// So NTFS (and most systems) sort the tree by default in lexicographical descending order. // For lookups, this often isn't that useful if you only know substrings of the path. // struct BTNode { // u16 key_count; // BTNode* keys; // u16 allocated; // }; // struct B_Tree { // BTNode* root; // Allocator allocator; // }; // A compact collection of data with sorting indices // Maybe we can make B+ trees for sorting according to // size and modtime. // It really doesn't make sense to store data in memory as a B-tree except // if we need ordered insertions and deletes. // // Returns offset force_inline u32 AddString_NoCount (Serializer* serializer, u8* data, u8 count) { // #TODO: , bool null_terminate=false u8* current_point = &serializer->data[serializer->count]; s64 final_count = serializer->allocated + (count * sizeof(u8)); if (serializer->allocated < final_count) { array_reserve(*serializer, final_count); } memcpy(current_point, data, count * sizeof(u8)); serializer->count += count * sizeof(u8); return (u32)serializer->count; } constexpr s64 DFS_Preallocation_Count = 4194304; // 2^22 // template struct DFS_Array { Serializer* wstrings; // Serializer* strings; ArenaArray* offsets; // offsets into strings->data ArenaArray* lengths; // this type may vary Not sure if I should make it a template argument. Seems yucky. ArenaArray* modtimes; ArenaArray* sizes; ArenaArray* parent_indices; // #Temporary arrays for linking files/dirs to their parent directory, if present. ArenaArray* record_ids; ArenaArray* parent_ids; }; s64 item_count (DFS_Array* dfsa) { return dfsa->offsets->count; } void initialize (DFS_Array* dfsa) { Assert(dfsa != nullptr); dfsa->wstrings = new_serializer(Arena_Reserve::Size_2G); // dfsa->strings = new_serializer(Arena_Reserve::Size_2G); dfsa->offsets = arena_array_new(DFS_Preallocation_Count, Arena_Reserve::Size_2G); dfsa->lengths = arena_array_new (DFS_Preallocation_Count, Arena_Reserve::Size_2G); dfsa->modtimes = arena_array_new(DFS_Preallocation_Count, Arena_Reserve::Size_2G); dfsa->sizes = arena_array_new(DFS_Preallocation_Count, Arena_Reserve::Size_2G); dfsa->record_ids = arena_array_new(DFS_Preallocation_Count, Arena_Reserve::Size_2G); dfsa->parent_ids = arena_array_new(DFS_Preallocation_Count, Arena_Reserve::Size_2G); dfsa->parent_indices = arena_array_new(DFS_Preallocation_Count, Arena_Reserve::Size_2G); } void release (DFS_Array* dfsa) { free_serializer(dfsa->wstrings); if (is_valid(dfsa->offsets)) { arena_array_free(*dfsa->offsets); } if (is_valid(dfsa->lengths)) { arena_array_free(*dfsa->lengths); } if (is_valid(dfsa->modtimes)) { arena_array_free(*dfsa->modtimes); } if (is_valid(dfsa->sizes)) { arena_array_free(*dfsa->sizes); } if (is_valid(dfsa->record_ids)) { arena_array_free(*dfsa->record_ids); } if (is_valid(dfsa->parent_ids)) { arena_array_free(*dfsa->parent_ids); } if (is_valid(dfsa->parent_indices)) { arena_array_free(*dfsa->parent_indices); } zero_struct(dfsa); } struct Dense_FS { // Link to OS_Drive DFS_Array paths; DFS_Array files; ArenaTable path_table; // . OS_Drive* drive; // backlink for reference. }; force_inline void release (Dense_FS* dfs) { Timed_Block_Print("release: Dense_FS*"); dfs->drive = nullptr; // just a link release(&dfs->paths); release(&dfs->files); if (table_is_valid(&dfs->path_table)) { table_release(&dfs->path_table); } } void initialize (Dense_FS* dfs, OS_Drive* drive) { Assert(drive != nullptr); Assert(dfs != nullptr); // Is there a less stupid way of doing this? dfs->drive = drive; drive->data = dfs; initialize(&dfs->paths); initialize(&dfs->files); table_init(&dfs->path_table, 1048576); // 2^20 // dfs->path_table.hash_function = table_hash_function_fnv1a; // default. dfs->path_table.hash_function = sdbm_hash; dfs->path_table.compare_function = u32_keys_match; } s32 find_previous_index (Dense_FS* dfs, u32 record_id, bool* success) { s32 result = -1; (*success) = table_find(&dfs->path_table, record_id, &result); return result; } void cleanup_after_enumeration(Dense_FS* dfs) { table_release(&dfs->path_table); reset_struct(&dfs->path_table); arena_array_free(*dfs->paths.record_ids); arena_array_free(*dfs->paths.parent_ids); arena_array_free(*dfs->files.record_ids); arena_array_free(*dfs->files.parent_ids); // Unfortunately, we have to do this whenever we free, because we're using raw pointers. dfs->paths.record_ids = nullptr; dfs->paths.parent_ids = nullptr; dfs->files.record_ids = nullptr; dfs->files.parent_ids = nullptr; }