Improve prescription block autodetection.

zaaReloaded2/Medication/Importer.cs

@@ -36,40 +36,118 @@ namespace zaaReloaded2.Medication
         /// The detected block is selected.
         /// </summary>
         /// <returns>True if a block was detected, false if not.</returns>
+        /// <remarks>
+        /// <para>
+        /// Autodetection works by examining the document paragraph by
+        /// paragraph, starting at the end of the document. The first
+        /// block of at least two lines that are identified as prescription
+        /// lines is selected.
+        /// </para>
+        /// <para>
+        /// It should be noted that every paragraph (a.k.a. line) may
+        /// be regarded as one of three things:
+        /// </para>
+        /// <list type="ol">
+        /// <item>A typical prescription line (in the form "Ramipril 5 mg 1-0-0")</item>
+        /// <item>A typical non-prescription text line</item>
+        /// <item>Something inbetween, e.g. a line with tab stops as in
+        /// "Ramipril 5 mg \t alle zwei Tage" or in "Prof. B. Oss \t Dr. A. Sistent"
+        /// </item>
+        /// </list>
+        /// <para>
+        /// It is the third type of line that may cause confusion. If such a line
+        /// is encountered at the start of a putative block of prescriptions, we
+        /// therefore enter a "fuzzy" state in the detection algorithm and take
+        /// it from there, i.e. disregard the block if there are no lines that
+        /// are clearly prescriptions lines, or accept the block if we do detect
+        /// adjacent lines with unequivocal prescriptions.
+        /// </para>
+        /// 
+        /// </remarks>
         public static bool AutoDetect(Document document)
         {
             Paragraph start = null;
             Paragraph end = null;
+            bool insideBlock = false;
+            bool fuzzy = false;
+            bool result = false;
             int i = document.Paragraphs.Count;
 
             while (i > 1)
             {
                 string line = document.Paragraphs[i].Range.Text;
-                if (Prescription.IsPrescriptionLine(line))
+
+                if (Prescription.IsCanonicalPrescriptionLine(line))
+                {
+                    // The current line is unequivocally a prescription line:
+                    // If we're not inside a block already, mark the bottom
+                    // of the block.
+                    // If we are inside a block already, make sure to leave
+                    // the 'fuzzy' state because this clearly now is a prescription
+                    // block.
+                    if (insideBlock)
+                    {
+                        fuzzy = false;
+                    }
+                    else
                     {
                         end = document.Paragraphs[i];
+                        insideBlock = true;
+                    }
+                }
+                else if (Prescription.IsPotentialPrescriptionLine(line))
+                {
+                    // The current line is a putative prescription line:
+                    // If we're not inside a block already, enter the
+                    // "fuzzy" state.
+                    // If we are inside a block, no special action is
+                    // needed, we can continue with the next paragraph.
+                    if (!insideBlock)
+                    {
+                        fuzzy = true;
+                        insideBlock = true;
+                        end = document.Paragraphs[i];
+                    }
+                }
+                else
+                {
+                    // The current line is not a prescription line:
+                    // If we are currently in a definitive block of prescriptions,
+                    // mark the line below the current line as the start of the block.
+                    // If we're in a putative block, discard the information
+                    // about the bottom end of the block and reset all flags.
+                    if (insideBlock)
+                    {
+                        if (!fuzzy)
+	                    {
+                            start = document.Paragraphs[i + 1];
                             break;
 	                    }
+                        else
+                        {
+                            fuzzy = false;
+                            insideBlock = false;
+                            end = null;
+                        }
+                    }
+                }
+
                 i--;
             }
 
             if (end != null)
 	        {
-                start = end;
-                while (i > 2)
+                // If we don't have a start paragraph,
+                // but do have an end paragraph, we set the start paragraph to the
+                // first paragraph of the document.
+                if (start == null)
                 {
-                    if (!Prescription.IsPrescriptionLine(document.Paragraphs[i - 1].Range.Text))
-                    {
-                        start = document.Paragraphs[i];
-                        break;
+                    start = document.Paragraphs[1];
                 }
-                    i--;
-                }
-
                 document.Range(start.Range.Start, end.Range.End).Select();
-                return true;
+                result = true;
 	        }
-            return false;
+            return result;
         }
 
         #endregion
@@ -102,7 +180,7 @@ namespace zaaReloaded2.Medication
             string[] lines = Helpers.SplitParagraphs(text);
             foreach (string line in lines)
             {
-                if (Prescription.IsPrescriptionLine(line))
+                if (Prescription.IsCanonicalPrescriptionLine(line))
                 {
                     addition = Prescription.ManyFromLine(line);
                     columns = System.Math.Max(columns, addition.Count);

zaaReloaded2/Medication/Prescription.cs

@@ -35,11 +35,16 @@ namespace zaaReloaded2.Medication
         /// </summary>
         /// <param name="line">Line to inspect.</param>
         /// <returns>True if the line contains prescriptions.</returns>
-        public static bool IsPrescriptionLine(string line)
+        public static bool IsCanonicalPrescriptionLine(string line)
         {
             return canonicalRegex.IsMatch(line);
         }
 
+        public static bool IsPotentialPrescriptionLine(string line)
+        {
+            return alternativeRegex.IsMatch(line);
+        }
+
         #endregion
 
         #region Factory