Cache Format

Cache Format

v

Format

itodag

an

Explanation

An example will be used here as it is easier to follow.

Let us say, the client wishes to fetch file type 2, file id 17.

First off, it will open the main_file_cache.idx2 file and seek to the index 17 * 6 (102). It will then read two tribytes.

fileSize = 1200
intialDataBlockId = 4

The client will now open the main_file_cache.dat file and seek to the index 4 * 520 (2080). The values it reads will be:

nextFileId = 17
currentFilePartId = 0
nextDataBlockId = 5
nextFileTypeId = 2
blockData = ...

It will read the first 512 bytes of the file and then knows that there is 688 bytes left. Therefore, it has to read the next block.

nextFileId = 17
currentFilePartId = 1
nextDataBlockId = 6
nextFileTypeId = 2
blockData ...

It reads these next 512 bytes of the file and now knows that there are 176 bytes left. So for a final time, it will read the next block.

nextFileId = 18
currentFilePartId = 2
nextDataBlockId = 7
nextFileTypeId = 2
blockData = ...

It can ignore most of these values (the next ones are meaningless at this stage) and read the final 176 bytes. The whole 1200 byte file has now been read.

Named Files

All the files in cache 0 have an archive-like format which allows named files (e.g. BADENC.TXT is a file which contains bad words in the wordenc archive).

Format

tribyte uncompressedsize
tribyte compressedsize

If the uncompressed and compressed sizes are equal, the whole file is not compressed but the individual entries are compressed using bzip2. If they are not equal, the entire file is compressed using bzip2 but the individual entries are not.

Also note, the magic id at the start of the bzip2 entries are not included in the cache. If you use an existing API to read the files and want to add this back, you must append the four characters: BZh1 before you decompress.

short fileCount

Each file entry has the format:

int nameHash
tribyte uncompressedSize
tribyte compressedSize

When you are looping through the files, you need to keep track of the file offset yourself. This psuedocode demonstrates how:

int offset = buffer.getCurrentOffset() + numFiles * 10;
for(int i = 0; i < numFiles; i++) {
   // read values
   int thisFileOffset = offset;
   offset += thisFileCompressedSize;
}

To get a named file by its name, you should first hash the name using this method:

public static int hash(String name) {
   int hash = 0;
   name = name.toUpperCase();
   for(int j = 0; j < name.length(); j++) {
       hash = (hash * 61 + name.charAt(j)) - 32;
   }
   return hash;
}

Then, loop through the file entries you loaded earlier to find a matching hash. Read the compressed file size from the offset. If the whole file is not compressed, you should decompress the individual entry.