handle narrow whitespaces in default text extractor #319

where the gap is small but much larger than all previous gaps at this font size (and still larger than some minimum threshold) then break the word at this gap boundary.
UglyToad · May 9, 2021 · 97e831c · 97e831c
1 parent 264cf7b
commit 97e831c
Showing 1 changed file with 37 additions and 2 deletions.
diff --git a/src/UglyToad.PdfPig/Util/DefaultWordExtractor.cs b/src/UglyToad.PdfPig/Util/DefaultWordExtractor.cs
@@ -21,6 +21,8 @@ public IEnumerable<Word> GetWords(IReadOnlyList<Letter> letters)
 
             var lettersSoFar = new List<Letter>(10);
 
+            var gapCountsSoFarByFontSize = new Dictionary<double, Dictionary<double, int>>();
+
             var y = default(double?);
             var lastX = default(double?);
             var lastLetter = default(Letter);
@@ -68,15 +70,48 @@ public IEnumerable<Word> GetWords(IReadOnlyList<Letter> letters)
                     continue;
                 }
 
+                var letterHeight = Math.Max(lastLetter.GlyphRectangle.Height, letter.GlyphRectangle.Height);
+
                 var gap = letter.Location.X - (lastLetter.Location.X + lastLetter.Width);
                 var nextToLeft = letter.Location.X < lastX.Value - 1;
-                var nextBigSpace = gap > Math.Max(lastLetter.GlyphRectangle.Height, letter.GlyphRectangle.Height) * 0.39;
+                var nextBigSpace = gap > letterHeight * 0.39;
                 var nextIsWhiteSpace = string.IsNullOrWhiteSpace(letter.Value);
                 var nextFontDiffers = !string.Equals(letter.FontName, lastLetter.FontName, StringComparison.OrdinalIgnoreCase) && gap > letter.Width * 0.1;
                 var nextFontSizeDiffers = Math.Abs(letter.FontSize - lastLetter.FontSize) > 0.1;
                 var nextTextOrientationDiffers = letter.TextOrientation != lastLetter.TextOrientation;
 
-                if (nextToLeft || nextBigSpace || nextIsWhiteSpace || nextFontDiffers || nextFontSizeDiffers || nextTextOrientationDiffers)
+                var suspectGap = false;
+
+                if (!nextFontSizeDiffers && letter.FontSize > 0 && gap >= 0)
+                {
+                    var fontSize = Math.Round(letter.FontSize);
+                    if (!gapCountsSoFarByFontSize.TryGetValue(fontSize, out var gapCounts))
+                    {
+                        gapCounts = new Dictionary<double, int>();
+                        gapCountsSoFarByFontSize[fontSize] = gapCounts;
+                    }
+
+                    var gapRounded = Math.Round(gap, 2);
+                    if (!gapCounts.ContainsKey(gapRounded))
+                    {
+                        gapCounts[gapRounded] = 0;
+                    }
+
+                    gapCounts[gapRounded]++;
+
+                    // More than one type of gap.
+                    if (gapCounts.Count > 1 && gap > letterHeight * 0.16)
+                    {
+                        var mostCommonGap = gapCounts.OrderByDescending(x => x.Value).First();
+
+                        if (gap > (mostCommonGap.Key * 5) && mostCommonGap.Value > 1)
+                        {
+                            suspectGap = true;
+                        }
+                    }
+                }
+
+                if (nextToLeft || nextBigSpace || nextIsWhiteSpace || nextFontDiffers || nextFontSizeDiffers || nextTextOrientationDiffers || suspectGap)
                 {
                     if (lettersSoFar.Count > 0)
                     {