struct File_Search_Filter_Params {
  bool search_directories = true;
  bool search_files       = true;
  bool exact_match        = false;
  bool order_most_recent  = false; // this should be an enum, but w/e :SortOrderComboBox
  bool match_word         = false;
  bool case_sensitive     = false;
  // #TODO: File data
  // Preprocess params
  // ArrayView<bool> include_drive; // defaults all to true.
  
};

struct File_Search {
  string query;
  File_Search_Filter_Params params;
  // Filesystem Data:
#if OS_WINDOWS 
  Win32_File_Enumeration* src;
  // #TODO: reserve space for result indices!
#endif
  Thread_Group thread_group;
};

Thread_Continue_Status file_search_thread_group_proc (Thread_Group* group, Thread* thread, void* work) {
  return Thread_Continue_Status::CONTINUE;
}

void prepare_file_search (File_Search* fs, string query) {
  // Note we will have to copy all the params out of File_Search into File_Search_Task/MT
  // Tokenizer t = get_tokenizer_from_string(query);
  // Tokenizer is the wrong move, we really just want to split up the query and analyze each bit separately.
  // 1. We want to split by spaces or other significant symbols like `|`
  Array<string> query_split;
  query_split.allocator = default_allocator();
  // 2. construct filters from splits
    // a. contains drive letters (can be independent, or include additional dirs, e.g. `C:\jai`
    // b. "NOT" operator `!`
}

void file_search_run_search_query (File_Search* fs, string query) {
  Assert(fs->src != nullptr); // check our data is actually there.
  if (fs->src == nullptr) return;
  
  prepare_file_search(fs, query);
  // 1. preprocess and configure filter parameters for search query:
  //   a. search for drive letters `C:` etc. and `C:\`
  // we need to tokenize the search query!
  // Tokenizer query_tokenizer = {query, 0};
  
  // Check if thread_group initted:
  if (!fs->thread_group.initialized) {
    // option1: os_cpu_physical_core_count, option2: os_cpu_logical_core_count
    push_allocator(default_allocator());
    bool enable_work_stealing = false;
    string thread_group_label = "File_Search";
    thread_group_init(&fs->thread_group, os_cpu_physical_core_count(), file_search_thread_group_proc, 
      thread_group_label, enable_work_stealing);
    // #TODO: split up work.
    // Assert(params->include_drive.count == fs->src->drive_data.count);
    // for_each(d, fs->src->drive_data) {
    //   if (!params->include_drive[d]) continue; // only include drives we're actually searching.
    //   // #current
    // }
    
    thread_group_start(&fs->thread_group);
  }
}

global Thread_Group bit_table_thread_group;
global s64 bit_table_thread_group_work_added = 0;
global s64 bit_table_thread_group_work_completed = 0;
global f64 bit_table_task_start_time = 0;
global f64 bit_table_task_end_time_main_thread = 0;
global s64 bit_table_task_result_count = 0;

void bit_table_task_reset_global_state () {
  bit_table_thread_group_work_added = 0;
  bit_table_thread_group_work_completed = 0;
  bit_table_task_start_time = 0;
  bit_table_task_end_time_main_thread = 0;
  bit_table_task_result_count = 0;
}

struct Bit_Table_Test_Work {
  Bit_Array bit_query;
  ArrayView<Bit_Array> bit_arrays;
  Array<u32> results;
  s32 index_offset;
  s32 sample_count;
  b32 directories; // either it's fst_dirs or fst_files.
  
  string query; // for debugging / printing
  // For memchr stuff: 
  Win32_File_Enumeration_Drive* drive;
  
};

void delete_task (Bit_Table_Test_Work* btt) {
  string_free(btt->query);
  bit_array_delete(&btt->bit_query);
  array_free(btt->results);
}

Thread_Continue_Status bit_table_testing_thread_group_proc (Thread_Group* group, Thread* thread, void* work) {
  Bit_Table_Test_Work* btt = thread_group_task(Bit_Table_Test_Work);
  // debug_break();
  auto_release_temp();
  // We need to use filter params here to check if this is correct.
  Assert(btt->query.count == 1);
  for (s64 b = 0; b < btt->sample_count; b += 1) {
    // For single ascii character: testing bit arrays
    string current_entry;
    if (btt->directories) {
      current_entry = directory_name_string_view(btt->drive, btt->index_offset + b);
    } else {
      current_entry = file_name_string_view(btt->drive, btt->index_offset + b);
    }
    // memchr (find characters in a buffer)
    if (is_ascii_letter(btt->query.data[0])) {
      u8 lower_case = to_lower_ascii(btt->query.data[0]);
      u8 upper_case = to_upper_ascii(btt->query.data[0]);
      if (memchr_avx2(current_entry.data, lower_case, (u16)current_entry.count) 
        || memchr_avx2(current_entry.data, upper_case, (u16)current_entry.count)) {
        array_add(btt->results, (u32)(btt->index_offset + b));
      }
    } else Assert(false); // not yet implemented for symbols!
    // For any X64 use memchr
    // if (memchr(current_entry.data, lower_case, current_entry.count) || memchr(current_entry.data, upper_case, current_entry.count)) {
    
    // for_each(b, btt->bit_arrays) {
    // if (contains_single_ascii(&btt->bit_arrays[b], &btt->bit_query)) {
    //   array_add(btt->results, (u32)b);
    // }
    // Testing memchr:
  }
  
  f64 elapsed_time = GetUnixTimestamp()-bit_table_task_start_time;
  push_allocator(temp());
  log("Done with query %s in %s", 
    btt->query.data, 
    format_time_seconds(elapsed_time).data);
  
  return Thread_Continue_Status::CONTINUE;
}

void bit_table_testing_init () {
  if (bit_table_thread_group.initialized) return;
  bool enable_work_stealing = true;
  thread_group_init(&bit_table_thread_group, os_cpu_logical_core_count(), bit_table_testing_thread_group_proc, 
    "bit_table_thread_group", enable_work_stealing);
        
  thread_group_start(&bit_table_thread_group);
}

void bit_table_testing_add_query (string query) {
  bit_table_testing_init();
  bit_table_task_start_time = GetUnixTimestamp();
  s32 thread_count = bit_table_thread_group.worker_info.count;
  // dirs...
  for_each(d, w32fe->drive_data) { // for each drive
    auto drive = &w32fe->drive_data[d];
    s64 dirs_per_thread = directory_count(drive) / thread_count;
    s64 start_offset = 0;
    auto src_array = to_view(drive->fst_dirs.bit_arrays);
    for (s32 i = 0; i < thread_count; i += 1) {
      auto btt = New<Bit_Table_Test_Work>();
      btt->drive = drive;
      btt->query = copy_string(query);
      bit_array_initialize(&btt->bit_query, 128);
      Bit_Array_Add_Value_Ascii_No_Case(&btt->bit_query, btt->query);
    
      if (i == thread_count - 1) {
        dirs_per_thread = (src_array.count - start_offset);
      }
      btt->bit_arrays = array_view(src_array, start_offset, dirs_per_thread);
      btt->index_offset = start_offset;
      btt->sample_count = dirs_per_thread;
      start_offset += dirs_per_thread;
      btt->results = Array<u32>(context_allocator(), btt->bit_arrays.count);
      btt->directories = true;
      
      add_work(&bit_table_thread_group, btt);
      bit_table_thread_group_work_added += 1;
    }
    Assert(start_offset == directory_count(drive));
  }
  // files...
  for_each(d, w32fe->drive_data) { // for each drive
    auto drive = &w32fe->drive_data[d];
    s64 files_per_thread = file_count(drive) / thread_count;
    s64 start_offset = 0;
    auto src_array = to_view(drive->fst_files.bit_arrays);
    for (s32 i = 0; i < thread_count; i += 1) {
      auto btt = New<Bit_Table_Test_Work>();
      btt->drive = drive;
      btt->query = copy_string(query);
      bit_array_initialize(&btt->bit_query, 128);
      Bit_Array_Add_Value_Ascii_No_Case(&btt->bit_query, btt->query);
    
      if (i == thread_count - 1) {
        files_per_thread = (src_array.count - start_offset);
      }
      btt->bit_arrays = array_view(src_array, start_offset, files_per_thread);
      btt->index_offset = start_offset;
      btt->sample_count = files_per_thread;
      start_offset += files_per_thread;
      btt->results = Array<u32>(context_allocator(), btt->bit_arrays.count);
      btt->directories = false;
      
      add_work(&bit_table_thread_group, btt);
      bit_table_thread_group_work_added += 1;
    }
    Assert(start_offset == file_count(drive));
  }
}