Improvements in minimize stage when cleaning the string table (#52)

Author: josesimoes

source/MetadataProcessor.Core/Tables/nanoStringTable.cs

@@ -35,7 +35,7 @@ public IEnumerable<string> Sort(
         /// <summary>
         /// Maps for each unique string and related identifier (offset in strings table).
         /// </summary>
-        private readonly Dictionary<string, ushort> _idsByStrings =
+        private Dictionary<string, ushort> _idsByStrings =
             new Dictionary<string, ushort>(StringComparer.Ordinal);
 
         /// <summary>
@@ -138,6 +138,7 @@ internal void MergeValues(
         public void RemoveUnusedItems(HashSet<MetadataToken> items)
         {
             var setAll = new HashSet<MetadataToken>();
+            List<string> usedStrings = new List<string>();
 
             // build a collection of the existing tokens
             foreach (var a in _idsByStrings)
@@ -148,14 +149,25 @@ public void RemoveUnusedItems(HashSet<MetadataToken> items)
             // remove the ones that are used
             foreach (var t in items)
             {
-                setAll.Remove(t);
+                if(setAll.Remove(t))
+                {
+                    usedStrings.Add(TryGetString((ushort)t.ToUInt32()));
+                }
             }
 
-            // the ones left are the ones not used, OK to remove them
-            foreach (var t in setAll)
+            // reset dictionary
+            _idsByStrings = new Dictionary<string, ushort>(StringComparer.Ordinal);
+            
+            // and last ID too
+            _lastAvailableId = 0;
+
+            // First item in string table always empty string
+            GetOrCreateStringId(string.Empty);
+
+            // fill in the dictionary with the used strings
+            foreach (var s in usedStrings)
             {
-                var itemToRemove = TryGetString((ushort)t.ToUInt32());
-                _idsByStrings.Remove(itemToRemove);
+                GetOrCreateStringId(s);
             }
         }
 

source/MetadataProcessor.Core/Tables/nanoTablesContext.cs

@@ -420,10 +420,5 @@ internal void ResetSignaturesTable()
         {
             SignaturesTable = new nanoSignaturesTable(this);
         }
-
-        internal void ResetStringsTable()
-        {
-            StringTable = new nanoStringTable();
-        }
     }
 }

source/MetadataProcessor.Core/nanoAssemblyBuilder.cs

@@ -92,7 +92,7 @@ public void Write(
         /// Minimizes the assembly, removing unwanted and unused elements.
         /// </summary>
         /// <remarks>
-        /// In order to minize an assembly it has to have been previously compiled.
+        /// In order to minimize an assembly it has to have been previously compiled.
         /// </remarks>
         public void Minimize()
         {
@@ -156,7 +156,6 @@ public void Minimize()
             }
 
             // need to reset several tables so they are recreated only with the used items
-            _tablesContext.ResetStringsTable();
             _tablesContext.AssemblyReferenceTable.RemoveUnusedItems(set);
             _tablesContext.TypeReferencesTable.RemoveUnusedItems(set);
             _tablesContext.FieldsTable.RemoveUnusedItems(set);
@@ -168,7 +167,8 @@ public void Minimize()
             _tablesContext.AttributesTable.RemoveUnusedItems(set);
             _tablesContext.ResetByteCodeTable();
             _tablesContext.ResetSignaturesTable();
-             
+            _tablesContext.StringTable.RemoveUnusedItems(set);
+            
             // renormalise type definitions look-up tables
             // by removing items that are not used