fix: #355, #361, #319: calculate text block gap and spacewidth

base/core/evaluator.js

@@ -987,6 +987,38 @@ var PartialEvaluator = (function PartialEvaluatorClosure() {
               bidiText.x += renderParams.vScale / 2;
               bidiText.y -= renderParams.vScale;
             }
+
+            // MQZ: Add font metrics for accurate spacing calculation
+            bidiText.fontName = font.loadedName || font.name;
+            bidiText.fontSize = textState.fontSize;
+
+            // Get fontMatrix once (used for both spaceWidth and textWidth calculations)
+            var fontMatrix = font.fontMatrix || FONT_IDENTITY_MATRIX;
+            var fontDirection = textState.fontDirection || 1;
+
+            // Scale spaceWidth to PDF coordinates using fontMatrix (NO textHScale)
+            // Must match canvas.js canvasWidth calculation (line 1258 - no textHScale)
+            bidiText.spaceWidth = font.spaceWidth * textState.fontSize * fontMatrix[0];
+            bidiText.charSpace = charSpace;
+            bidiText.wordSpace = wordSpace;
+            bidiText.textHScale = textState.textHScale;
+
+            // Calculate actual text width using font glyph widths
+            // Match canvas.js calculation exactly (lines 1210-1211, 1258, canvasWidth does NOT include textHScale)
+            var textWidth = 0;
+            var glyphs = font.charsToGlyphs(chunk);
+            for (var i = 0, ii = glyphs.length; i < ii; i++) {
+              var glyph = glyphs[i];
+              // Use glyph.width if available, otherwise font.defaultWidth (like canvas.js does)
+              var glyphWidth = (glyph && glyph.width) || font.defaultWidth || 0;
+              // Match canvas.js line 1210-1211: width * fontSize * fontMatrix[0] + charSpacing * fontDirection
+              var charWidth = glyphWidth * textState.fontSize * fontMatrix[0] + charSpace * fontDirection;
+              textWidth += charWidth;
+            }
+            // DO NOT apply textHScale - canvasWidth is in unscaled coordinates
+            // (bidiText.x is scaled, but bidiText.width matches JSON w property which is unscaled)
+            bidiText.width = textWidth;
+
             bidiTexts.push(bidiText);
 
             chunk = '';

lib/pdf.js

@@ -327,18 +327,64 @@ export default class PDFJSClass extends EventEmitter {
 
     this.rawTextContents.forEach((textContent, index) => {
       let prevText = null;
+
       textContent.bidiTexts.forEach((textObj, idx) => {
-        if (prevText) {
-          if (Math.abs(textObj.y - prevText.y) <= 9) {
-            prevText.str += textObj.str;
-          } else {
-            retVal += `${prevText.str}\r\n`;
-            prevText = textObj;
+        // Check if on same line
+        // Use a tolerance relative to font size for better accuracy
+        // Typical line spacing is 120% of font size, so 10-15% tolerance is reasonable
+        const tolerance = prevText ? (prevText.fontSize || 12) * 0.15 : 2;
+        const sameLine = prevText && Math.abs(textObj.y - prevText.y) <= tolerance;
+
+        if (sameLine) {
+          // spaceWidth is in unscaled coordinates (no textHScale, matching JSON w property)
+          const { spaceWidth, startX, width, textHScale } = prevText;
+
+          // Use actual calculated text width (from glyph widths)
+          // width is in unscaled coordinates, but startX is in scaled coordinates
+          // So we must apply textHScale to width before adding to startX
+          // This matches canvas.js: current.x += x * textHScale (line 1267)
+          const prevTextEndX = startX + (width * textHScale);
+
+          // Calculate gap between end of previous text and start of current text
+          // gap is in SCALED coordinates (both textObj.x and prevTextEndX are scaled)
+          const gap = textObj.x - prevTextEndX;
+
+          // Scale spaceWidth to match gap's coordinate system
+          const scaledSpaceWidth = spaceWidth * textHScale;
+
+          // Add spaces if gap is positive and significant (> 30% of scaled space width)
+          if (gap > scaledSpaceWidth * 0.3) {
+            const numSpaces = Math.round(gap / scaledSpaceWidth);
+            prevText.str += ' '.repeat(Math.max(1, numSpaces));
           }
+
+          // Append current text
+          prevText.str += textObj.str;
+
+          // Update prevText to track current text for next iteration
+          prevText.startX = textObj.x;
+          prevText.width = textObj.width;
+          prevText.spaceWidth = textObj.spaceWidth;
+          prevText.textHScale = textObj.textHScale;
         } else {
-          prevText = textObj;
+          // Different line or first text
+          if (prevText) {
+            retVal += `${prevText.str}\r\n`;
+          }
+
+          // Initialize new text object with font metrics
+          prevText = { 
+            str: textObj.str, 
+            y: textObj.y, 
+            startX: textObj.x,
+            width: textObj.width,
+            spaceWidth: textObj.spaceWidth,
+            textHScale: textObj.textHScale,
+            fontSize: textObj.fontSize  // Keep for tolerance calculation
+          };
         }
       });
+
       if (prevText) {
         retVal += prevText.str;
       }