From cbbfc2f11ad49bfa849cd2ec3585bbb9699953ba Mon Sep 17 00:00:00 2001
From: Stenzek <stenzek@gmail.com>
Date: Tue, 13 Aug 2024 14:47:11 +1000
Subject: [PATCH] StringUtil: Add BytePatternSearch()

---
 src/common/string_util.cpp | 98 +++++++++++++++++++++++++++++++++++++-
 src/common/string_util.h   |  6 ++-
 2 files changed, 102 insertions(+), 2 deletions(-)

diff --git a/src/common/string_util.cpp b/src/common/string_util.cpp
index 5a92fb5293..7b84400535 100644
--- a/src/common/string_util.cpp
+++ b/src/common/string_util.cpp
@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
+// SPDX-FileCopyrightText: 2019-2024 Connor McLaughlin <stenzek@gmail.com>
 // SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
 
 #include "string_util.h"
@@ -9,6 +9,12 @@
 #include <cstdio>
 #include <sstream>
 
+#ifndef __APPLE__
+#include <malloc.h> // alloca
+#else
+#include <alloca.h>
+#endif
+
 #ifdef _WIN32
 #include "windows_headers.h"
 #endif
@@ -427,6 +433,96 @@ void StringUtil::EllipsiseInPlace(std::string& str, u32 max_length, const char*
   }
 }
 
+std::optional<size_t> StringUtil::BytePatternSearch(const std::span<const u8> bytes, const std::string_view pattern)
+{
+  // Parse the pattern into a bytemask.
+  size_t pattern_length = 0;
+  bool hinibble = true;
+  for (size_t i = 0; i < pattern.size(); i++)
+  {
+    if ((pattern[i] >= '0' && pattern[i] <= '9') || (pattern[i] >= 'a' && pattern[i] <= 'f') ||
+        (pattern[i] >= 'A' && pattern[i] <= 'F') || pattern[i] == '?')
+    {
+      hinibble ^= true;
+      if (hinibble)
+        pattern_length++;
+    }
+    else if (pattern[i] == ' ' || pattern[i] == '\r' || pattern[i] == '\n')
+    {
+      continue;
+    }
+    else
+    {
+      break;
+    }
+  }
+  if (pattern_length == 0)
+    return std::nullopt;
+
+  const bool allocate_on_heap = (pattern_length >= 512);
+  u8* match_bytes = allocate_on_heap ? static_cast<u8*>(alloca(pattern_length * 2)) : new u8[pattern_length * 2];
+  u8* match_masks = match_bytes + pattern_length;
+
+  hinibble = true;
+  u8 match_byte = 0;
+  u8 match_mask = 0;
+  for (size_t i = 0, match_len = 0; i < pattern.size(); i++)
+  {
+    u8 nibble = 0, nibble_mask = 0xF;
+    if (pattern[i] >= '0' && pattern[i] <= '9')
+      nibble = pattern[i] - '0';
+    else if (pattern[i] >= 'a' && pattern[i] <= 'f')
+      nibble = pattern[i] - 'a' + 0xa;
+    else if (pattern[i] >= 'A' && pattern[i] <= 'F')
+      nibble = pattern[i] - 'A' + 0xa;
+    else if (pattern[i] == '?')
+      nibble_mask = 0;
+    else if (pattern[i] == ' ' || pattern[i] == '\r' || pattern[i] == '\n')
+      continue;
+    else
+      break;
+
+    hinibble ^= true;
+    if (hinibble)
+    {
+      match_bytes[match_len] = nibble | (match_byte << 4);
+      match_masks[match_len] = nibble_mask | (match_mask << 4);
+      match_len++;
+    }
+    else
+    {
+      match_byte = nibble;
+      match_mask = nibble_mask;
+    }
+  }
+  if (pattern_length == 0)
+    return std::nullopt;
+
+  std::optional<size_t> ret;
+  const size_t max_search_offset = bytes.size() - pattern_length;
+  for (size_t offset = 0; offset < max_search_offset; offset++)
+  {
+    const u8* start = bytes.data() + offset;
+    for (size_t match_offset = 0;;)
+    {
+      if ((start[match_offset] & match_masks[match_offset]) != match_bytes[match_offset])
+        break;
+
+      match_offset++;
+      if (match_offset == pattern_length)
+      {
+        // found it!
+        ret = offset;
+      }
+    }
+  }
+
+  if (allocate_on_heap)
+    delete[] match_bytes;
+
+  return ret;
+}
+
 size_t StringUtil::DecodeUTF8(const std::string_view str, size_t offset, char32_t* ch)
 {
   return DecodeUTF8(str.data() + offset, str.length() - offset, ch);
diff --git a/src/common/string_util.h b/src/common/string_util.h
index 9e80fec4ad..388bd70df3 100644
--- a/src/common/string_util.h
+++ b/src/common/string_util.h
@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
+// SPDX-FileCopyrightText: 2019-2024 Connor McLaughlin <stenzek@gmail.com>
 // SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
 
 #pragma once
@@ -8,6 +8,7 @@
 #include <cstring>
 #include <iomanip>
 #include <optional>
+#include <span>
 #include <string>
 #include <string_view>
 #include <vector>
@@ -275,6 +276,9 @@ size_t DecodeUTF8(const std::string& str, size_t offset, char32_t* ch);
 std::string Ellipsise(const std::string_view str, u32 max_length, const char* ellipsis = "...");
 void EllipsiseInPlace(std::string& str, u32 max_length, const char* ellipsis = "...");
 
+/// Searches for the specified byte pattern in the given memory span. Wildcards (i.e. ??) are supported.
+std::optional<size_t> BytePatternSearch(const std::span<const u8> bytes, const std::string_view pattern);
+
 /// Strided memcpy/memcmp.
 ALWAYS_INLINE static void StrideMemCpy(void* dst, std::size_t dst_stride, const void* src, std::size_t src_stride,
                                        std::size_t copy_size, std::size_t count)