vendor: import KillerPDF 1.7.5 source (GPL-3.0) as the base for MMD PDF

Services/SearchService.cs

@@ -0,0 +1,91 @@
+using UglyToad.PdfPig;
+
+namespace KillerPDF.Services
+{
+    internal sealed class SearchResult
+    {
+        public Dictionary<int, List<(double Left, double Bottom, double Right, double Top)>> PageRects { get; } = [];
+        public List<int> ResultPages { get; } = [];
+        public int TotalHits { get; set; }
+    }
+
+    internal sealed class SearchService
+    {
+        /// <summary>
+        /// Scans every page of <paramref name="filePath"/> for <paramref name="query"/> (case-insensitive).
+        /// Returns an empty result when query is blank or the file cannot be opened.
+        /// </summary>
+        public SearchResult Search(string filePath, string query)
+        {
+            var result = new SearchResult();
+            if (string.IsNullOrWhiteSpace(query) || string.IsNullOrWhiteSpace(filePath))
+                return result;
+
+            try
+            {
+                using var doc = PdfDocument.Open(filePath);
+                for (int pi = 0; pi < doc.NumberOfPages; pi++)
+                {
+                    var page = doc.GetPage(pi + 1);
+                    var hits = FindMatchesOnPage(page, query);
+                    if (hits.Count > 0)
+                    {
+                        result.PageRects[pi] = hits;
+                        result.ResultPages.Add(pi);
+                        result.TotalHits += hits.Count;
+                    }
+                }
+            }
+            catch { /* return whatever was collected so far */ }
+
+            return result;
+        }
+
+        internal static List<(double Left, double Bottom, double Right, double Top)> FindMatchesOnPage(
+            UglyToad.PdfPig.Content.Page page, string query)
+        {
+            var result = new List<(double, double, double, double)>();
+            var words = page.GetWords().ToList();
+
+            // A multi-word phrase can span adjacent words; a single token never does, so for single-token
+            // queries we only do the per-word substring match (the cross-word union box below would
+            // otherwise highlight whole runs of words leading up to the matching one).
+            bool isPhrase = query.Trim().IndexOf(' ') >= 0;
+
+            for (int i = 0; i < words.Count; i++)
+            {
+                if (words[i].Text.IndexOf(query, System.StringComparison.OrdinalIgnoreCase) >= 0)
+                {
+                    var bb = words[i].BoundingBox;
+                    result.Add((bb.Left, bb.Bottom, bb.Right, bb.Top));
+                    continue;
+                }
+
+                if (!isPhrase) continue;
+
+                // Multi-word match
+                string combined = words[i].Text;
+                for (int j = i + 1; j < words.Count && combined.Length < query.Length + 20; j++)
+                {
+                    combined += " " + words[j].Text;
+                    if (combined.IndexOf(query, System.StringComparison.OrdinalIgnoreCase) >= 0)
+                    {
+                        double minX = double.MaxValue, minY = double.MaxValue;
+                        double maxX = double.MinValue, maxY = double.MinValue;
+                        for (int k = i; k <= j; k++)
+                        {
+                            var wbb = words[k].BoundingBox;
+                            minX = Math.Min(minX, wbb.Left);
+                            minY = Math.Min(minY, wbb.Bottom);
+                            maxX = Math.Max(maxX, wbb.Right);
+                            maxY = Math.Max(maxY, wbb.Top);
+                        }
+                        result.Add((minX, minY, maxX, maxY));
+                        break;
+                    }
+                }
+            }
+            return result;
+        }
+    }
+}