Rework string table minimization (#56)

Author: josesimoes

source/MetadataProcessor.Core/Tables/nanoByteCodeTable.cs

@@ -32,7 +32,7 @@ public sealed class nanoByteCodeTable : InanoTable
         /// <summary>
         /// Temporary string table for code generators used during initial load.
         /// </summary>
-        private readonly nanoStringTable _fakeStringTable = new nanoStringTable();
+        private readonly nanoStringTable _fakeStringTable;
 
         /// <summary>
         /// Assembly tables context - contains all tables used for building target assembly.
@@ -54,6 +54,8 @@ public nanoByteCodeTable(
             nanoTablesContext context)
         {
             _context = context;
+
+            _fakeStringTable = new nanoStringTable(_context);
         }
 
         /// <summary>

source/MetadataProcessor.Core/Tables/nanoStringTable.cs

@@ -48,17 +48,30 @@ public IEnumerable<string> Sort(
         /// </summary>
         private ushort _lastAvailableId;
 
+        private ushort _lastPreAllocatedId;
+
+        /// <summary>
+        /// Assembly tables context - contains all tables used for building target assembly.
+        /// </summary>
+        private readonly nanoTablesContext _context;
+
+        /// <summary>
+        /// Last pre-allocated string identifier.
+        /// </summary>
+        public ushort LastPreAllocatedId { get => _lastPreAllocatedId; }
+
         /// <summary>
         /// Creates new instance of <see cref="nanoStringTable"/> object.
         /// </summary>
         public nanoStringTable(
+            nanoTablesContext context,
             ICustomStringSorter stringSorter = null)
         {
             GetOrCreateStringId(string.Empty); // First item in string table always empty string
             _stringSorter = stringSorter ?? new EmptyStringSorter();
-        }
 
-        
+            _context = context;
+        }
 
         /// <summary>
         /// Gets existing or creates new string reference identifier related to passed string value.
@@ -164,6 +177,46 @@ public void RemoveUnusedItems(HashSet<MetadataToken> items)
             // First item in string table always empty string
             GetOrCreateStringId(string.Empty);
 
+            // Pre-allocate strings from some tables
+            _context.AssemblyReferenceTable.AllocateStrings();
+            _context.TypeReferencesTable.AllocateStrings();
+
+
+            var memberReferences = _context.AssemblyDefinition.MainModule.GetMemberReferences();
+            List<MemberReference> memberReferencesInUse = new List<MemberReference>();
+
+            foreach (var item in memberReferences)
+            {
+                var memberRef = _context.TypeReferencesTable.Items.FirstOrDefault(m => m.FullName.Contains(item.DeclaringType.FullName));
+                if (memberRef != null)
+                {
+                    memberReferencesInUse.Add(item);
+                }
+            }
+
+            foreach (var item in memberReferencesInUse)
+            {
+                GetOrCreateStringId(item.Name);
+            }
+
+            foreach (var item in _context.TypeDefinitionTable.Items)
+            {
+                GetOrCreateStringId(item.Namespace);
+                GetOrCreateStringId(item.Name);
+
+                foreach (var f in item.Fields)
+                {
+                    GetOrCreateStringId(f.Name);
+                }
+
+                foreach (var m in item.Methods)
+                {
+                    GetOrCreateStringId(m.Name);
+                }
+            }
+
+            _lastPreAllocatedId = _idsByStrings.Last().Value;
+
             // fill in the dictionary with the used strings
             foreach (var s in usedStrings)
             {

source/MetadataProcessor.Core/Tables/nanoTablesContext.cs

@@ -159,7 +159,9 @@ public nanoTablesContext(
             // Strings and signatures
 
             SignaturesTable = new nanoSignaturesTable(this);
-            StringTable = new nanoStringTable(stringSorter);
+            StringTable = new nanoStringTable(
+                this,
+                stringSorter);
 
             // Byte code table
             ByteCodeTable = new nanoByteCodeTable(this);

source/MetadataProcessor.Core/nanoDumperGenerator.cs

@@ -85,7 +85,10 @@ private void DumpCustomAttributes(DumpAllTable dumpTable)
 
         private void DumpUserStrings(DumpAllTable dumpTable)
         {
-            foreach (var s in _tablesContext.StringTable.GetItems().OrderBy(i => i.Value))
+            // start at 1, because 0 is the empty string entry
+            int tokenId = 1;
+
+            foreach (var s in _tablesContext.StringTable.GetItems().OrderBy(i => i.Value).Where(i => i.Value > _tablesContext.StringTable.LastPreAllocatedId))
             {
                 // don't output the empty string
                 if(s.Value == 0)
@@ -94,7 +97,7 @@ private void DumpUserStrings(DumpAllTable dumpTable)
                 }
 
                 // fake the metadata token from the ID
-                var stringMetadataToken = new MetadataToken(TokenType.String, s.Value);
+                var stringMetadataToken = new MetadataToken(TokenType.String, tokenId++);
 
                 dumpTable.UserStrings.Add(
                     new UserString()

source/version.json

@@ -1,6 +1,6 @@
 {
   "$schema": "https://raw.githubusercontent.com/AArnott/Nerdbank.GitVersioning/master/src/NerdBank.GitVersioning/version.schema.json",
-  "version": "2.18.0",
+  "version": "2.19.0",
   "release": {
     "branchName" : "release-v{version}",
     "versionIncrement" : "minor",