HCL (Huffman Compression Library)
HCL a custom C++ library that allows file compression using Huffman Compression algorithm.
Table of Contents
Features
- Fast and efficient
- Supports all kinds of files (images, pdf, video, text, etc)
- High level abstraction methods to be used
- No need of understanding how compression algorithm works
- Light weight library
- Compression ratio of upto 50%
Benchmarks
Below is the output for huffmantool.benchmark()
method
Usage
There are 2 ways to try the HCL library
1. Try Online
Try the HCL library online on my custom repl.
2. Try Locally
-
Clone this repo to your project folder
git clone https://github.com/bhumijgupta/huffman-compression-library.git hcl
-
Import Huffman tool header and create new
huffmantool
object
// Filename main.cpp
#include "hcl/huffmantool.h"
int main(){
huffmantool ht;
.
.
.
}
- Compile and run file
g++ main.cpp && ./a.out
Methods
compressFile
compressFile(string sourceFile, string compressedFile = "") : string
Returns the destination of compressed file.
If the sourceFile
provided is invalid or cannot be opened, it returns an empty string.
If no compressedFile
is provided, compressedFile
is present at the same dir as sourcefile
with prefix compressed_
decompressFile
decompressFile(string compressedFile, string retrievedFile = "") : string
Returns the destination of decompressed file.
If the compressedFile
provided is invalid or cannot be opened, it reurns an empty string.
If no retrievedFile
is provided, retrievedFile
is present at the same dir as compressedFile
with prefix decompressed_
and removed compressed_
prefix (if present)
benchmark
benchmark(string sourcefile) : void
This method performs compression and decompression on the sourcefile provided and prints the following stats to the stdout.
--------------------------------------------------------------------------------
B E N C H M A R K
--------------------------------------------------------------------------------
Filetype Filename Filesize in bytes
Original
Compressed
Decompressed
--------------------------------------------------------------------------------
Time taken to compress file: XYZ microseconds
Time taken to decompress file: XYZ microseconds
Compression: XY.Z%
Note: For detailed understanding of code refer to Wiki.
Performance Comparison
The newfile.txt
is populated using lorem ipsum data from lipsum.com and the following data is noticed after executing main.cpp
.
Number of characters | Original file size (in bytes) | Compressed file size (in bytes) |
---|---|---|
10 | 10 | 36 |
100 | 100 | 116 |
1000 | 1000 | 642 |
10000 | 10000 | 5445 |
100000 | 100000 | 53354 |
Conclusion
The compression ratio and performance of the Huffman coding depends on the size of input text and the frequency of distinct characters in the file. From the above comparison we can see that the current implementation of huffman coding does not produce positive compression ratio for small files due to the overhead of storing Huffman Tree
in the compressed file, which is useful at the time of decompression.
But we notice, as the size of file increases (>=1000 bytes), the compression ratio stays at almost 50%.
Examples
Following are the list of examples
Run tests
The tests are stored in tests/
directory. You need to install googletest framework to run test.
cd tests/
cmake CMakeLists.txt
make
./runTests
License
This library is licensed under MIT License. The license file can be found here
Author
Bhumij Gupta
if(repo.isAwesome || repo.isHelpful){
StarRepo();
}