runStart, runStart);
502 return;
503 }
504 int textBoxEnd = textBoxStart + m_textBox->len();
505 int runEnd = min(textBoxEnd, end);
506
507 // Determine what the next text box will be, but don't advance yet
508 InlineTextBox* nextTextBox = 0;
509 if (renderer->containsReversedText()) {
510 if (m_sortedTextBoxesPosition + 1 < m_sortedTextBoxes.size())
511 nextTextBox = m_sortedTextBoxes[m_sortedTextBoxesPosition + 1];
512 } else
513 nextTextBox = m_textBox->nextTextBox();
514
515 if (runStart < runEnd) {
516 // Handle either a single newline character (which becomes a space),
517 // or a run of characters that does not include a newline.
518 // This effectively translates newlines to spaces without copying the text.
519 if (str[runStart] == '\n') {
520 emitCharacter(' ', m_node, 0, runStart, runStart + 1);
521 m_offset = runStart + 1;
522 } else {
523 int subrunEnd = str.find('\n', runStart);
524 if (subrunEnd == -1 || subrunEnd > runEnd)
525 subrunEnd = runEnd;
526
527 m_offset = subrunEnd;
528 emitText(m_node, runStart, subrunEnd);
529 }
530
531 // If we are doing a subrun that doesn't go to the end of the text box,
532 // come back again to finish handling this text box; don't advance to the next one.
533 if (m_positionEndOffset < textBoxEnd)
534 return;
535
536 // Advance and return
537 int nextRunStart = nextTextBox ? nextTextBox->start() : str.length();
538 if (nextRunStart > runEnd)
539 m_lastTextNodeEndedWithCollapsedSpace = true; // collapsed space between runs or at the end
540 m_textBox = nextTextBox;
541 if (renderer->containsReversedText())
542 ++m_sortedTextBoxesPosition;
543 return;
544 }
545 // Advance and continue
546 m_textBox = nextTextBox;
547 if (renderer->containsReversedText())
548 ++m_sortedTextBoxesPosition;
549 }
550 }
551
handleReplacedElement()552 bool TextIterator::handleReplacedElement()
553 {
554 if (m_fullyClippedStack.top())
555 return false;
556
557 RenderObject* renderer = m_node->renderer();
558 if (renderer->style()->visibility() != VISIBLE)
559 return false;
560
561 if (m_lastTextNodeEndedWithCollapsedSpace) {
562 emitCharacter(' ', m_lastTextNode->parentNode(), m_lastTextNode, 1, 1);
563 return false;
564 }
565
566 if (m_enterTextControls && renderer->isTextControl()) {
567 if (HTMLElement* innerTextElement = toRenderTextControl(renderer)->innerTextElement()) {
568 m_node = innerTextElement->shadowTreeRootNode();
569 pushFullyClippedState(m_fullyClippedStack, m_node);
570 m_offset = 0;
571 return false;
572 }
573 }
574
575 m_haveEmitted = true;
576
577 if (m_emitCharactersBetweenAllVisiblePositions) {
578 // We want replaced elements to behave like punctuation for boundary
579 // finding, and to simply take up space for the selection preservation
580 // code in moveParagraphs, so we use a comma.
581 emitCharacter(',', m_node->parentNode(), m_node, 0, 1);
582 return true;
583 }
584
585 m_positionNode = m_node->parentNode();
586 m_positionOffsetBaseNode = m_node;
587 m_positionStartOffset = 0;
588 m_positionEndOffset = 1;
589
590 m_textCharacters = 0;
591 m_textLength = 0;
592
593 m_lastCharacter = 0;
594
595 return true;
596 }
597
shouldEmitTabBeforeNode(Node * node)598 static bool shouldEmitTabBeforeNode(Node* node)
599 {
600 RenderObject* r = node->renderer();
601
602 // Table cells are delimited by tabs.
603 if (!r || !isTableCell(node))
604 return false;
605
606 // Want a tab before every cell other than the first one
607 RenderTableCell* rc = toRenderTableCell(r);
608 RenderTable* t = rc->table();
609 return t && (t->cellBefore(rc) || t->cellAbove(rc));
610 }
611
shouldEmitNewlineForNode(Node * node)612 static bool shouldEmitNewlineForNode(Node* node)
613 {
614 // br elements are represented by a single newline.
615 RenderObject* r = node->renderer();
616 if (!r)
617 return node->hasTagName(brTag);
618
619 return r->isBR();
620 }
621
shouldEmitNewlinesBeforeAndAfterNode(Node * node)622 static bool shouldEmitNewlinesBeforeAndAfterNode(Node* node)
623 {
624 // Block flow (versus inline flow) is represented by having
625 // a newline both before and after the element.
626 RenderObject* r = node->renderer();
627 if (!r) {
628 return (node->hasTagName(blockquoteTag)
629 || node->hasTagName(ddTag)
630 || node->hasTagName(divTag)
631 || node->hasTagName(dlTag)
632 || node->hasTagName(dtTag)
633 || node->hasTagName(h1Tag)
634 || node->hasTagName(h2Tag)
635 || node->hasTagName(h3Tag)
636 || node->hasTagName(h4Tag)
637 || node->hasTagName(h5Tag)
638 || node->hasTagName(h6Tag)
639 || node->hasTagName(hrTag)
640 || node->hasTagName(liTag)
641 || node->hasTagName(listingTag)
642 || node->hasTagName(olTag)
643 || node->hasTagName(pTag)
644 || node->hasTagName(preTag)
645 || node->hasTagName(trTag)
646 || node->hasTagName(ulTag));
647 }
648
649 // Need to make an exception for table cells, because they are blocks, but we
650 // want them tab-delimited rather than having newlines before and after.
651 if (isTableCell(node))
652 return false;
653
654 // Need to make an exception for table row elements, because they are neither
655 // "inline" or "RenderBlock", but we want newlines for them.
656 if (r->isTableRow()) {
657 RenderTable* t = toRenderTableRow(r)->table();
658 if (t && !t->isInline())
659 return true;
660 }
661
662 return !r->isInline() && r->isRenderBlock() && !r->isFloatingOrPositioned() && !r->isBody();
663 }
664
shouldEmitNewlineAfterNode(Node * node)665 static bool shouldEmitNewlineAfterNode(Node* node)
666 {
667 // FIXME: It should be better but slower to create a VisiblePosition here.
668 if (!shouldEmitNewlinesBeforeAndAfterNode(node))
669 return false;
670 // Check if this is the very last renderer in the document.
671 // If so, then we should not emit a newline.
672 while ((node = node->traverseNextSibling()))
673 if (node->renderer())
674 return true;
675 return false;
676 }
677
shouldEmitNewlineBeforeNode(Node * node)678 static bool shouldEmitNewlineBeforeNode(Node* node)
679 {
680 return shouldEmitNewlinesBeforeAndAfterNode(node);
681 }
682
shouldEmitExtraNewlineForNode(Node * node)683 static bool shouldEmitExtraNewlineForNode(Node* node)
684 {
685 // When there is a significant collapsed bottom margin, emit an extra
686 // newline for a more realistic result. We end up getting the right
687 // result even without margin collapsing. For example: <div><p>text</p></div>
688 // will work right even if both the <div> and the <p> have bottom margins.
689 RenderObject* r = node->renderer();
690 if (!r || !r->isBox())
691 return false;
692
693 // NOTE: We only do this for a select set of nodes, and fwiw WinIE appears
694 // not to do this at all
695 if (node->hasTagName(h1Tag)
696 || node->hasTagName(h2Tag)
697 || node->hasTagName(h3Tag)
698 || node->hasTagName(h4Tag)
699 || node->hasTagName(h5Tag)
700 || node->hasTagName(h6Tag)
701 || node->hasTagName(pTag)) {
702 RenderStyle* style = r->style();
703 if (style) {
704 int bottomMargin = toRenderBox(r)->collapsedMarginBottom();
705 int fontSize = style->fontDescription().computedPixelSize();
706 if (bottomMargin * 2 >= fontSize)
707 return true;
708 }
709 }
710
711 return false;
712 }
713
collapsedSpaceLength(RenderText * renderer,int textEnd)714 static int collapsedSpaceLength(RenderText* renderer, int textEnd)
715 {
716 const UChar* characters = renderer->text()->characters();
717 int length = renderer->text()->length();
718 for (int i = textEnd; i < length; ++i) {
719 if (!renderer->style()->isCollapsibleWhiteSpace(characters[i]))
720 return i - textEnd;
721 }
722
723 return length - textEnd;
724 }
725
maxOffsetIncludingCollapsedSpaces(Node * node)726 static int maxOffsetIncludingCollapsedSpaces(Node* node)
727 {
728 int offset = caretMaxOffset(node);
729
730 if (node->renderer() && node->renderer()->isText())
731 offset += collapsedSpaceLength(toRenderText(node->renderer()), offset);
732
733 return offset;
734 }
735
736 // Whether or not we should emit a character as we enter m_node (if it's a container) or as we hit it (if it's atomic).
shouldRepresentNodeOffsetZero()737 bool TextIterator::shouldRepresentNodeOffsetZero()
738 {
739 if (m_emitCharactersBetweenAllVisiblePositions && m_node->renderer() && m_node->renderer()->isTable())
740 return true;
741
742 // Leave element positioned flush with start of a paragraph
743 // (e.g. do not insert tab before a table cell at the start of a paragraph)
744 if (m_lastCharacter == '\n')
745 return false;
746
747 // Otherwise, show the position if we have emitted any characters
748 if (m_haveEmitted)
749 return true;
750
751 // We've not emitted anything yet. Generally, there is no need for any positioning then.
752 // The only exception is when the element is visually not in the same line as
753 // the start of the range (e.g. the range starts at the end of the previous paragraph).
754 // NOTE: Creating VisiblePositions and comparing them is relatively expensive, so we
755 // make quicker checks to possibly avoid that. Another check that we could make is
756 // is whether the inline vs block flow changed since the previous visible element.
757 // I think we're already in a special enough case that that won't be needed, tho.
758
759 // No character needed if this is the first node in the range.
760 if (m_node == m_startContainer)
761 return false;
762
763 // If we are outside the start container's subtree, assume we need to emit.
764 // FIXME: m_startContainer could be an inline block
765 if (!m_node->isDescendantOf(m_startContainer))
766 return true;
767
768 // If we started as m_startContainer offset 0 and the current node is a descendant of
769 // the start container, we already had enough context to correctly decide whether to
770 // emit after a preceding block. We chose not to emit (m_haveEmitted is false),
771 // so don't second guess that now.
772 // NOTE: Is this really correct when m_node is not a leftmost descendant? Probably
773 // immaterial since we likely would have already emitted something by now.
774 if (m_startOffset == 0)
775 return false;
776
777 // If this node is unrendered or invisible the VisiblePosition checks below won't have much meaning.
778 // Additionally, if the range we are iterating over contains huge sections of unrendered content,
779 // we would create VisiblePositions on every call to this function without this check.
780 if (!m_node->renderer() || m_node->renderer()->style()->visibility() != VISIBLE)
781 return false;
782
783 // The startPos.isNotNull() check is needed because the start could be before the body,
784 // and in that case we'll get null. We don't want to put in newlines at the start in that case.
785 // The currPos.isNotNull() check is needed because positions in non-HTML content
786 // (like SVG) do not have visible positions, and we don't want to emit for them either.
787 VisiblePosition startPos = VisiblePosition(m_startContainer, m_startOffset, DOWNSTREAM);
788 VisiblePosition currPos = VisiblePosition(m_node, 0, DOWNSTREAM);
789 return startPos.isNotNull() && currPos.isNotNull() && !inSameLine(startPos, currPos);
790 }
791
shouldEmitSpaceBeforeAndAfterNode(Node * node)792 bool TextIterator::shouldEmitSpaceBeforeAndAfterNode(Node* node)
793 {
794 return node->renderer() && node->renderer()->isTable() && (node->renderer()->isInline() || m_emitCharactersBetweenAllVisiblePositions);
795 }
796
representNodeOffsetZero()797 void TextIterator::representNodeOffsetZero()
798 {
799 // Emit a character to show the positioning of m_node.
800
801 // When we haven't been emitting any characters, shouldRepresentNodeOffsetZero() can
802 // create VisiblePositions, which is expensive. So, we perform the inexpensive checks
803 // on m_node to see if it necessitates emitting a character first and will early return
804 // before encountering shouldRepresentNodeOffsetZero()s worse case behavior.
805 if (shouldEmitTabBeforeNode(m_node)) {
806 if (shouldRepresentNodeOffsetZero())
807 emitCharacter('\t', m_node->parentNode(), m_node, 0, 0);
808 } else if (shouldEmitNewlineBeforeNode(m_node)) {
809 if (shouldRepresentNodeOffsetZero())
810 emitCharacter('\n', m_node->parentNode(), m_node, 0, 0);
811 } else if (shouldEmitSpaceBeforeAndAfterNode(m_node)) {
812 if (shouldRepresentNodeOffsetZero())
813 emitCharacter(' ', m_node->parentNode(), m_node, 0, 0);
814 }
815 }
816
handleNonTextNode()817 bool TextIterator::handleNonTextNode()
818 {
819 if (shouldEmitNewlineForNode(m_node))
820 emitCharacter('\n', m_node->parentNode(), m_node, 0, 1);
821 else if (m_emitCharactersBetweenAllVisiblePositions && m_node->renderer() && m_node->renderer()->isHR())
822 emitCharacter(' ', m_node->parentNode(), m_node, 0, 1);
823 else
824 representNodeOffsetZero();
825
826 return true;
827 }
828
exitNode()829 void TextIterator::exitNode()
830 {
831 // prevent emitting a newline when exiting a collapsed block at beginning of the range
832 // FIXME: !m_haveEmitted does not necessarily mean there was a collapsed block... it could
833 // have been an hr (e.g.). Also, a collapsed block could have height (e.g. a table) and
834 // therefore look like a blank line.
835 if (!m_haveEmitted)
836 return;
837
838 // Emit with a position *inside* m_node, after m_node's contents, in
839 // case it is a block, because the run should start where the
840 // emitted character is positioned visually.
841 Node* baseNode = m_node->lastChild() ? m_node->lastChild() : m_node;
842 // FIXME: This shouldn't require the m_lastTextNode to be true, but we can't change that without making
843 // the logic in _web_attributedStringFromRange match. We'll get that for free when we switch to use
844 // TextIterator in _web_attributedStringFromRange.
845 // See <rdar://problem/5428427> for an example of how this mismatch will cause problems.
846 if (m_lastTextNode && shouldEmitNewlineAfterNode(m_node)) {
847 // use extra newline to represent margin bottom, as needed
848 bool addNewline = shouldEmitExtraNewlineForNode(m_node);
849
850 // FIXME: We need to emit a '\n' as we leave an empty block(s) that
851 // contain a VisiblePosition when doing selection preservation.
852 if (m_lastCharacter != '\n') {
853 // insert a newline with a position following this block's contents.
854 emitCharacter('\n', baseNode->parentNode(), baseNode, 1, 1);
855 // remember whether to later add a newline for the current node
856 ASSERT(!m_needAnotherNewline);
857 m_needAnotherNewline = addNewline;
858 } else if (addNewline)
859 // insert a newline with a position following this block's contents.
860 emitCharacter('\n', baseNode->parentNode(), baseNode, 1, 1);
861 }
862
863 // If nothing was emitted, see if we need to emit a space.
864 if (!m_positionNode && shouldEmitSpaceBeforeAndAfterNode(m_node))
865 emitCharacter(' ', baseNode->parentNode(), baseNode, 1, 1);
866 }
867
emitCharacter(UChar c,Node * textNode,Node * offsetBaseNode,int textStartOffset,int textEndOffset)868 void TextIterator::emitCharacter(UChar c, Node* textNode, Node* offsetBaseNode, int textStartOffset, int textEndOffset)
869 {
870 m_haveEmitted = true;
871
872 // remember information with which to construct the TextIterator::range()
873 // NOTE: textNode is often not a text node, so the range will specify child nodes of positionNode
874 m_positionNode = textNode;
875 m_positionOffsetBaseNode = offsetBaseNode;
876 m_positionStartOffset = textStartOffset;
877 m_positionEndOffset = textEndOffset;
878
879 // remember information with which to construct the TextIterator::characters() and length()
880 m_singleCharacterBuffer = c;
881 m_textCharacters = &m_singleCharacterBuffer;
882 m_textLength = 1;
883
884 // remember some iteration state
885 m_lastTextNodeEndedWithCollapsedSpace = false;
886 m_lastCharacter = c;
887 }
888
emitText(Node * textNode,int textStartOffset,int textEndOffset)889 void TextIterator::emitText(Node* textNode, int textStartOffset, int textEndOffset)
890 {
891 RenderText* renderer = toRenderText(m_node->renderer());
892 String str = renderer->text();
893 ASSERT(str.characters());
894
895 m_positionNode = textNode;
896 m_positionOffsetBaseNode = 0;
897 m_positionStartOffset = textStartOffset;
898 m_positionEndOffset = textEndOffset;
899 m_textCharacters = str.characters() + textStartOffset;
900 m_textLength = textEndOffset - textStartOffset;
901 m_lastCharacter = str[textEndOffset - 1];
902
903 m_lastTextNodeEndedWithCollapsedSpace = false;
904 m_haveEmitted = true;
905 }
906
range() const907 PassRefPtr<Range> TextIterator::range() const
908 {
909 // use the current run information, if we have it
910 if (m_positionNode) {
911 if (m_positionOffsetBaseNode) {
912 int index = m_positionOffsetBaseNode->nodeIndex();
913 m_positionStartOffset += index;
914 m_positionEndOffset += index;
915 m_positionOffsetBaseNode = 0;
916 }
917 return Range::create(m_positionNode->document(), m_positionNode, m_positionStartOffset, m_positionNode, m_positionEndOffset);
918 }
919
920 // otherwise, return the end of the overall range we were given
921 if (m_endContainer)
922 return Range::create(m_endContainer->document(), m_endContainer, m_endOffset, m_endContainer, m_endOffset);
923
924 return 0;
925 }
926
node() const927 Node* TextIterator::node() const
928 {
929 RefPtr<Range> textRange = range();
930 if (!textRange)
931 return 0;
932
933 Node* node = textRange->startContainer();
934 if (!node)
935 return 0;
936 if (node->offsetInCharacters())
937 return node;
938
939 return node->childNode(textRange->startOffset());
940 }
941
942 // --------
943
SimplifiedBackwardsTextIterator()944 SimplifiedBackwardsTextIterator::SimplifiedBackwardsTextIterator()
945 : m_positionNode(0)
946 {
947 }
948
SimplifiedBackwardsTextIterator(const Range * r)949 SimplifiedBackwardsTextIterator::SimplifiedBackwardsTextIterator(const Range* r)
950 : m_positionNode(0)
951 {
952 if (!r)
953 return;
954
955 Node* startNode = r->startContainer();
956 if (!startNode)
957 return;
958 Node* endNode = r->endContainer();
959 int startOffset = r->startOffset();
960 int endOffset = r->endOffset();
961
962 if (!startNode->offsetInCharacters()) {
963 if (startOffset >= 0 && startOffset < static_cast<int>(startNode->childNodeCount())) {
964 startNode = startNode->childNode(startOffset);
965 startOffset = 0;
966 }
967 }
968 if (!endNode->offsetInCharacters()) {
969 if (endOffset > 0 && endOffset <= static_cast<int>(endNode->childNodeCount())) {
970 endNode = endNode->childNode(endOffset - 1);
971 endOffset = lastOffsetInNode(endNode);
972 }
973 }
974
975 m_node = endNode;
976 setUpFullyClippedStack(m_fullyClippedStack, m_node);
977 m_offset = endOffset;
978 m_handledNode = false;
979 m_handledChildren = endOffset == 0;
980
981 m_startNode = startNode;
982 m_startOffset = startOffset;
983 m_endNode = endNode;
984 m_endOffset = endOffset;
985
986 #ifndef NDEBUG
987 // Need this just because of the assert.
988 m_positionNode = endNode;
989 #endif
990
991 m_lastTextNode = 0;
992 m_lastCharacter = '\n';
993
994 m_pastStartNode = previousInPostOrderCrossingShadowBoundaries(startNode, startOffset);
995
996 advance();
997 }
998
advance()999 void SimplifiedBackwardsTextIterator::advance()
1000 {
1001 ASSERT(m_positionNode);
1002
1003 m_positionNode = 0;
1004 m_textLength = 0;
1005
1006 while (m_node && m_node != m_pastStartNode) {
1007 // Don't handle node if we start iterating at [node, 0].
1008 if (!m_handledNode && !(m_node == m_endNode && m_endOffset == 0)) {
1009 RenderObject* renderer = m_node->renderer();
1010 if (renderer && renderer->isText() && m_node->nodeType() == Node::TEXT_NODE) {
1011 // FIXME: What about CDATA_SECTION_NODE?
1012 if (renderer->style()->visibility() == VISIBLE && m_offset > 0)
1013 m_handledNode = handleTextNode();
1014 } else if (renderer && (renderer->isImage() || renderer->isWidget())) {
1015 if (renderer->style()->visibility() == VISIBLE && m_offset > 0)
1016 m_handledNode = handleReplacedElement();
1017 } else
1018 m_handledNode = handleNonTextNode();
1019 if (m_positionNode)
1020 return;
1021 }
1022
1023 Node* next = m_handledChildren ? 0 : m_node->lastChild();
1024 if (!next) {
1025 // Exit empty containers as we pass over them or containers
1026 // where [container, 0] is where we started iterating.
1027 if (!m_handledNode &&
1028 canHaveChildrenForEditing(m_node) &&
1029 m_node->parentNode() &&
1030 (!m_node->lastChild() || (m_node == m_endNode && m_endOffset == 0))) {
1031 exitNode();
1032 if (m_positionNode) {
1033 m_handledNode = true;
1034 m_handledChildren = true;
1035 return;
1036 }
1037 }
1038 // Exit all other containers.
1039 next = m_node->previousSibling();
1040 while (!next) {
1041 Node* parentNode = parentCrossingShadowBoundaries(m_node);
1042 if (!parentNode)
1043 break;
1044 m_node = parentNode;
1045 m_fullyClippedStack.pop();
1046 exitNode();
1047 if (m_positionNode) {
1048 m_handledNode = true;
1049 m_handledChildren = true;
1050 return;
1051 }
1052 next = m_node->previousSibling();
1053 }
1054 m_fullyClippedStack.pop();
1055 }
1056
1057 m_node = next;
1058 if (m_node)
1059 pushFullyClippedState(m_fullyClippedStack, m_node);
1060 // For the purpose of word boundary detection,
1061 // we should iterate all visible text and trailing (collapsed) whitespaces.
1062 m_offset = m_node ? maxOffsetIncludingCollapsedSpaces(m_node) : 0;
1063 m_handledNode = false;
1064 m_handledChildren = false;
1065
1066 if (m_positionNode)
1067 return;
1068 }
1069 }
1070
handleTextNode()1071 bool SimplifiedBackwardsTextIterator::handleTextNode()
1072 {
1073 m_lastTextNode = m_node;
1074
1075 RenderText* renderer = toRenderText(m_node->renderer());
1076 String str = renderer->text();
1077
1078 if (!renderer->firstTextBox() && str.length() > 0)
1079 return true;
1080
1081 m_positionEndOffset = m_offset;
1082
1083 m_offset = (m_node == m_startNode) ? m_startOffset : 0;
1084 m_positionNode = m_node;
1085 m_positionStartOffset = m_offset;
1086 m_textLength = m_positionEndOffset - m_positionStartOffset;
1087 m_textCharacters = str.characters() + m_positionStartOffset;
1088
1089 m_lastCharacter = str[m_positionEndOffset - 1];
1090
1091 return true;
1092 }
1093
handleReplacedElement()1094 bool SimplifiedBackwardsTextIterator::handleReplacedElement()
1095 {
1096 unsigned index = m_node->nodeIndex();
1097 // We want replaced elements to behave like punctuation for boundary
1098 // finding, and to simply take up space for the selection preservation
1099 // code in moveParagraphs, so we use a comma. Unconditionally emit
1100 // here because this iterator is only used for boundary finding.
1101 emitCharacter(',', m_node->parentNode(), index, index + 1);
1102 return true;
1103 }
1104
handleNonTextNode()1105 bool SimplifiedBackwardsTextIterator::handleNonTextNode()
1106 {
1107 // We can use a linefeed in place of a tab because this simple iterator is only used to
1108 // find boundaries, not actual content. A linefeed breaks words, sentences, and paragraphs.
1109 if (shouldEmitNewlineForNode(m_node) || shouldEmitNewlineAfterNode(m_node) || shouldEmitTabBeforeNode(m_node)) {
1110 unsigned index = m_node->nodeIndex();
1111 // The start of this emitted range is wrong. Ensuring correctness would require
1112 // VisiblePositions and so would be slow. previousBoundary expects this.
1113 emitCharacter('\n', m_node->parentNode(), index + 1, index + 1);
1114 }
1115 return true;
1116 }
1117
exitNode()1118 void SimplifiedBackwardsTextIterator::exitNode()
1119 {
1120 if (shouldEmitNewlineForNode(m_node) || shouldEmitNewlineBeforeNode(m_node) || shouldEmitTabBeforeNode(m_node)) {
1121 // The start of this emitted range is wrong. Ensuring correctness would require
1122 // VisiblePositions and so would be slow. previousBoundary expects this.
1123 emitCharacter('\n', m_node, 0, 0);
1124 }
1125 }
1126
emitCharacter(UChar c,Node * node,int startOffset,int endOffset)1127 void SimplifiedBackwardsTextIterator::emitCharacter(UChar c, Node* node, int startOffset, int endOffset)
1128 {
1129 m_singleCharacterBuffer = c;
1130 m_positionNode = node;
1131 m_positionStartOffset = startOffset;
1132 m_positionEndOffset = endOffset;
1133 m_textCharacters = &m_singleCharacterBuffer;
1134 m_textLength = 1;
1135 m_lastCharacter = c;
1136 }
1137
range() const1138 PassRefPtr<Range> SimplifiedBackwardsTextIterator::range() const
1139 {
1140 if (m_positionNode)
1141 return Range::create(m_positionNode->document(), m_positionNode, m_positionStartOffset, m_positionNode, m_positionEndOffset);
1142
1143 return Range::create(m_startNode->document(), m_startNode, m_startOffset, m_startNode, m_startOffset);
1144 }
1145
1146 // --------
1147
CharacterIterator()1148 CharacterIterator::CharacterIterator()
1149 : m_offset(0)
1150 , m_runOffset(0)
1151 , m_atBreak(true)
1152 {
1153 }
1154
CharacterIterator(const Range * r,bool emitCharactersBetweenAllVisiblePositions,bool enterTextControls)1155 CharacterIterator::CharacterIterator(const Range* r, bool emitCharactersBetweenAllVisiblePositions, bool enterTextControls)
1156 : m_offset(0)
1157 , m_runOffset(0)
1158 , m_atBreak(true)
1159 , m_textIterator(r, emitCharactersBetweenAllVisiblePositions, enterTextControls)
1160 {
1161 while (!atEnd() && m_textIterator.length() == 0)
1162 m_textIterator.advance();
1163 }
1164
range() const1165 PassRefPtr<Range> CharacterIterator::range() const
1166 {
1167 RefPtr<Range> r = m_textIterator.range();
1168 if (!m_textIterator.atEnd()) {
1169 if (m_textIterator.length() <= 1) {
1170 ASSERT(m_runOffset == 0);
1171 } else {
1172 Node* n = r->startContainer();
1173 ASSERT(n == r->endContainer());
1174 int offset = r->startOffset() + m_runOffset;
1175 ExceptionCode ec = 0;
1176 r->setStart(n, offset, ec);
1177 r->setEnd(n, offset + 1, ec);
1178 ASSERT(!ec);
1179 }
1180 }
1181 return r.release();
1182 }
1183
advance(int count)1184 void CharacterIterator::advance(int count)
1185 {
1186 if (count <= 0) {
1187 ASSERT(count == 0);
1188 return;
1189 }
1190
1191 m_atBreak = false;
1192
1193 // easy if there is enough left in the current m_textIterator run
1194 int remaining = m_textIterator.length() - m_runOffset;
1195 if (count < remaining) {
1196 m_runOffset += count;
1197 m_offset += count;
1198 return;
1199 }
1200
1201 // exhaust the current m_textIterator run
1202 count -= remaining;
1203 m_offset += remaining;
1204
1205 // move to a subsequent m_textIterator run
1206 for (m_textIterator.advance(); !atEnd(); m_textIterator.advance()) {
1207 int runLength = m_textIterator.length();
1208 if (runLength == 0)
1209 m_atBreak = true;
1210 else {
1211 // see whether this is m_textIterator to use
1212 if (count < runLength) {
1213 m_runOffset = count;
1214 m_offset += count;
1215 return;
1216 }
1217
1218 // exhaust this m_textIterator run
1219 count -= runLength;
1220 m_offset += runLength;
1221 }
1222 }
1223
1224 // ran to the end of the m_textIterator... no more runs left
1225 m_atBreak = true;
1226 m_runOffset = 0;
1227 }
1228
string(int numChars)1229 String CharacterIterator::string(int numChars)
1230 {
1231 Vector<UChar> result;
1232 result.reserveInitialCapacity(numChars);
1233 while (numChars > 0 && !atEnd()) {
1234 int runSize = min(numChars, length());
1235 result.append(characters(), runSize);
1236 numChars -= runSize;
1237 advance(runSize);
1238 }
1239 return String::adopt(result);
1240 }
1241
characterSubrange(CharacterIterator & it,int offset,int length)1242 static PassRefPtr<Range> characterSubrange(CharacterIterator& it, int offset, int length)
1243 {
1244 it.advance(offset);
1245 RefPtr<Range> start = it.range();
1246
1247 if (length > 1)
1248 it.advance(length - 1);
1249 RefPtr<Range> end = it.range();
1250
1251 return Range::create(start->startContainer()->document(),
1252 start->startContainer(), start->startOffset(),
1253 end->endContainer(), end->endOffset());
1254 }
1255
BackwardsCharacterIterator()1256 BackwardsCharacterIterator::BackwardsCharacterIterator()
1257 : m_offset(0)
1258 , m_runOffset(0)
1259 , m_atBreak(true)
1260 {
1261 }
1262
BackwardsCharacterIterator(const Range * range)1263 BackwardsCharacterIterator::BackwardsCharacterIterator(const Range* range)
1264 : m_offset(0)
1265 , m_runOffset(0)
1266 , m_atBreak(true)
1267 , m_textIterator(range)
1268 {
1269 while (!atEnd() && !m_textIterator.length())
1270 m_textIterator.advance();
1271 }
1272
range() const1273 PassRefPtr<Range> BackwardsCharacterIterator::range() const
1274 {
1275 RefPtr<Range> r = m_textIterator.range();
1276 if (!m_textIterator.atEnd()) {
1277 if (m_textIterator.length() <= 1)
1278 ASSERT(m_runOffset == 0);
1279 else {
1280 Node* n = r->startContainer();
1281 ASSERT(n == r->endContainer());
1282 int offset = r->endOffset() - m_runOffset;
1283 ExceptionCode ec = 0;
1284 r->setStart(n, offset - 1, ec);
1285 r->setEnd(n, offset, ec);
1286 ASSERT(!ec);
1287 }
1288 }
1289 return r.release();
1290 }
1291
advance(int count)1292 void BackwardsCharacterIterator::advance(int count)
1293 {
1294 if (count <= 0) {
1295 ASSERT(!count);
1296 return;
1297 }
1298
1299 m_atBreak = false;
1300
1301 int remaining = m_textIterator.length() - m_runOffset;
1302 if (count < remaining) {
1303 m_runOffset += count;
1304 m_offset += count;
1305 return;
1306 }
1307
1308 count -= remaining;
1309 m_offset += remaining;
1310
1311 for (m_textIterator.advance(); !atEnd(); m_textIterator.advance()) {
1312 int runLength = m_textIterator.length();
1313 if (runLength == 0)
1314 m_atBreak = true;
1315 else {
1316 if (count < runLength) {
1317 m_runOffset = count;
1318 m_offset += count;
1319 return;
1320 }
1321
1322 count -= runLength;
1323 m_offset += runLength;
1324 }
1325 }
1326
1327 m_atBreak = true;
1328 m_runOffset = 0;
1329 }
1330
1331 // --------
1332
WordAwareIterator()1333 WordAwareIterator::WordAwareIterator()
1334 : m_previousText(0)
1335 , m_didLookAhead(false)
1336 {
1337 }
1338
WordAwareIterator(const Range * r)1339 WordAwareIterator::WordAwareIterator(const Range* r)
1340 : m_previousText(0)
1341 , m_didLookAhead(true) // so we consider the first chunk from the text iterator
1342 , m_textIterator(r)
1343 {
1344 advance(); // get in position over the first chunk of text
1345 }
1346
1347 // We're always in one of these modes:
1348 // - The current chunk in the text iterator is our current chunk
1349 // (typically its a piece of whitespace, or text that ended with whitespace)
1350 // - The previous chunk in the text iterator is our current chunk
1351 // (we looked ahead to the next chunk and found a word boundary)
1352 // - We built up our own chunk of text from many chunks from the text iterator
1353
1354 // FIXME: Performance could be bad for huge spans next to each other that don't fall on word boundaries.
1355
advance()1356 void WordAwareIterator::advance()
1357 {
1358 m_previousText = 0;
1359 m_buffer.clear(); // toss any old buffer we built up
1360
1361 // If last time we did a look-ahead, start with that looked-ahead chunk now
1362 if (!m_didLookAhead) {
1363 ASSERT(!m_textIterator.atEnd());
1364 m_textIterator.advance();
1365 }
1366 m_didLookAhead = false;
1367
1368 // Go to next non-empty chunk
1369 while (!m_textIterator.atEnd() && m_textIterator.length() == 0)
1370 m_textIterator.advance();
1371 m_range = m_textIterator.range();
1372
1373 if (m_textIterator.atEnd())
1374 return;
1375
1376 while (1) {
1377 // If this chunk ends in whitespace we can just use it as our chunk.
1378 if (isSpaceOrNewline(m_textIterator.characters()[m_textIterator.length() - 1]))
1379 return;
1380
1381 // If this is the first chunk that failed, save it in previousText before look ahead
1382 if (m_buffer.isEmpty()) {
1383 m_previousText = m_textIterator.characters();
1384 m_previousLength = m_textIterator.length();
1385 }
1386
1387 // Look ahead to next chunk. If it is whitespace or a break, we can use the previous stuff
1388 m_textIterator.advance();
1389 if (m_textIterator.atEnd() || m_textIterator.length() == 0 || isSpaceOrNewline(m_textIterator.characters()[0])) {
1390 m_didLookAhead = true;
1391 return;
1392 }
1393
1394 if (m_buffer.isEmpty()) {
1395 // Start gobbling chunks until we get to a suitable stopping point
1396 m_buffer.append(m_previousText, m_previousLength);
1397 m_previousText = 0;
1398 }
1399 m_buffer.append(m_textIterator.characters(), m_textIterator.length());
1400 int exception = 0;
1401 m_range->setEnd(m_textIterator.range()->endContainer(), m_textIterator.range()->endOffset(), exception);
1402 }
1403 }
1404
length() const1405 int WordAwareIterator::length() const
1406 {
1407 if (!m_buffer.isEmpty())
1408 return m_buffer.size();
1409 if (m_previousText)
1410 return m_previousLength;
1411 return m_textIterator.length();
1412 }
1413
characters() const1414 const UChar* WordAwareIterator::characters() const
1415 {
1416 if (!m_buffer.isEmpty())
1417 return m_buffer.data();
1418 if (m_previousText)
1419 return m_previousText;
1420 return m_textIterator.characters();
1421 }
1422
1423 // --------
1424
foldQuoteMark(UChar c)1425 static inline UChar foldQuoteMark(UChar c)
1426 {
1427 switch (c) {
1428 case hebrewPunctuationGershayim:
1429 case leftDoubleQuotationMark:
1430 case rightDoubleQuotationMark:
1431 return '"';
1432 case hebrewPunctuationGeresh:
1433 case leftSingleQuotationMark:
1434 case rightSingleQuotationMark:
1435 return '\'';
1436 default:
1437 return c;
1438 }
1439 }
1440
foldQuoteMarks(String & s)1441 static inline void foldQuoteMarks(String& s)
1442 {
1443 s.replace(hebrewPunctuationGeresh, '\'');
1444 s.replace(hebrewPunctuationGershayim, '"');
1445 s.replace(leftDoubleQuotationMark, '"');
1446 s.replace(leftSingleQuotationMark, '\'');
1447 s.replace(rightDoubleQuotationMark, '"');
1448 s.replace(rightSingleQuotationMark, '\'');
1449 }
1450
1451 #if USE(ICU_UNICODE) && !UCONFIG_NO_COLLATION
1452
foldQuoteMarks(UChar * data,size_t length)1453 static inline void foldQuoteMarks(UChar* data, size_t length)
1454 {
1455 for (size_t i = 0; i < length; ++i)
1456 data[i] = foldQuoteMark(data[i]);
1457 }
1458
1459 static const size_t minimumSearchBufferSize = 8192;
1460
1461 #ifndef NDEBUG
1462 static bool searcherInUse;
1463 #endif
1464
createSearcher()1465 static UStringSearch* createSearcher()
1466 {
1467 // Provide a non-empty pattern and non-empty text so usearch_open will not fail,
1468 // but it doesn't matter exactly what it is, since we don't perform any searches
1469 // without setting both the pattern and the text.
1470 UErrorCode status = U_ZERO_ERROR;
1471 UStringSearch* searcher = usearch_open(&newlineCharacter, 1, &newlineCharacter, 1, currentSearchLocaleID(), 0, &status);
1472 ASSERT(status == U_ZERO_ERROR || status == U_USING_FALLBACK_WARNING || status == U_USING_DEFAULT_WARNING);
1473 return searcher;
1474 }
1475
searcher()1476 static UStringSearch* searcher()
1477 {
1478 static UStringSearch* searcher = createSearcher();
1479 return searcher;
1480 }
1481
lockSearcher()1482 static inline void lockSearcher()
1483 {
1484 #ifndef NDEBUG
1485 ASSERT(!searcherInUse);
1486 searcherInUse = true;
1487 #endif
1488 }
1489
unlockSearcher()1490 static inline void unlockSearcher()
1491 {
1492 #ifndef NDEBUG
1493 ASSERT(searcherInUse);
1494 searcherInUse = false;
1495 #endif
1496 }
1497
1498 // ICU's search ignores the distinction between small kana letters and ones
1499 // that are not small, and also characters that differ only in the voicing
1500 // marks when considering only primary collation strength diffrences.
1501 // This is not helpful for end users, since these differences make words
1502 // distinct, so for our purposes we need these to be considered.
1503 // The Unicode folks do not think the collation algorithm should be
1504 // changed. To work around this, we would like to tailor the ICU searcher,
1505 // but we can't get that to work yet. So instead, we check for cases where
1506 // these differences occur, and skip those matches.
1507
1508 // We refer to the above technique as the "kana workaround". The next few
1509 // functions are helper functinos for the kana workaround.
1510
isKanaLetter(UChar character)1511 static inline bool isKanaLetter(UChar character)
1512 {
1513 // Hiragana letters.
1514 if (character >= 0x3041 && character <= 0x3096)
1515 return true;
1516
1517 // Katakana letters.
1518 if (character >= 0x30A1 && character <= 0x30FA)
1519 return true;
1520 if (character >= 0x31F0 && character <= 0x31FF)
1521 return true;
1522
1523 // Halfwidth katakana letters.
1524 if (character >= 0xFF66 && character <= 0xFF9D && character != 0xFF70)
1525 return true;
1526
1527 return false;
1528 }
1529
isSmallKanaLetter(UChar character)1530 static inline bool isSmallKanaLetter(UChar character)
1531 {
1532 ASSERT(isKanaLetter(character));
1533
1534 switch (character) {
1535 case 0x3041: // HIRAGANA LETTER SMALL A
1536 case 0x3043: // HIRAGANA LETTER SMALL I
1537 case 0x3045: // HIRAGANA LETTER SMALL U
1538 case 0x3047: // HIRAGANA LETTER SMALL E
1539 case 0x3049: // HIRAGANA LETTER SMALL O
1540 case 0x3063: // HIRAGANA LETTER SMALL TU
1541 case 0x3083: // HIRAGANA LETTER SMALL YA
1542 case 0x3085: // HIRAGANA LETTER SMALL YU
1543 case 0x3087: // HIRAGANA LETTER SMALL YO
1544 case 0x308E: // HIRAGANA LETTER SMALL WA
1545 case 0x3095: // HIRAGANA LETTER SMALL KA
1546 case 0x3096: // HIRAGANA LETTER SMALL KE
1547 case 0x30A1: // KATAKANA LETTER SMALL A
1548 case 0x30A3: // KATAKANA LETTER SMALL I
1549 case 0x30A5: // KATAKANA LETTER SMALL U
1550 case 0x30A7: // KATAKANA LETTER SMALL E
1551 case 0x30A9: // KATAKANA LETTER SMALL O
1552 case 0x30C3: // KATAKANA LETTER SMALL TU
1553 case 0x30E3: // KATAKANA LETTER SMALL YA
1554 case 0x30E5: // KATAKANA LETTER SMALL YU
1555 case 0x30E7: // KATAKANA LETTER SMALL YO
1556 case 0x30EE: // KATAKANA LETTER SMALL WA
1557 case 0x30F5: // KATAKANA LETTER SMALL KA
1558 case 0x30F6: // KATAKANA LETTER SMALL KE
1559 case 0x31F0: // KATAKANA LETTER SMALL KU
1560 case 0x31F1: // KATAKANA LETTER SMALL SI
1561 case 0x31F2: // KATAKANA LETTER SMALL SU
1562 case 0x31F3: // KATAKANA LETTER SMALL TO
1563 case 0x31F4: // KATAKANA LETTER SMALL NU
1564 case 0x31F5: // KATAKANA LETTER SMALL HA
1565 case 0x31F6: // KATAKANA LETTER SMALL HI
1566 case 0x31F7: // KATAKANA LETTER SMALL HU
1567 case 0x31F8: // KATAKANA LETTER SMALL HE
1568 case 0x31F9: // KATAKANA LETTER SMALL HO
1569 case 0x31FA: // KATAKANA LETTER SMALL MU
1570 case 0x31FB: // KATAKANA LETTER SMALL RA
1571 case 0x31FC: // KATAKANA LETTER SMALL RI
1572 case 0x31FD: // KATAKANA LETTER SMALL RU
1573 case 0x31FE: // KATAKANA LETTER SMALL RE
1574 case 0x31FF: // KATAKANA LETTER SMALL RO
1575 case 0xFF67: // HALFWIDTH KATAKANA LETTER SMALL A
1576 case 0xFF68: // HALFWIDTH KATAKANA LETTER SMALL I
1577 case 0xFF69: // HALFWIDTH KATAKANA LETTER SMALL U
1578 case 0xFF6A: // HALFWIDTH KATAKANA LETTER SMALL E
1579 case 0xFF6B: // HALFWIDTH KATAKANA LETTER SMALL O
1580 case 0xFF6C: // HALFWIDTH KATAKANA LETTER SMALL YA
1581 case 0xFF6D: // HALFWIDTH KATAKANA LETTER SMALL YU
1582 case 0xFF6E: // HALFWIDTH KATAKANA LETTER SMALL YO
1583 case 0xFF6F: // HALFWIDTH KATAKANA LETTER SMALL TU
1584 return true;
1585 }
1586 return false;
1587 }
1588
1589 enum VoicedSoundMarkType { NoVoicedSoundMark, VoicedSoundMark, SemiVoicedSoundMark };
1590
composedVoicedSoundMark(UChar character)1591 static inline VoicedSoundMarkType composedVoicedSoundMark(UChar character)
1592 {
1593 ASSERT(isKanaLetter(character));
1594
1595 switch (character) {
1596 case 0x304C: // HIRAGANA LETTER GA
1597 case 0x304E: // HIRAGANA LETTER GI
1598 case 0x3050: // HIRAGANA LETTER GU
1599 case 0x3052: // HIRAGANA LETTER GE
1600 case 0x3054: // HIRAGANA LETTER GO
1601 case 0x3056: // HIRAGANA LETTER ZA
1602 case 0x3058: // HIRAGANA LETTER ZI
1603 case 0x305A: // HIRAGANA LETTER ZU
1604 case 0x305C: // HIRAGANA LETTER ZE
1605 case 0x305E: // HIRAGANA LETTER ZO
1606 case 0x3060: // HIRAGANA LETTER DA
1607 case 0x3062: // HIRAGANA LETTER DI
1608 case 0x3065: // HIRAGANA LETTER DU
1609 case 0x3067: // HIRAGANA LETTER DE
1610 case 0x3069: // HIRAGANA LETTER DO
1611 case 0x3070: // HIRAGANA LETTER BA
1612 case 0x3073: // HIRAGANA LETTER BI
1613 case 0x3076: // HIRAGANA LETTER BU
1614 case 0x3079: // HIRAGANA LETTER BE
1615 case 0x307C: // HIRAGANA LETTER BO
1616 case 0x3094: // HIRAGANA LETTER VU
1617 case 0x30AC: // KATAKANA LETTER GA
1618 case 0x30AE: // KATAKANA LETTER GI
1619 case 0x30B0: // KATAKANA LETTER GU
1620 case 0x30B2: // KATAKANA LETTER GE
1621 case 0x30B4: // KATAKANA LETTER GO
1622 case 0x30B6: // KATAKANA LETTER ZA
1623 case 0x30B8: // KATAKANA LETTER ZI
1624 case 0x30BA: // KATAKANA LETTER ZU
1625 case 0x30BC: // KATAKANA LETTER ZE
1626 case 0x30BE: // KATAKANA LETTER ZO
1627 case 0x30C0: // KATAKANA LETTER DA
1628 case 0x30C2: // KATAKANA LETTER DI
1629 case 0x30C5: // KATAKANA LETTER DU
1630 case 0x30C7: // KATAKANA LETTER DE
1631 case 0x30C9: // KATAKANA LETTER DO
1632 case 0x30D0: // KATAKANA LETTER BA
1633 case 0x30D3: // KATAKANA LETTER BI
1634 case 0x30D6: // KATAKANA LETTER BU
1635 case 0x30D9: // KATAKANA LETTER BE
1636 case 0x30DC: // KATAKANA LETTER BO
1637 case 0x30F4: // KATAKANA LETTER VU
1638 case 0x30F7: // KATAKANA LETTER VA
1639 case 0x30F8: // KATAKANA LETTER VI
1640 case 0x30F9: // KATAKANA LETTER VE
1641 case 0x30FA: // KATAKANA LETTER VO
1642 return VoicedSoundMark;
1643 case 0x3071: // HIRAGANA LETTER PA
1644 case 0x3074: // HIRAGANA LETTER PI
1645 case 0x3077: // HIRAGANA LETTER PU
1646 case 0x307A: // HIRAGANA LETTER PE
1647 case 0x307D: // HIRAGANA LETTER PO
1648 case 0x30D1: // KATAKANA LETTER PA
1649 case 0x30D4: // KATAKANA LETTER PI
1650 case 0x30D7: // KATAKANA LETTER PU
1651 case 0x30DA: // KATAKANA LETTER PE
1652 case 0x30DD: // KATAKANA LETTER PO
1653 return SemiVoicedSoundMark;
1654 }
1655 return NoVoicedSoundMark;
1656 }
1657
isCombiningVoicedSoundMark(UChar character)1658 static inline bool isCombiningVoicedSoundMark(UChar character)
1659 {
1660 switch (character) {
1661 case 0x3099: // COMBINING KATAKANA-HIRAGANA VOICED SOUND MARK
1662 case 0x309A: // COMBINING KATAKANA-HIRAGANA SEMI-VOICED SOUND MARK
1663 return true;
1664 }
1665 return false;
1666 }
1667
containsKanaLetters(const String & pattern)1668 static inline bool containsKanaLetters(const String& pattern)
1669 {
1670 const UChar* characters = pattern.characters();
1671 unsigned length = pattern.length();
1672 for (unsigned i = 0; i < length; ++i) {
1673 if (isKanaLetter(characters[i]))
1674 return true;
1675 }
1676 return false;
1677 }
1678
normalizeCharacters(const UChar * characters,unsigned length,Vector<UChar> & buffer)1679 static void normalizeCharacters(const UChar* characters, unsigned length, Vector<UChar>& buffer)
1680 {
1681 ASSERT(length);
1682
1683 buffer.resize(length);
1684
1685 UErrorCode status = U_ZERO_ERROR;
1686 size_t bufferSize = unorm_normalize(characters, length, UNORM_NFC, 0, buffer.data(), length, &status);
1687 ASSERT(status == U_ZERO_ERROR || status == U_STRING_NOT_TERMINATED_WARNING || status == U_BUFFER_OVERFLOW_ERROR);
1688 ASSERT(bufferSize);
1689
1690 buffer.resize(bufferSize);
1691
1692 if (status == U_ZERO_ERROR || status == U_STRING_NOT_TERMINATED_WARNING)
1693 return;
1694
1695 status = U_ZERO_ERROR;
1696 unorm_normalize(characters, length, UNORM_NFC, 0, buffer.data(), bufferSize, &status);
1697 ASSERT(status == U_STRING_NOT_TERMINATED_WARNING);
1698 }
1699
SearchBuffer(const String & target,bool isCaseSensitive)1700 inline SearchBuffer::SearchBuffer(const String& target, bool isCaseSensitive)
1701 : m_target(target)
1702 , m_atBreak(true)
1703 , m_targetRequiresKanaWorkaround(containsKanaLetters(m_target))
1704 {
1705 ASSERT(!m_target.isEmpty());
1706
1707 // FIXME: We'd like to tailor the searcher to fold quote marks for us instead
1708 // of doing it in a separate replacement pass here, but ICU doesn't offer a way
1709 // to add tailoring on top of the locale-specific tailoring as of this writing.
1710 foldQuoteMarks(m_target);
1711
1712 size_t targetLength = m_target.length();
1713 m_buffer.reserveInitialCapacity(max(targetLength * 8, minimumSearchBufferSize));
1714 m_overlap = m_buffer.capacity() / 4;
1715
1716 // Grab the single global searcher.
1717 // If we ever have a reason to do more than once search buffer at once, we'll have
1718 // to move to multiple searchers.
1719 lockSearcher();
1720
1721 UStringSearch* searcher = WebCore::searcher();
1722 UCollator* collator = usearch_getCollator(searcher);
1723
1724 UCollationStrength strength = isCaseSensitive ? UCOL_TERTIARY : UCOL_PRIMARY;
1725 if (ucol_getStrength(collator) != strength) {
1726 ucol_setStrength(collator, strength);
1727 usearch_reset(searcher);
1728 }
1729
1730 UErrorCode status = U_ZERO_ERROR;
1731 usearch_setPattern(searcher, m_target.characters(), targetLength, &status);
1732 ASSERT(status == U_ZERO_ERROR);
1733
1734 // The kana workaround requires a normalized copy of the target string.
1735 if (m_targetRequiresKanaWorkaround)
1736 normalizeCharacters(m_target.characters(), m_target.length(), m_normalizedTarget);
1737 }
1738
~SearchBuffer()1739 inline SearchBuffer::~SearchBuffer()
1740 {
1741 unlockSearcher();
1742 }
1743
append(const UChar * characters,size_t length)1744 inline size_t SearchBuffer::append(const UChar* characters, size_t length)
1745 {
1746 ASSERT(length);
1747
1748 if (m_atBreak) {
1749 m_buffer.shrink(0);
1750 m_atBreak = false;
1751 } else if (m_buffer.size() == m_buffer.capacity()) {
1752 memcpy(m_buffer.data(), m_buffer.data() + m_buffer.size() - m_overlap, m_overlap * sizeof(UChar));
1753 m_buffer.shrink(m_overlap);
1754 }
1755
1756 size_t oldLength = m_buffer.size();
1757 size_t usableLength = min(m_buffer.capacity() - oldLength, length);
1758 ASSERT(usableLength);
1759 m_buffer.append(characters, usableLength);
1760 foldQuoteMarks(m_buffer.data() + oldLength, usableLength);
1761 return usableLength;
1762 }
1763
atBreak() const1764 inline bool SearchBuffer::atBreak() const
1765 {
1766 return m_atBreak;
1767 }
1768
reachedBreak()1769 inline void SearchBuffer::reachedBreak()
1770 {
1771 m_atBreak = true;
1772 }
1773
isBadMatch(const UChar * match,size_t matchLength) const1774 inline bool SearchBuffer::isBadMatch(const UChar* match, size_t matchLength) const
1775 {
1776 // This function implements the kana workaround. If usearch treats
1777 // it as a match, but we do not want to, then it's a "bad match".
1778 if (!m_targetRequiresKanaWorkaround)
1779 return false;
1780
1781 // Normalize into a match buffer. We reuse a single buffer rather than
1782 // creating a new one each time.
1783 normalizeCharacters(match, matchLength, m_normalizedMatch);
1784
1785 const UChar* a = m_normalizedTarget.begin();
1786 const UChar* aEnd = m_normalizedTarget.end();
1787
1788 const UChar* b = m_normalizedMatch.begin();
1789 const UChar* bEnd = m_normalizedMatch.end();
1790
1791 while (true) {
1792 // Skip runs of non-kana-letter characters. This is necessary so we can
1793 // correctly handle strings where the target and match have different-length
1794 // runs of characters that match, while still double checking the correctness
1795 // of matches of kana letters with other kana letters.
1796 while (a != aEnd && !isKanaLetter(*a))
1797 ++a;
1798 while (b != bEnd && !isKanaLetter(*b))
1799 ++b;
1800
1801 // If we reached the end of either the target or the match, we should have
1802 // reached the end of both; both should have the same number of kana letters.
1803 if (a == aEnd || b == bEnd) {
1804 ASSERT(a == aEnd);
1805 ASSERT(b == bEnd);
1806 return false;
1807 }
1808
1809 // Check for differences in the kana letter character itself.
1810 if (isSmallKanaLetter(*a) != isSmallKanaLetter(*b))
1811 return true;
1812 if (composedVoicedSoundMark(*a) != composedVoicedSoundMark(*b))
1813 return true;
1814 ++a;
1815 ++b;
1816
1817 // Check for differences in combining voiced sound marks found after the letter.
1818 while (1) {
1819 if (!(a != aEnd && isCombiningVoicedSoundMark(*a))) {
1820 if (b != bEnd && isCombiningVoicedSoundMark(*b))
1821 return true;
1822 break;
1823 }
1824 if (!(b != bEnd && isCombiningVoicedSoundMark(*b)))
1825 return true;
1826 if (*a != *b)
1827 return true;
1828 ++a;
1829 ++b;
1830 }
1831 }
1832 }
1833
search(size_t & start)1834 inline size_t SearchBuffer::search(size_t& start)
1835 {
1836 size_t size = m_buffer.size();
1837 if (m_atBreak) {
1838 if (!size)
1839 return 0;
1840 } else {
1841 if (size != m_buffer.capacity())
1842 return 0;
1843 }
1844
1845 UStringSearch* searcher = WebCore::searcher();
1846
1847 UErrorCode status = U_ZERO_ERROR;
1848 usearch_setText(searcher, m_buffer.data(), size, &status);
1849 ASSERT(status == U_ZERO_ERROR);
1850
1851 int matchStart = usearch_first(searcher, &status);
1852 ASSERT(status == U_ZERO_ERROR);
1853
1854 nextMatch:
1855 if (!(matchStart >= 0 && static_cast<size_t>(matchStart) < size)) {
1856 ASSERT(matchStart == USEARCH_DONE);
1857 return 0;
1858 }
1859
1860 // Matches that start in the overlap area are only tentative.
1861 // The same match may appear later, matching more characters,
1862 // possibly including a combining character that's not yet in the buffer.
1863 if (!m_atBreak && static_cast<size_t>(matchStart) >= size - m_overlap) {
1864 memcpy(m_buffer.data(), m_buffer.data() + size - m_overlap, m_overlap * sizeof(UChar));
1865 m_buffer.shrink(m_overlap);
1866 return 0;
1867 }
1868
1869 size_t matchedLength = usearch_getMatchedLength(searcher);
1870 ASSERT(matchStart + matchedLength <= size);
1871
1872 // If this match is "bad", move on to the next match.
1873 if (isBadMatch(m_buffer.data() + matchStart, matchedLength)) {
1874 matchStart = usearch_next(searcher, &status);
1875 ASSERT(status == U_ZERO_ERROR);
1876 goto nextMatch;
1877 }
1878
1879 size_t newSize = size - (matchStart + 1);
1880 memmove(m_buffer.data(), m_buffer.data() + matchStart + 1, newSize * sizeof(UChar));
1881 m_buffer.shrink(newSize);
1882
1883 start = size - matchStart;
1884 return matchedLength;
1885 }
1886
1887 #else // !ICU_UNICODE
1888
SearchBuffer(const String & target,bool isCaseSensitive)1889 inline SearchBuffer::SearchBuffer(const String& target, bool isCaseSensitive)
1890 : m_target(isCaseSensitive ? target : target.foldCase())
1891 , m_isCaseSensitive(isCaseSensitive)
1892 , m_buffer(m_target.length())
1893 , m_isCharacterStartBuffer(m_target.length())
1894 , m_isBufferFull(false)
1895 , m_cursor(0)
1896 {
1897 ASSERT(!m_target.isEmpty());
1898 m_target.replace(noBreakSpace, ' ');
1899 foldQuoteMarks(m_target);
1900 }
1901
~SearchBuffer()1902 inline SearchBuffer::~SearchBuffer()
1903 {
1904 }
1905
reachedBreak()1906 inline void SearchBuffer::reachedBreak()
1907 {
1908 m_cursor = 0;
1909 m_isBufferFull = false;
1910 }
1911
atBreak() const1912 inline bool SearchBuffer::atBreak() const
1913 {
1914 return !m_cursor && !m_isBufferFull;
1915 }
1916
append(UChar c,bool isStart)1917 inline void SearchBuffer::append(UChar c, bool isStart)
1918 {
1919 m_buffer[m_cursor] = c == noBreakSpace ? ' ' : foldQuoteMark(c);
1920 m_isCharacterStartBuffer[m_cursor] = isStart;
1921 if (++m_cursor == m_target.length()) {
1922 m_cursor = 0;
1923 m_isBufferFull = true;
1924 }
1925 }
1926
append(const UChar * characters,size_t length)1927 inline size_t SearchBuffer::append(const UChar* characters, size_t length)
1928 {
1929 ASSERT(length);
1930 if (m_isCaseSensitive) {
1931 append(characters[0], true);
1932 return 1;
1933 }
1934 const int maxFoldedCharacters = 16; // sensible maximum is 3, this should be more than enough
1935 UChar foldedCharacters[maxFoldedCharacters];
1936 bool error;
1937 int numFoldedCharacters = foldCase(foldedCharacters, maxFoldedCharacters, characters, 1, &error);
1938 ASSERT(!error);
1939 ASSERT(numFoldedCharacters);
1940 ASSERT(numFoldedCharacters <= maxFoldedCharacters);
1941 if (!error && numFoldedCharacters) {
1942 numFoldedCharacters = min(numFoldedCharacters, maxFoldedCharacters);
1943 append(foldedCharacters[0], true);
1944 for (int i = 1; i < numFoldedCharacters; ++i)
1945 append(foldedCharacters[i], false);
1946 }
1947 return 1;
1948 }
1949
search(size_t & start)1950 inline size_t SearchBuffer::search(size_t& start)
1951 {
1952 if (!m_isBufferFull)
1953 return 0;
1954 if (!m_isCharacterStartBuffer[m_cursor])
1955 return 0;
1956
1957 size_t tailSpace = m_target.length() - m_cursor;
1958 if (memcmp(&m_buffer[m_cursor], m_target.characters(), tailSpace * sizeof(UChar)) != 0)
1959 return 0;
1960 if (memcmp(&m_buffer[0], m_target.characters() + tailSpace, m_cursor * sizeof(UChar)) != 0)
1961 return 0;
1962
1963 start = length();
1964
1965 // Now that we've found a match once, we don't want to find it again, because those
1966 // are the SearchBuffer semantics, allowing for a buffer where you append more than one
1967 // character at a time. To do this we take advantage of m_isCharacterStartBuffer, but if
1968 // we want to get rid of that in the future we could track this with a separate boolean
1969 // or even move the characters to the start of the buffer and set m_isBufferFull to false.
1970 m_isCharacterStartBuffer[m_cursor] = false;
1971
1972 return start;
1973 }
1974
1975 // Returns the number of characters that were appended to the buffer (what we are searching in).
1976 // That's not necessarily the same length as the passed-in target string, because case folding
1977 // can make two strings match even though they're not the same length.
length() const1978 size_t SearchBuffer::length() const
1979 {
1980 size_t bufferSize = m_target.length();
1981 size_t length = 0;
1982 for (size_t i = 0; i < bufferSize; ++i)
1983 length += m_isCharacterStartBuffer[i];
1984 return length;
1985 }
1986
1987 #endif // !ICU_UNICODE
1988
1989 // --------
1990
rangeLength(const Range * r,bool forSelectionPreservation)1991 int TextIterator::rangeLength(const Range* r, bool forSelectionPreservation)
1992 {
1993 int length = 0;
1994 for (TextIterator it(r, forSelectionPreservation); !it.atEnd(); it.advance())
1995 length += it.length();
1996
1997 return length;
1998 }
1999
subrange(Range * entireRange,int characterOffset,int characterCount)2000 PassRefPtr<Range> TextIterator::subrange(Range* entireRange, int characterOffset, int characterCount)
2001 {
2002 CharacterIterator entireRangeIterator(entireRange);
2003 return characterSubrange(entireRangeIterator, characterOffset, characterCount);
2004 }
2005
rangeFromLocationAndLength(Element * scope,int rangeLocation,int rangeLength,bool forSelectionPreservation)2006 PassRefPtr<Range> TextIterator::rangeFromLocationAndLength(Element* scope, int rangeLocation, int rangeLength, bool forSelectionPreservation)
2007 {
2008 RefPtr<Range> resultRange = scope->document()->createRange();
2009
2010 int docTextPosition = 0;
2011 int rangeEnd = rangeLocation + rangeLength;
2012 bool startRangeFound = false;
2013
2014 RefPtr<Range> textRunRange;
2015
2016 TextIterator it(rangeOfContents(scope).get(), forSelectionPreservation);
2017
2018 // FIXME: the atEnd() check shouldn't be necessary, workaround for <http://bugs.webkit.org/show_bug.cgi?id=6289>.
2019 if (rangeLocation == 0 && rangeLength == 0 && it.atEnd()) {
2020 textRunRange = it.range();
2021
2022 ExceptionCode ec = 0;
2023 resultRange->setStart(textRunRange->startContainer(), 0, ec);
2024 ASSERT(!ec);
2025 resultRange->setEnd(textRunRange->startContainer(), 0, ec);
2026 ASSERT(!ec);
2027
2028 return resultRange.release();
2029 }
2030
2031 for (; !it.atEnd(); it.advance()) {
2032 int len = it.length();
2033 textRunRange = it.range();
2034
2035 bool foundStart = rangeLocation >= docTextPosition && rangeLocation <= docTextPosition + len;
2036 bool foundEnd = rangeEnd >= docTextPosition && rangeEnd <= docTextPosition + len;
2037
2038 // Fix textRunRange->endPosition(), but only if foundStart || foundEnd, because it is only
2039 // in those cases that textRunRange is used.
2040 if (foundStart || foundEnd) {
2041 // FIXME: This is a workaround for the fact that the end of a run is often at the wrong
2042 // position for emitted '\n's.
2043 if (len == 1 && it.characters()[0] == '\n') {
2044 Position runStart = textRunRange->startPosition();
2045 Position runEnd = VisiblePosition(runStart).next().deepEquivalent();
2046 if (runEnd.isNotNull()) {
2047 ExceptionCode ec = 0;
2048 textRunRange->setEnd(runEnd.node(), runEnd.deprecatedEditingOffset(), ec);
2049 ASSERT(!ec);
2050 }
2051 }
2052 }
2053
2054 if (foundStart) {
2055 startRangeFound = true;
2056 int exception = 0;
2057 if (textRunRange->startContainer()->isTextNode()) {
2058 int offset = rangeLocation - docTextPosition;
2059 resultRange->setStart(textRunRange->startContainer(), offset + textRunRange->startOffset(), exception);
2060 } else {
2061 if (rangeLocation == docTextPosition)
2062 resultRange->setStart(textRunRange->startContainer(), textRunRange->startOffset(), exception);
2063 else
2064 resultRange->setStart(textRunRange->endContainer(), textRunRange->endOffset(), exception);
2065 }
2066 }
2067
2068 if (foundEnd) {
2069 int exception = 0;
2070 if (textRunRange->startContainer()->isTextNode()) {
2071 int offset = rangeEnd - docTextPosition;
2072 resultRange->setEnd(textRunRange->startContainer(), offset + textRunRange->startOffset(), exception);
2073 } else {
2074 if (rangeEnd == docTextPosition)
2075 resultRange->setEnd(textRunRange->startContainer(), textRunRange->startOffset(), exception);
2076 else
2077 resultRange->setEnd(textRunRange->endContainer(), textRunRange->endOffset(), exception);
2078 }
2079 docTextPosition += len;
2080 break;
2081 }
2082 docTextPosition += len;
2083 }
2084
2085 if (!startRangeFound)
2086 return 0;
2087
2088 if (rangeLength != 0 && rangeEnd > docTextPosition) { // rangeEnd is out of bounds
2089 int exception = 0;
2090 resultRange->setEnd(textRunRange->endContainer(), textRunRange->endOffset(), exception);
2091 }
2092
2093 return resultRange.release();
2094 }
2095
2096 // --------
2097
plainTextToMallocAllocatedBuffer(const Range * r,unsigned & bufferLength,bool isDisplayString)2098 UChar* plainTextToMallocAllocatedBuffer(const Range* r, unsigned& bufferLength, bool isDisplayString)
2099 {
2100 UChar* result = 0;
2101
2102 // Do this in pieces to avoid massive reallocations if there is a large amount of text.
2103 // Use system malloc for buffers since they can consume lots of memory and current TCMalloc is unable return it back to OS.
2104 static const unsigned cMaxSegmentSize = 1 << 16;
2105 bufferLength = 0;
2106 typedef pair<UChar*, unsigned> TextSegment;
2107 Vector<TextSegment>* textSegments = 0;
2108 Vector<UChar> textBuffer;
2109 textBuffer.reserveInitialCapacity(cMaxSegmentSize);
2110 for (TextIterator it(r); !it.atEnd(); it.advance()) {
2111 if (textBuffer.size() && textBuffer.size() + it.length() > cMaxSegmentSize) {
2112 UChar* newSegmentBuffer = static_cast<UChar*>(malloc(textBuffer.size() * sizeof(UChar)));
2113 if (!newSegmentBuffer)
2114 goto exit;
2115 memcpy(newSegmentBuffer, textBuffer.data(), textBuffer.size() * sizeof(UChar));
2116 if (!textSegments)
2117 textSegments = new Vector<TextSegment>;
2118 textSegments->append(make_pair(newSegmentBuffer, (unsigned)textBuffer.size()));
2119 textBuffer.clear();
2120 }
2121 textBuffer.append(it.characters(), it.length());
2122 bufferLength += it.length();
2123 }
2124
2125 if (!bufferLength)
2126 return 0;
2127
2128 // Since we know the size now, we can make a single buffer out of the pieces with one big alloc
2129 result = static_cast<UChar*>(malloc(bufferLength * sizeof(UChar)));
2130 if (!result)
2131 goto exit;
2132
2133 {
2134 UChar* resultPos = result;
2135 if (textSegments) {
2136 unsigned size = textSegments->size();
2137 for (unsigned i = 0; i < size; ++i) {
2138 const TextSegment& segment = textSegments->at(i);
2139 memcpy(resultPos, segment.first, segment.second * sizeof(UChar));
2140 resultPos += segment.second;
2141 }
2142 }
2143 memcpy(resultPos, textBuffer.data(), textBuffer.size() * sizeof(UChar));
2144 }
2145
2146 exit:
2147 if (textSegments) {
2148 unsigned size = textSegments->size();
2149 for (unsigned i = 0; i < size; ++i)
2150 free(textSegments->at(i).first);
2151 delete textSegments;
2152 }
2153
2154 if (isDisplayString && r->ownerDocument())
2155 r->ownerDocument()->displayBufferModifiedByEncoding(result, bufferLength);
2156
2157 return result;
2158 }
2159
plainText(const Range * r)2160 String plainText(const Range* r)
2161 {
2162 unsigned length;
2163 UChar* buf = plainTextToMallocAllocatedBuffer(r, length, false);
2164 if (!buf)
2165 return "";
2166 String result(buf, length);
2167 free(buf);
2168 return result;
2169 }
2170
isAllCollapsibleWhitespace(const String & string)2171 static inline bool isAllCollapsibleWhitespace(const String& string)
2172 {
2173 const UChar* characters = string.characters();
2174 unsigned length = string.length();
2175 for (unsigned i = 0; i < length; ++i) {
2176 if (!isCollapsibleWhitespace(characters[i]))
2177 return false;
2178 }
2179 return true;
2180 }
2181
collapsedToBoundary(const Range * range,bool forward)2182 static PassRefPtr<Range> collapsedToBoundary(const Range* range, bool forward)
2183 {
2184 ExceptionCode ec = 0;
2185 RefPtr<Range> result = range->cloneRange(ec);
2186 ASSERT(!ec);
2187 result->collapse(!forward, ec);
2188 ASSERT(!ec);
2189 return result.release();
2190 }
2191
findPlainText(CharacterIterator & it,const String & target,bool forward,bool caseSensitive,size_t & matchStart)2192 static size_t findPlainText(CharacterIterator& it, const String& target, bool forward, bool caseSensitive, size_t& matchStart)
2193 {
2194 matchStart = 0;
2195 size_t matchLength = 0;
2196
2197 SearchBuffer buffer(target, caseSensitive);
2198
2199 while (!it.atEnd()) {
2200 it.advance(buffer.append(it.characters(), it.length()));
2201 tryAgain:
2202 size_t matchStartOffset;
2203 if (size_t newMatchLength = buffer.search(matchStartOffset)) {
2204 // Note that we found a match, and where we found it.
2205 size_t lastCharacterInBufferOffset = it.characterOffset();
2206 ASSERT(lastCharacterInBufferOffset >= matchStartOffset);
2207 matchStart = lastCharacterInBufferOffset - matchStartOffset;
2208 matchLength = newMatchLength;
2209 // If searching forward, stop on the first match.
2210 // If searching backward, don't stop, so we end up with the last match.
2211 if (forward)
2212 break;
2213 goto tryAgain;
2214 }
2215 if (it.atBreak() && !buffer.atBreak()) {
2216 buffer.reachedBreak();
2217 goto tryAgain;
2218 }
2219 }
2220
2221 return matchLength;
2222 }
2223
findPlainText(const Range * range,const String & target,bool forward,bool caseSensitive)2224 PassRefPtr<Range> findPlainText(const Range* range, const String& target, bool forward, bool caseSensitive)
2225 {
2226 // First, find the text.
2227 size_t matchStart;
2228 size_t matchLength;
2229 {
2230 CharacterIterator findIterator(range, false, true);
2231 matchLength = findPlainText(findIterator, target, forward, caseSensitive, matchStart);
2232 if (!matchLength)
2233 return collapsedToBoundary(range, forward);
2234 }
2235
2236 // Then, find the document position of the start and the end of the text.
2237 CharacterIterator computeRangeIterator(range, false, true);
2238 return characterSubrange(computeRangeIterator, matchStart, matchLength);
2239 }
2240
2241 }
2242