AK: Make hashing of UTF-16 strings cheaper

No need to iterate every byte of the string, we can iterate the code
units instead.

We must also actually record that we have cached the hash :^)

Author: trflynn89

AK/StringBase.h

@@ -219,7 +219,7 @@ inline u32 StringBase::hash() const
         return string_hash(reinterpret_cast<char const*>(bytes.data()), bytes.size());
     }
     if (!m_impl.data)
-        return string_hash(nullptr, 0);
+        return string_hash<char>(nullptr, 0);
     return data_without_union_member_assertion()->hash();
 }
 

AK/StringHash.h

@@ -6,6 +6,7 @@
 
 #pragma once
 
+#include <AK/Concepts.h>
 #include <AK/Types.h>
 
 namespace AK {
@@ -14,7 +15,8 @@ namespace AK {
 //        We can't use SipHash since that depends on runtime parameters,
 //        but some string hashes like IPC endpoint magic numbers need to be deterministic.
 //        Maybe use a SipHash with a statically-known key?
-constexpr u32 string_hash(char const* characters, size_t length, u32 seed = 0)
+template<OneOf<char, char16_t> T>
+constexpr u32 string_hash(T const* characters, size_t length, u32 seed = 0)
 {
     u32 hash = seed;
     for (size_t i = 0; i < length; ++i) {

AK/Utf16StringBase.h

@@ -176,7 +176,7 @@ class Utf16StringBase {
 
         if (auto const* data = data_without_union_member_assertion())
             return data->hash();
-        return string_hash(nullptr, 0);
+        return string_hash<char16_t>(nullptr, 0);
     }
 
     [[nodiscard]] ALWAYS_INLINE bool is_empty() const { return length_in_code_units() == 0uz; }

AK/Utf16StringData.h

@@ -78,8 +78,11 @@ class Utf16StringData final : public RefCounted<Utf16StringData> {
 
     ALWAYS_INLINE u32 hash() const
     {
-        if (!m_has_hash)
-            m_hash = calculate_hash();
+        if (!m_has_hash) {
+            m_hash = utf16_view().hash();
+            m_has_hash = true;
+        }
+
         return m_hash;
     }
 
@@ -128,13 +131,6 @@ class Utf16StringData final : public RefCounted<Utf16StringData> {
 
     [[nodiscard]] size_t calculate_code_point_length() const;
 
-    [[nodiscard]] ALWAYS_INLINE u32 calculate_hash() const
-    {
-        if (has_ascii_storage())
-            return ascii_view().hash();
-        return utf16_view().hash();
-    }
-
     // We store whether this string has ASCII or UTF-16 storage by setting the most significant bit of m_length_in_code_units
     // to 1 for UTF-16 storage. This shrinks the size of most UTF-16 string related classes, at the cost of not being
     // allowed to create a string larger than 2**63 - 1.

AK/Utf16View.h

@@ -324,7 +324,7 @@ class Utf16View {
             return 0;
         if (has_ascii_storage())
             return string_hash(m_string.ascii, length_in_code_units());
-        return string_hash(reinterpret_cast<char const*>(m_string.utf16), length_in_code_units() * sizeof(char16_t));
+        return string_hash(m_string.utf16, length_in_code_units());
     }
 
     [[nodiscard]] constexpr bool is_null() const