Add support for removing elements out of StringMap.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@34185 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/llvm/ADT/StringMap.h b/include/llvm/ADT/StringMap.h
index 7cc9ce7b6ce48ee9707c3052e5e5645f63b8ad30..243f9cd13fce7d68a98e68bd844ac0fbea74bebb 100644 (file)
--- a/include/llvm/ADT/StringMap.h
+++ b/include/llvm/ADT/StringMap.h
@@ -52,6 +52,7 @@ protected:
   ItemBucket *TheTable;
   unsigned NumBuckets;
   unsigned NumItems;
+  unsigned NumTombstones;
   unsigned ItemSize;
 protected:
   StringMapImpl(unsigned InitSize, unsigned ItemSize);
@@ -68,6 +69,14 @@ protected:
   /// in the map, return the bucket number of the key.  Otherwise return -1.
   /// This does not modify the map.
   int FindKey(const char *KeyStart, const char *KeyEnd) const;
+
+  /// RemoveKey - Remove the specified StringMapEntry from the table, but do not
+  /// delete it.  This aborts if the value isn't in the table.
+  void RemoveKey(StringMapEntryBase *V);
+
+  /// RemoveKey - Remove the StringMapEntry for the specified key from the
+  /// table, returning it.  If the key is not in the table, this returns null.
+  StringMapEntryBase *RemoveKey(const char *KeyStart, const char *KeyEnd);
   
 public:
   static StringMapEntryBase *getTombstoneVal() {
@@ -193,6 +202,28 @@ public:
     return const_iterator(TheTable+Bucket);
   }
   
+  /// insert - Insert the specified key/value pair into the map.  If the key
+  /// already exists in the map, return false and ignore the request, otherwise
+  /// insert it and return true.
+  bool insert(MapEntryTy *KeyValue) {
+    unsigned BucketNo =
+      LookupBucketFor(KeyValue->getKeyData(),
+                      KeyValue->getKeyData()+KeyValue->getKeyLength());
+    ItemBucket &Bucket = TheTable[BucketNo];
+    if (Bucket.Item && Bucket.Item != getTombstoneVal()) 
+      return false;  // Already exists in map.
+    
+    if (Bucket.Item == getTombstoneVal())
+      --NumTombstones;
+    Bucket.Item = KeyValue;
+    ++NumItems;
+    
+    // If the hash table is now more than 3/4 full, rehash into a larger table.
+    if (NumItems > NumBuckets*3/4)
+      RehashTable();
+    return true;
+  }
+  
   /// GetOrCreateValue - Look up the specified key in the table.  If a value
   /// exists, return it.  Otherwise, default construct a value, insert it, and
   /// return.
@@ -200,10 +231,13 @@ public:
                                             const char *KeyEnd) {
     unsigned BucketNo = LookupBucketFor(KeyStart, KeyEnd);
     ItemBucket &Bucket = TheTable[BucketNo];
-    if (Bucket.Item)
+    if (Bucket.Item && Bucket.Item != getTombstoneVal())
       return *static_cast<MapEntryTy*>(Bucket.Item);
     
     MapEntryTy *NewItem = MapEntryTy::Create(KeyStart, KeyEnd, Allocator);
+    
+    if (Bucket.Item == getTombstoneVal())
+      --NumTombstones;
     ++NumItems;
     
     // Fill in the bucket for the hash table.  The FullHashValue was already
@@ -216,6 +250,12 @@ public:
     return *NewItem;
   }
   
+  /// remove - Remove the specified key/value pair from the map, but do not
+  /// erase it.  This aborts if the key is not in the map.
+  void remove(MapEntryTy *KeyValue) {
+    RemoveKey(KeyValue);
+  }
+  
   ~StringMap() {
     for (ItemBucket *I = TheTable, *E = TheTable+NumBuckets; I != E; ++I) {
       if (MapEntryTy *Id = static_cast<MapEntryTy*>(I->Item))
@@ -228,6 +268,7 @@ public:
 
 template<typename ValueTy>
 class StringMapConstIterator {
+protected:
   StringMapImpl::ItemBucket *Ptr;
 public:
   StringMapConstIterator(StringMapImpl::ItemBucket *Bucket) : Ptr(Bucket) {

diff --git a/lib/Support/StringMap.cpp b/lib/Support/StringMap.cpp
index 02d42b9b49bb3b7d24aafbbd0ec0aa24e80d7a4c..95ff2bbe94a2ca0071a0c8d2af3fe4da202edaa5 100644 (file)
--- a/lib/Support/StringMap.cpp
+++ b/lib/Support/StringMap.cpp
@@ -21,6 +21,7 @@ StringMapImpl::StringMapImpl(unsigned InitSize, unsigned itemSize) {
   NumBuckets = InitSize ? InitSize : 512;
   ItemSize = itemSize;
   NumItems = 0;
+  NumTombstones = 0;
   
   TheTable = new ItemBucket[NumBuckets+1]();
   memset(TheTable, 0, NumBuckets*sizeof(ItemBucket));
@@ -57,20 +58,32 @@ unsigned StringMapImpl::LookupBucketFor(const char *NameStart,
   unsigned BucketNo = FullHashValue & (HTSize-1);
   
   unsigned ProbeAmt = 1;
+  int FirstTombstone = -1;
   while (1) {
     ItemBucket &Bucket = TheTable[BucketNo];
     StringMapEntryBase *BucketItem = Bucket.Item;
     // If we found an empty bucket, this key isn't in the table yet, return it.
     if (BucketItem == 0) {
+      // If we found a tombstone, we want to reuse the tombstone instead of an
+      // empty bucket.  This reduces probing.
+      if (FirstTombstone != -1) {
+        TheTable[FirstTombstone].FullHashValue = FullHashValue;
+        return FirstTombstone;
+      }
+      
       Bucket.FullHashValue = FullHashValue;
       return BucketNo;
     }
     
-    // If the full hash value matches, check deeply for a match.  The common
-    // case here is that we are only looking at the buckets (for item info
-    // being non-null and for the full hash value) not at the items.  This
-    // is important for cache locality.
-    if (Bucket.FullHashValue == FullHashValue) {
+    if (BucketItem == getTombstoneVal()) {
+      // Skip over tombstones.  However, remember the first one we see.
+      if (FirstTombstone == -1) FirstTombstone = BucketNo;
+    } else if (Bucket.FullHashValue == FullHashValue) {
+      // If the full hash value matches, check deeply for a match.  The common
+      // case here is that we are only looking at the buckets (for item info
+      // being non-null and for the full hash value) not at the items.  This
+      // is important for cache locality.
+      
       // Do the comparison like this because NameStart isn't necessarily
       // null-terminated!
       char *ItemStr = (char*)BucketItem+ItemSize;
@@ -108,11 +121,14 @@ int StringMapImpl::FindKey(const char *KeyStart, const char *KeyEnd) const {
     if (BucketItem == 0)
       return -1;
     
-    // If the full hash value matches, check deeply for a match.  The common
-    // case here is that we are only looking at the buckets (for item info
-    // being non-null and for the full hash value) not at the items.  This
-    // is important for cache locality.
-    if (Bucket.FullHashValue == FullHashValue) {
+    if (BucketItem == getTombstoneVal()) {
+      // Ignore tombstones.
+    } else if (Bucket.FullHashValue == FullHashValue) {
+      // If the full hash value matches, check deeply for a match.  The common
+      // case here is that we are only looking at the buckets (for item info
+      // being non-null and for the full hash value) not at the items.  This
+      // is important for cache locality.
+      
       // Do the comparison like this because NameStart isn't necessarily
       // null-terminated!
       char *ItemStr = (char*)BucketItem+ItemSize;
@@ -133,6 +149,30 @@ int StringMapImpl::FindKey(const char *KeyStart, const char *KeyEnd) const {
   }
 }
 
+/// RemoveKey - Remove the specified StringMapEntry from the table, but do not
+/// delete it.  This aborts if the value isn't in the table.
+void StringMapImpl::RemoveKey(StringMapEntryBase *V) {
+  const char *VStr = (char*)V + ItemSize;
+  StringMapEntryBase *V2 = RemoveKey(VStr, VStr+V->getKeyLength());
+  V2 = V2;
+  assert(V == V2 && "Didn't find key?");
+}
+
+/// RemoveKey - Remove the StringMapEntry for the specified key from the
+/// table, returning it.  If the key is not in the table, this returns null.
+StringMapEntryBase *StringMapImpl::RemoveKey(const char *KeyStart,
+                                             const char *KeyEnd) {
+  int Bucket = FindKey(KeyStart, KeyEnd);
+  if (Bucket == -1) return 0;
+  
+  StringMapEntryBase *Result = TheTable[Bucket].Item;
+  TheTable[Bucket].Item = getTombstoneVal();
+  --NumItems;
+  ++NumTombstones;
+  return Result;
+}
+
+
 
 /// RehashTable - Grow the table, redistributing values into the buckets with
 /// the appropriate mod-of-hashtable-size.
@@ -147,7 +187,7 @@ void StringMapImpl::RehashTable() {
   // Rehash all the items into their new buckets.  Luckily :) we already have
   // the hash values available, so we don't have to rehash any strings.
   for (ItemBucket *IB = TheTable, *E = TheTable+NumBuckets; IB != E; ++IB) {
-    if (IB->Item) {
+    if (IB->Item && IB->Item != getTombstoneVal()) {
       // Fast case, bucket available.
       unsigned FullHash = IB->FullHashValue;
       unsigned NewBucket = FullHash & (NewSize-1);
@@ -157,6 +197,7 @@ void StringMapImpl::RehashTable() {
         continue;
       }
       
+      // Otherwise probe for a spot.
       unsigned ProbeSize = 1;
       do {
         NewBucket = (NewBucket + ProbeSize++) & (NewSize-1);