Refactor sentence parsing (#1215)

* Rename sentenceExtent with sentenceScanExtent

* Update TextScanner.setOptions

* Change function argument order

* Rename quote map variables

* Fix edge case quote handling

* Update terminator maps to support character inclusion

ext/fg/js/frontend.js

@@ -312,7 +312,7 @@ class Frontend {
     async _updateOptionsInternal() {
         const optionsContext = await this._getOptionsContext();
         const options = await api.optionsGet(optionsContext);
-        const scanningOptions = options.scanning;
+        const {scanning: scanningOptions, sentenceParsing: sentenceParsingOptions} = options;
         this._options = options;
 
         await this._updatePopup();
@@ -326,7 +326,7 @@ class Frontend {
             touchInputEnabled: scanningOptions.touchInputEnabled,
             pointerEventsEnabled: scanningOptions.pointerEventsEnabled,
             scanLength: scanningOptions.length,
-            sentenceExtent: options.sentenceParsing.scanExtent,
+            sentenceScanExtent: sentenceParsingOptions.scanExtent,
             layoutAwareScan: scanningOptions.layoutAwareScan,
             preventMiddleMouse
         });

ext/mixed/js/display.js

@@ -309,7 +309,7 @@ class Display extends EventDispatcher {
 
     async updateOptions() {
         const options = await api.optionsGet(this.getOptionsContext());
-        const scanning = options.scanning;
+        const {scanning: scanningOptions, sentenceParsing: sentenceParsingOptions} = options;
         this._options = options;
 
         this._updateDocumentOptions(options);
@@ -320,16 +320,16 @@ class Display extends EventDispatcher {
             selectedParser: options.parsing.selectedParser,
             termSpacing: options.parsing.termSpacing,
             scanning: {
-                inputs: scanning.inputs,
-                deepContentScan: scanning.deepDomScan,
-                selectText: scanning.selectText,
-                delay: scanning.delay,
-                touchInputEnabled: scanning.touchInputEnabled,
-                pointerEventsEnabled: scanning.pointerEventsEnabled,
-                scanLength: scanning.length,
-                sentenceExtent: options.sentenceParsing.scanExtent,
-                layoutAwareScan: scanning.layoutAwareScan,
-                preventMiddleMouse: scanning.preventMiddleMouse.onSearchQuery
+                inputs: scanningOptions.inputs,
+                deepContentScan: scanningOptions.deepDomScan,
+                selectText: scanningOptions.selectText,
+                delay: scanningOptions.delay,
+                touchInputEnabled: scanningOptions.touchInputEnabled,
+                pointerEventsEnabled: scanningOptions.pointerEventsEnabled,
+                scanLength: scanningOptions.length,
+                sentenceScanExtent: sentenceParsingOptions.scanExtent,
+                layoutAwareScan: scanningOptions.layoutAwareScan,
+                preventMiddleMouse: scanningOptions.preventMiddleMouse.onSearchQuery
             }
         });
 
@@ -1810,7 +1810,7 @@ class Display extends EventDispatcher {
             this._definitionTextScanner.on('searched', this._onDefinitionTextScannerSearched.bind(this));
         }
 
-        const scanningOptions = options.scanning;
+        const {scanning: scanningOptions, sentenceParsing: sentenceParsingOptions} = options;
         this._definitionTextScanner.setOptions({
             inputs: [{
                 include: 'mouse0',
@@ -1832,7 +1832,7 @@ class Display extends EventDispatcher {
             touchInputEnabled: false,
             pointerEventsEnabled: false,
             scanLength: scanningOptions.length,
-            sentenceExtent: options.sentenceParsing.scanExtent,
+            sentenceScanExtent: sentenceParsingOptions.scanExtent,
             layoutAwareScan: scanningOptions.layoutAwareScan,
             preventMiddleMouse: false
         });

ext/mixed/js/document-util.js

@@ -31,12 +31,16 @@ class DocumentUtil {
             ['\'', '\''],
             ['"', '"']
         ];
-        this._terminatorSet = new Set(['…', '。', '．', '.', '？', '?', '！', '!']);
-        this._startQuoteMap = new Map();
-        this._endQuoteMap = new Map();
+        const terminatorString = '…。．.？?！!';
+        this._terminatorMap = new Map();
+        for (const char of terminatorString) {
+            this._terminatorMap.set(char, [false, true]);
+        }
+        this._forwardQuoteMap = new Map();
+        this._backwardQuoteMap = new Map();
         for (const [char1, char2] of quoteArray) {
-            this._startQuoteMap.set(char1, char2);
-            this._endQuoteMap.set(char2, char1);
+            this._forwardQuoteMap.set(char1, [char2, false]);
+            this._backwardQuoteMap.set(char2, [char1, false]);
         }
     }
 
@@ -77,10 +81,10 @@ class DocumentUtil {
         }
     }
 
-    extractSentence(source, extent, layoutAwareScan) {
-        const terminatorSet = this._terminatorSet;
-        const startQuoteMap = this._startQuoteMap;
-        const endQuoteMap = this._endQuoteMap;
+    extractSentence(source, layoutAwareScan, extent) {
+        const terminatorMap = this._terminatorMap;
+        const forwardQuoteMap = this._forwardQuoteMap;
+        const backwardQuoteMap = this._backwardQuoteMap;
 
         // Scan text
         source = source.clone();
@@ -98,22 +102,28 @@ class DocumentUtil {
             const c = text[pos1 - 1];
             if (c === '\n') { break; }
 
-            if (quoteStack.length === 0 && terminatorSet.has(c)) {
+            if (quoteStack.length === 0) {
+                const terminatorInfo = terminatorMap.get(c);
+                if (typeof terminatorInfo !== 'undefined') {
+                    if (terminatorInfo[0]) { --pos1; }
                     break;
                 }
+            }
 
-            let otherQuote = startQuoteMap.get(c);
-            if (typeof otherQuote !== 'undefined') {
+            let quoteInfo = forwardQuoteMap.get(c);
+            if (typeof quoteInfo !== 'undefined') {
                 if (quoteStack.length === 0) {
+                    if (quoteInfo[1]) { --pos1; }
                     break;
                 } else if (quoteStack[0] === c) {
                     quoteStack.pop();
+                    continue;
                 }
-            } else {
-                otherQuote = endQuoteMap.get(c);
-                if (typeof otherQuote !== 'undefined') {
-                    quoteStack.unshift(otherQuote);
             }
+
+            quoteInfo = backwardQuoteMap.get(c);
+            if (typeof quoteInfo !== 'undefined') {
+                quoteStack.unshift(quoteInfo[0]);
             }
         }
 
@@ -123,23 +133,28 @@ class DocumentUtil {
             const c = text[pos2];
             if (c === '\n') { break; }
 
-            if (quoteStack.length === 0 && terminatorSet.has(c)) {
-                ++pos2;
+            if (quoteStack.length === 0) {
+                const terminatorInfo = terminatorMap.get(c);
+                if (typeof terminatorInfo !== 'undefined') {
+                    if (terminatorInfo[1]) { ++pos2; }
                     break;
                 }
+            }
 
-            let otherQuote = endQuoteMap.get(c);
-            if (typeof otherQuote !== 'undefined') {
+            let quoteInfo = backwardQuoteMap.get(c);
+            if (typeof quoteInfo !== 'undefined') {
                 if (quoteStack.length === 0) {
+                    if (quoteInfo[1]) { ++pos2; }
                     break;
                 } else if (quoteStack[0] === c) {
                     quoteStack.pop();
+                    continue;
                 }
-            } else {
-                otherQuote = startQuoteMap.get(c);
-                if (typeof otherQuote !== 'undefined') {
-                    quoteStack.unshift(otherQuote);
             }
+
+            quoteInfo = forwardQuoteMap.get(c);
+            if (typeof quoteInfo !== 'undefined') {
+                quoteStack.unshift(quoteInfo[0]);
             }
         }
 

ext/mixed/js/text-scanner.js

@@ -59,7 +59,7 @@ class TextScanner extends EventDispatcher {
         this._touchInputEnabled = false;
         this._pointerEventsEnabled = false;
         this._scanLength = 1;
-        this._sentenceExtent = 1;
+        this._sentenceScanExtent = 1;
         this._layoutAwareScan = false;
         this._preventMiddleMouse = false;
         this._inputs = [];
@@ -134,7 +134,18 @@ class TextScanner extends EventDispatcher {
         }
     }
 
-    setOptions({inputs, deepContentScan, selectText, delay, touchInputEnabled, pointerEventsEnabled, scanLength, sentenceExtent, layoutAwareScan, preventMiddleMouse}) {
+    setOptions({
+        inputs,
+        deepContentScan,
+        selectText,
+        delay,
+        touchInputEnabled,
+        pointerEventsEnabled,
+        scanLength,
+        sentenceScanExtent,
+        layoutAwareScan,
+        preventMiddleMouse
+    }) {
         if (Array.isArray(inputs)) {
             this._inputs = inputs.map(({
                 include,
@@ -182,8 +193,8 @@ class TextScanner extends EventDispatcher {
         if (typeof scanLength === 'number') {
             this._scanLength = scanLength;
         }
-        if (typeof sentenceExtent === 'number') {
-            this._sentenceExtent = sentenceExtent;
+        if (typeof sentenceScanExtent === 'number') {
+            this._sentenceScanExtent = sentenceScanExtent;
         }
         if (typeof layoutAwareScan === 'boolean') {
             this._layoutAwareScan = layoutAwareScan;
@@ -711,7 +722,7 @@ class TextScanner extends EventDispatcher {
 
     async _findTerms(textSource, optionsContext) {
         const scanLength = this._scanLength;
-        const sentenceExtent = this._sentenceExtent;
+        const sentenceScanExtent = this._sentenceScanExtent;
         const layoutAwareScan = this._layoutAwareScan;
         const searchText = this.getTextSourceContent(textSource, scanLength, layoutAwareScan);
         if (searchText.length === 0) { return null; }
@@ -720,13 +731,13 @@ class TextScanner extends EventDispatcher {
         if (definitions.length === 0) { return null; }
 
         textSource.setEndOffset(length, layoutAwareScan);
-        const sentence = this._documentUtil.extractSentence(textSource, sentenceExtent, layoutAwareScan);
+        const sentence = this._documentUtil.extractSentence(textSource, layoutAwareScan, sentenceScanExtent);
 
         return {definitions, sentence, type: 'terms'};
     }
 
     async _findKanji(textSource, optionsContext) {
-        const sentenceExtent = this._sentenceExtent;
+        const sentenceScanExtent = this._sentenceScanExtent;
         const layoutAwareScan = this._layoutAwareScan;
         const searchText = this.getTextSourceContent(textSource, 1, layoutAwareScan);
         if (searchText.length === 0) { return null; }
@@ -735,7 +746,7 @@ class TextScanner extends EventDispatcher {
         if (definitions.length === 0) { return null; }
 
         textSource.setEndOffset(1, layoutAwareScan);
-        const sentence = this._documentUtil.extractSentence(textSource, sentenceExtent, layoutAwareScan);
+        const sentence = this._documentUtil.extractSentence(textSource, layoutAwareScan, sentenceScanExtent);
 
         return {definitions, sentence, type: 'kanji'};
     }

test/data/html/test-document1.html

@@ -21,7 +21,7 @@
         data-end-node-selector="span"
         data-end-offset="0"
         data-result-type="TextSourceRange",
-        data-sentence-extent="100"
+        data-sentence-scan-extent="100"
         data-sentence="真白「心配してくださって、ありがとございます」"
     >
         <span>真白「心配してくださって、ありがとございます」</span>
@@ -37,7 +37,7 @@
         data-end-node-selector="span"
         data-end-offset="5"
         data-result-type="TextSourceRange",
-        data-sentence-extent="100"
+        data-sentence-scan-extent="100"
         data-sentence="心配してくださって、ありがとございます"
     >
         <span>真白「心配してくださって、ありがとございます」</span>
@@ -53,7 +53,7 @@
         data-end-node-selector="span"
         data-end-offset="16"
         data-result-type="TextSourceRange",
-        data-sentence-extent="100"
+        data-sentence-scan-extent="100"
         data-sentence="心配して「くださって」、ありがと「ございます」"
     >
         <span>真白「心配して「くださって」、ありがと「ございます」」</span>
@@ -69,7 +69,7 @@
         data-end-node-selector="span"
         data-end-offset="4"
         data-result-type="TextSourceRange",
-        data-sentence-extent="100"
+        data-sentence-scan-extent="100"
         data-sentence="ありがとございます。"
     >
         <span>ありがとございます。ありがとございます。</span>
@@ -85,7 +85,7 @@
         data-end-node-selector="span"
         data-end-offset="14"
         data-result-type="TextSourceRange",
-        data-sentence-extent="100"
+        data-sentence-scan-extent="100"
         data-sentence="ありがとございます。"
     >
         <span>ありがとございます。ありがとございます。</span>
@@ -101,7 +101,7 @@
         data-end-node-selector="input"
         data-end-offset="0"
         data-result-type="TextSourceRange",
-        data-sentence-extent="100"
+        data-sentence-scan-extent="100"
         data-sentence="真白「心配してくださって、ありがとございます」"
         data-has-imposter="true"
     >
@@ -118,7 +118,7 @@
         data-end-node-selector="textarea"
         data-end-offset="0"
         data-result-type="TextSourceRange",
-        data-sentence-extent="100"
+        data-sentence-scan-extent="100"
         data-sentence="真白「心配してくださって、ありがとございます」"
         data-has-imposter="true"
     >
@@ -135,7 +135,7 @@
         data-end-node-selector="button"
         data-end-offset="0"
         data-result-type="TextSourceElement",
-        data-sentence-extent="100"
+        data-sentence-scan-extent="100"
         data-sentence="よみちゃん"
     >
         <button style="width: 100%; box-sizing: border-box; font-family: inherit; font-size: inherit; border: 1px solid #d8d8d8; background-color: #f0f0f0; padding: 0.2em;">よみちゃん</button>
@@ -151,7 +151,7 @@
         data-end-node-selector="img"
         data-end-offset="0"
         data-result-type="TextSourceElement"
-        data-sentence-extent="100"
+        data-sentence-scan-extent="100"
         data-sentence="よみちゃん"
     >
         <img src="data:image/gif;base64,R0lGODdhBwAHAIABAAAAAP///ywAAAAABwAHAAACDIRvEaC32FpCbEkKCgA7" alt="よみちゃん" title="よみちゃん" style="width: 70px; height: 70px; image-rendering: crisp-edges; image-rendering: pixelated; display: block;" />

test/test-document-util.js

@@ -127,7 +127,7 @@ async function testDocumentTextScanningFunctions(dom, {DocumentUtil, TextSourceR
             endNodeSelector,
             endOffset,
             resultType,
-            sentenceExtent,
+            sentenceScanExtent,
             sentence,
             hasImposter
         } = testElement.dataset;
@@ -139,7 +139,7 @@ async function testDocumentTextScanningFunctions(dom, {DocumentUtil, TextSourceR
 
         startOffset = parseInt(startOffset, 10);
         endOffset = parseInt(endOffset, 10);
-        sentenceExtent = parseInt(sentenceExtent, 10);
+        sentenceScanExtent = parseInt(sentenceScanExtent, 10);
 
         assert.notStrictEqual(elementFromPointValue, null);
         assert.notStrictEqual(caretRangeFromPointValue, null);
@@ -182,7 +182,7 @@ async function testDocumentTextScanningFunctions(dom, {DocumentUtil, TextSourceR
         if (source === null) { continue; }
 
         // Test docSentenceExtract
-        const sentenceActual = documentUtil.extractSentence(source, sentenceExtent, false).text;
+        const sentenceActual = documentUtil.extractSentence(source, false, sentenceScanExtent).text;
         assert.strictEqual(sentenceActual, sentence);
 
         // Clean