This program is used to compress or decompress a file. The source code is build in an API style then you can easily write your own compression tool using the Huffman encoding.
ha compress (and decompress) a file using the Huffman coding technique. The input file is divided into regular piece named block and this block has a fixed size named block-size which is 10MB by default.
The compression ratio it-self is not very good because using only the Huffman code is not really efficient compare to other tool like bzip2 or xz. The minimal compressed size is eight of the original input size due to the nature of the Huffman code it-self but this only append if your file contain the same character.
The advantage of this program is its speed:
- It is multi-thread using the pthread API
- The Huffman code is easy to compute
ha input_filename -o output_filename
This is just a basic usage example but you can see all options by typing ha -h. With this command line you will compress the input_filename using a block-size of 10MB, a maximum memory usage of 100MB and n threads where n represent the number of cores on your computer.
The block-size, the memory maximum usage and the number of thread can be specified either directly with the program argument or by using the environment variable.
If you are using this tool and you don't have the possibility the set the program argument (e.g. if you use it with the tar program) you can set them indirectly by using the following environment variable:
- HA_THREAD: set the number of thread that need to be used
- HA_BLOCK_SIZE: set the block-size value
- HA_MEMLIMIT: set the maximum memory usage
This program is compatible with the tar archiver by using the following command line:
tar -I /path/to/ha -cf archive.tar.hc dir
This command line will archive and compress the directory dir using the ha compression program.
To decompress you can use this command line:
tar -I /path/to/ha -xvf archive.tar.hc
This table describe the format of the compressed file. The file is written in little endian order.
| Name | Size in bytes | Value | Description |
|---|---|---|---|
| Magic number | 4 | 0x52B5218B | Use to recognize the file |
| Version major | 1 | 1 | Current major version |
| Version minor | 1 | 0 | Current minor version |
| Version patch | 1 | 1 | Current patch version |
From this point all the fields are repeated until the end of the file (EOF).
| Name | Size in bytes | Value | Description |
|---|---|---|---|
| Block size | 4 | --block-size | The block size of each compressed block |
| CRC-32 | 4 | Depends data | This crc is computed using all the following field |
| Number of leaf | 1 | Depends data | Size of the Huffman tree |
The 2 following fields are repeated number of leaf times.
| Name | Size in bytes | Value | Description |
|---|---|---|---|
| Weight | 4 | Depends data | Number of character in the block-size |
| Byte value | 1 | 0 - 255 | Character corresponding to the weight |
| Name | Size in bytes | Value | Description |
|---|---|---|---|
| Padding | 1 | 0 - 7 | Number of padding bits at the end of the compressed memory |
| Memory size | 4 | 0 - Block size | Size of the following compressed memory |
| Data | memory size | Depends data | Compressed data |