Title: [260578] trunk
Revision
260578
Author
sihui_...@apple.com
Date
2020-04-23 10:23:54 -0700 (Thu, 23 Apr 2020)

Log Message

TextManipulationController should set range of paragraph using token's positions
https://bugs.webkit.org/show_bug.cgi?id=210866
<rdar://problem/60646283>

Reviewed by Wenson Hsieh.

Source/WebCore:

Set the range of paragraph using positions of first token and last token in the paragraph because:
1. Accurate range makes token matching in TextManipulationController::replace() easier, as TextIterator could
visit different positions with different ranges or different conditions. For example, in our previous
implementation, start of a paragraph can be set as the first visible position of document, while position of
first token is after that. Then in replace(), TextManipulationController may extract a word before the position
of first token and return error. See added test TextManipulation.CompleteTextManipulationCorrectParagraphRange.
2. TextManipulationController can handle fewer content and this is less error-prone. For example, svg elements
before/after the paragraph text will not be identified as tokens [] in a paragraph now. See updated API tests
for example.

New test: TextManipulation.CompleteTextManipulationCorrectParagraphRange

* editing/TextManipulationController.cpp:
(WebCore::ParagraphContentIterator::moveCurrentNodeForward): m_currentNodeForFindingInvisibleContent should not
be advanced if it is already at the end.
(WebCore::containsOnlyHTMLSpaces):
(WebCore::TextManipulationController::observeParagraphs):Set the paragraph start as the position of the first
token and end as the position of last token. If the paragraph is split with <br>, the end will be extended to
position of <br> so that we can add this node back later; otherwise, <br> can be removed after original
text of paragraph is removed in TextManipulationController::replace(). Also, stop identifying spaces as tokens
because non-text Node can emit spaces.
(WebCore::TextManipulationController::replace): Only identify tokens from content with meaningful text.

Tools:

* TestWebKitAPI/Tests/WebKitCocoa/TextManipulation.mm:
(TestWebKitAPI::TEST):

Modified Paths

Diff

Modified: trunk/Source/WebCore/ChangeLog (260577 => 260578)


--- trunk/Source/WebCore/ChangeLog	2020-04-23 17:23:18 UTC (rev 260577)
+++ trunk/Source/WebCore/ChangeLog	2020-04-23 17:23:54 UTC (rev 260578)
@@ -1,3 +1,34 @@
+2020-04-23  Sihui Liu  <sihui_...@apple.com>
+
+        TextManipulationController should set range of paragraph using token's positions
+        https://bugs.webkit.org/show_bug.cgi?id=210866
+        <rdar://problem/60646283>
+
+        Reviewed by Wenson Hsieh.
+
+        Set the range of paragraph using positions of first token and last token in the paragraph because:
+        1. Accurate range makes token matching in TextManipulationController::replace() easier, as TextIterator could 
+        visit different positions with different ranges or different conditions. For example, in our previous 
+        implementation, start of a paragraph can be set as the first visible position of document, while position of 
+        first token is after that. Then in replace(), TextManipulationController may extract a word before the position 
+        of first token and return error. See added test TextManipulation.CompleteTextManipulationCorrectParagraphRange.
+        2. TextManipulationController can handle fewer content and this is less error-prone. For example, svg elements 
+        before/after the paragraph text will not be identified as tokens [] in a paragraph now. See updated API tests 
+        for example.
+
+        New test: TextManipulation.CompleteTextManipulationCorrectParagraphRange
+
+        * editing/TextManipulationController.cpp:
+        (WebCore::ParagraphContentIterator::moveCurrentNodeForward): m_currentNodeForFindingInvisibleContent should not 
+        be advanced if it is already at the end.
+        (WebCore::containsOnlyHTMLSpaces):
+        (WebCore::TextManipulationController::observeParagraphs):Set the paragraph start as the position of the first 
+        token and end as the position of last token. If the paragraph is split with <br>, the end will be extended to 
+        position of <br> so that we can add this node back later; otherwise, <br> can be removed after original 
+        text of paragraph is removed in TextManipulationController::replace(). Also, stop identifying spaces as tokens 
+        because non-text Node can emit spaces.
+        (WebCore::TextManipulationController::replace): Only identify tokens from content with meaningful text.
+
 2020-04-23  Chris Dumez  <cdu...@apple.com>
 
         [ Mac wk2 ] imported/w3c/web-platform-tests/notifications/event-onclose.html is flaky failing.

Modified: trunk/Source/WebCore/editing/TextManipulationController.cpp (260577 => 260578)


--- trunk/Source/WebCore/editing/TextManipulationController.cpp	2020-04-23 17:23:18 UTC (rev 260577)
+++ trunk/Source/WebCore/editing/TextManipulationController.cpp	2020-04-23 17:23:54 UTC (rev 260578)
@@ -30,8 +30,10 @@
 #include "Editing.h"
 #include "ElementAncestorIterator.h"
 #include "EventLoop.h"
+#include "HTMLBRElement.h"
 #include "HTMLElement.h"
 #include "HTMLNames.h"
+#include "HTMLParserIdioms.h"
 #include "NodeTraversal.h"
 #include "PseudoElement.h"
 #include "Range.h"
@@ -207,6 +209,9 @@
 private:
     void moveCurrentNodeForward()
     {
+        if (m_currentNodeForFindingInvisibleContent == m_pastEndNode)
+            return;
+
         m_currentNodeForFindingInvisibleContent = NodeTraversal::next(*m_currentNodeForFindingInvisibleContent);
         if (!m_currentNodeForFindingInvisibleContent)
             m_currentNodeForFindingInvisibleContent = m_pastEndNode;
@@ -242,6 +247,16 @@
     return element.hasTagName(HTMLNames::titleTag) || element.hasTagName(HTMLNames::optionTag);
 }
 
+static bool containsOnlyHTMLSpaces(StringView text)
+{
+    for (unsigned index = 0; index < text.length(); ++index) {
+        if (isNotHTMLSpace(text[index]))
+            return false;
+    }
+
+    return true;
+}
+
 void TextManipulationController::observeParagraphs(const Position& start, const Position& end)
 {
     if (start.isNull() || end.isNull())
@@ -257,7 +272,8 @@
 
     ExclusionRuleMatcher exclusionRuleMatcher(m_exclusionRules);
     Vector<ManipulationToken> tokensInCurrentParagraph;
-    Position startOfCurrentParagraph = visibleStart.deepEquivalent();
+    Position startOfCurrentParagraph;
+    Position endOfCurrentParagraph;
     for (; !iterator.atEnd(); iterator.advance()) {
         auto content = iterator.currentContent();
         if (content.node) {
@@ -281,15 +297,20 @@
                     }
                 }
             }
-
-            if (startOfCurrentParagraph.isNull() && content.isTextContent)
-                startOfCurrentParagraph = iterator.startPosition();
         }
 
         if (content.isReplacedContent) {
-            if (startOfCurrentParagraph.isNull())
-                startOfCurrentParagraph = positionBeforeNode(content.node.get());
-            tokensInCurrentParagraph.append(ManipulationToken { m_tokenIdentifier.generate(), "[]", true /* isExcluded */});
+            if (tokensInCurrentParagraph.isEmpty())
+                continue;
+
+            auto currentEndOfCurrentParagraph = positionAfterNode(content.node.get());
+            // This is at the same Node as last token, so it is already included in current range.
+            if (!is<Text>(content.node) && currentEndOfCurrentParagraph.equals(endOfCurrentParagraph))
+                continue;
+
+            endOfCurrentParagraph = currentEndOfCurrentParagraph;
+            tokensInCurrentParagraph.append(ManipulationToken { m_tokenIdentifier.generate(), "[]", true });
+
             continue;
         }
 
@@ -300,31 +321,39 @@
         size_t offsetOfNextNewLine = 0;
         StringView currentText = content.text;
         while ((offsetOfNextNewLine = currentText.find('\n', startOfCurrentLine)) != notFound) {
-            if (startOfCurrentLine < offsetOfNextNewLine) {
+            if (is<Text>(content.node) && startOfCurrentLine < offsetOfNextNewLine) {
                 auto stringUntilEndOfLine = currentText.substring(startOfCurrentLine, offsetOfNextNewLine - startOfCurrentLine).toString();
+                auto& textNode = downcast<Text>(*content.node);
+                endOfCurrentParagraph = Position(&textNode, offsetOfNextNewLine);
+                if (tokensInCurrentParagraph.isEmpty())
+                    startOfCurrentParagraph = Position(&textNode, startOfCurrentLine);
+
                 tokensInCurrentParagraph.append(ManipulationToken { m_tokenIdentifier.generate(), stringUntilEndOfLine, exclusionRuleMatcher.isExcluded(content.node.get()) });
             }
 
             if (!tokensInCurrentParagraph.isEmpty()) {
-                Position endOfCurrentParagraph = iterator.endPosition();
-                if (is<Text>(content.node)) {
-                    auto& textNode = downcast<Text>(*content.node);
-                    endOfCurrentParagraph = Position(&textNode, offsetOfNextNewLine);
-                    startOfCurrentParagraph = Position(&textNode, offsetOfNextNewLine + 1);
-                }
+                if (is<HTMLBRElement>(content.node))
+                    endOfCurrentParagraph = positionAfterNode(content.node.get());
                 addItem(ManipulationItemData { startOfCurrentParagraph, endOfCurrentParagraph, nullptr, nullQName(), std::exchange(tokensInCurrentParagraph, { }) });
-                startOfCurrentParagraph.clear();
             }
             startOfCurrentLine = offsetOfNextNewLine + 1;
         }
 
         auto remainingText = currentText.substring(startOfCurrentLine);
-        if (remainingText.length())
+        if (!containsOnlyHTMLSpaces(remainingText)) {
+            if (tokensInCurrentParagraph.isEmpty()) {
+                if (startOfCurrentLine && is<Text>(content.node))
+                    startOfCurrentParagraph = Position(&downcast<Text>(*content.node), startOfCurrentLine);
+                else
+                    startOfCurrentParagraph = iterator.startPosition();
+            }
+            endOfCurrentParagraph = iterator.endPosition();
             tokensInCurrentParagraph.append(ManipulationToken { m_tokenIdentifier.generate(), remainingText.toString(), exclusionRuleMatcher.isExcluded(content.node.get()) });
+        }
     }
 
     if (!tokensInCurrentParagraph.isEmpty())
-        addItem(ManipulationItemData { startOfCurrentParagraph, visibleEnd.deepEquivalent(), nullptr, nullQName(), WTFMove(tokensInCurrentParagraph) });
+        addItem(ManipulationItemData { startOfCurrentParagraph, endOfCurrentParagraph, nullptr, nullQName(), WTFMove(tokensInCurrentParagraph) });
 }
 
 void TextManipulationController::didCreateRendererForElement(Element& element)
@@ -519,6 +548,13 @@
             return ManipulationFailureType::ContentChanged;
         }
 
+        if (content.isTextContent && containsOnlyHTMLSpaces(content.text)) {
+            // <br> should not exist in the middle of a paragraph.
+            if (is<HTMLBRElement>(content.node))
+                return ManipulationFailureType::ContentChanged;
+            continue;
+        }
+
         auto& currentToken = item.tokens[currentTokenIndex];
         if (!content.isReplacedContent && content.text != currentToken.content)
             return ManipulationFailureType::ContentChanged;

Modified: trunk/Tools/ChangeLog (260577 => 260578)


--- trunk/Tools/ChangeLog	2020-04-23 17:23:18 UTC (rev 260577)
+++ trunk/Tools/ChangeLog	2020-04-23 17:23:54 UTC (rev 260578)
@@ -1,3 +1,14 @@
+2020-04-23  Sihui Liu  <sihui_...@apple.com>
+
+        TextManipulationController should set range of paragraph using token's positions
+        https://bugs.webkit.org/show_bug.cgi?id=210866
+        <rdar://problem/60646283>
+
+        Reviewed by Wenson Hsieh.
+
+        * TestWebKitAPI/Tests/WebKitCocoa/TextManipulation.mm:
+        (TestWebKitAPI::TEST):
+
 2020-04-23  Emilio Cobos Álvarez  <emi...@crisal.io>
 
         Unreviewed, add my bugzilla / slack nick to contributors.json

Modified: trunk/Tools/TestWebKitAPI/Tests/WebKitCocoa/TextManipulation.mm (260577 => 260578)


--- trunk/Tools/TestWebKitAPI/Tests/WebKitCocoa/TextManipulation.mm	2020-04-23 17:23:18 UTC (rev 260577)
+++ trunk/Tools/TestWebKitAPI/Tests/WebKitCocoa/TextManipulation.mm	2020-04-23 17:23:54 UTC (rev 260578)
@@ -988,22 +988,14 @@
     TestWebKitAPI::Util::run(&done);
 
     auto *items = [delegate items];
-    EXPECT_EQ(items.count, 2UL);
-    EXPECT_EQ(items[0].tokens.count, 2UL);
-    EXPECT_STREQ("[]", items[0].tokens[0].content.UTF8String);
-    EXPECT_TRUE(items[0].tokens[0].isExcluded);
-    EXPECT_STREQ("[]", items[0].tokens[1].content.UTF8String);
-    EXPECT_TRUE(items[0].tokens[1].isExcluded);
+    EXPECT_EQ(items.count, 1UL);
+    EXPECT_EQ(items[0].tokens.count, 1UL);
+    EXPECT_STREQ("helllo world", items[0].tokens[0].content.UTF8String);
+    EXPECT_TRUE(!items[0].tokens[0].isExcluded);
 
-    auto *tokens = items[1].tokens;
-    EXPECT_EQ(tokens.count, 1UL);
-    EXPECT_STREQ("helllo world", tokens[0].content.UTF8String);
-    EXPECT_FALSE(tokens[0].isExcluded);
-
     done = false;
     [webView _completeTextManipulationForItems:@[
-        (_WKTextManipulationItem *)createItem(items[0].identifier, { { items[0].tokens[0].identifier, nil } }),
-        (_WKTextManipulationItem *)createItem(items[1].identifier, { { items[1].tokens[0].identifier, @"hello, world" } }),
+        (_WKTextManipulationItem *)createItem(items[0].identifier, { { items[0].tokens[0].identifier, @"hello, world" } }),
     ] completion:^(NSArray<NSError *> *errors) {
         EXPECT_EQ(errors, nil);
         done = true;
@@ -1076,16 +1068,13 @@
     TestWebKitAPI::Util::run(&done);
 
     auto *items = [delegate items];
-    EXPECT_EQ(items.count, 2UL);
-    EXPECT_EQ(items[1].tokens.count, 1UL);
-    EXPECT_STREQ("[]", items[0].tokens[0].content.UTF8String);
+    EXPECT_EQ(items.count, 1UL);
     EXPECT_EQ(items[0].tokens.count, 1UL);
-    EXPECT_STREQ("hello world", items[1].tokens[0].content.UTF8String);
+    EXPECT_STREQ("hello world", items[0].tokens[0].content.UTF8String);
 
     done = false;
     [webView _completeTextManipulationForItems:@[
-        (_WKTextManipulationItem *)createItem(items[0].identifier, { { items[0].tokens[0].identifier, nil } }),
-        (_WKTextManipulationItem *)createItem(items[1].identifier, { { items[1].tokens[0].identifier, @"hello, world" } }),
+        (_WKTextManipulationItem *)createItem(items[0].identifier, { { items[0].tokens[0].identifier, @"hello, world" } }),
     ] completion:^(NSArray<NSError *> *errors) {
         EXPECT_EQ(errors, nil);
         done = true;
@@ -1162,19 +1151,16 @@
     TestWebKitAPI::Util::run(&done);
 
     auto *items = [delegate items];
-    EXPECT_EQ(items.count, 3UL);
+    EXPECT_EQ(items.count, 2UL);
     EXPECT_EQ(items[0].tokens.count, 1UL);
     EXPECT_STREQ("heeey", items[0].tokens[0].content.UTF8String);
     EXPECT_EQ(items[1].tokens.count, 1UL);
-    EXPECT_STREQ("[]", items[1].tokens[0].content.UTF8String);
-    EXPECT_EQ(items[2].tokens.count, 1UL);
-    EXPECT_STREQ("woorld", items[2].tokens[0].content.UTF8String);
+    EXPECT_STREQ("woorld", items[1].tokens[0].content.UTF8String);
 
     done = false;
     [webView _completeTextManipulationForItems:@[
         (_WKTextManipulationItem *)createItem(items[0].identifier, { { items[0].tokens[0].identifier, @"hello" } }),
-        (_WKTextManipulationItem *)createItem(items[1].identifier, { { items[1].tokens[0].identifier, nil } }),
-        (_WKTextManipulationItem *)createItem(items[2].identifier, { { items[2].tokens[0].identifier, @"world" } }),
+        (_WKTextManipulationItem *)createItem(items[1].identifier, { { items[1].tokens[0].identifier, @"world" } }),
     ] completion:^(NSArray<NSError *> *errors) {
         EXPECT_EQ(errors, nil);
         done = true;
@@ -1684,6 +1670,40 @@
     EXPECT_WK_STREQ("<p>bar <strong>garply</strong> foo</p>", [webView stringByEvaluatingJavaScript:@"document.body.innerHTML"]);
 }
 
+TEST(TextManipulation, CompleteTextManipulationCorrectParagraphRange)
+{
+    auto delegate = adoptNS([[TextManipulationDelegate alloc] init]);
+    auto webView = adoptNS([[TestWKWebView alloc] initWithFrame:NSMakeRect(0, 0, 400, 400)]);
+    [webView _setTextManipulationDelegate:delegate.get()];
+
+    [webView synchronouslyLoadHTMLString:@"<head><style>ul{display:block}li{display:inline-block}.inline {float: left;}.subframe {height: 42px;}.frame {position: absolute;top: -9999px;}</style></head><body><div class='frame'><div class='subframe'></div></div><style></style><div class='inline'><div><li><a href=''>holle</a></li><li><a href=''>wdrlo</a></li></div></div><div class='frame'><div class='subframe'></div></div></body>"];
+
+    RetainPtr<_WKTextManipulationConfiguration> configuration = adoptNS([[_WKTextManipulationConfiguration alloc] init]);
+    done = false;
+    [webView _startTextManipulationsWithConfiguration:configuration.get() completion:^{
+        done = true;
+    }];
+    TestWebKitAPI::Util::run(&done);
+
+    auto *items = [delegate items];
+    EXPECT_EQ(items.count, 1UL);
+    EXPECT_EQ(items[0].tokens.count, 2UL);
+    EXPECT_STREQ("holle", items[0].tokens[0].content.UTF8String);
+    EXPECT_STREQ("wdrlo", items[0].tokens[1].content.UTF8String);
+
+    done = false;
+    [webView _completeTextManipulationForItems:@[(_WKTextManipulationItem *)createItem(items[0].identifier, {
+        { items[0].tokens[0].identifier, @"hello" },
+        { items[0].tokens[1].identifier, @"world" },
+    })] completion:^(NSArray<NSError *> *errors) {
+        EXPECT_EQ(errors, nil);
+        done = true;
+    }];
+    TestWebKitAPI::Util::run(&done);
+
+    EXPECT_WK_STREQ("<div class=\"frame\"><div class=\"subframe\"></div></div><style></style><div class=\"inline\"><div><li><a href="" href="" class=\"frame\"><div class=\"subframe\"></div></div>", [webView stringByEvaluatingJavaScript:@"document.body.innerHTML"]);
+}
+
 TEST(TextManipulation, InsertingContentIntoAlreadyManipulatedContentDoesNotCreateTextManipulationItem)
 {
     auto delegate = adoptNS([[TextManipulationDelegate alloc] init]);
_______________________________________________
webkit-changes mailing list
webkit-changes@lists.webkit.org
https://lists.webkit.org/mailman/listinfo/webkit-changes

Reply via email to