Algorithm for compressing small files (345 Bytes) [closed] - algorithm

Question

Closed. This question does not meet Stack Overflow guidelines. It is not currently accepting answers.
We don’t allow questions seeking recommendations for books, tools, software libraries, and more. You can edit the question so it can be answered with facts and citations.
Closed 7 years ago.
Improve this question
I'm looking for a software that can losslessy compress a small amount of noisy data; example:
ZmNiYWNma3F5gYqSmqCkpqenpKGdmJONiIN/e3d1c3FxcXFxcnN0dXZ3eXp7fHx9fn5/gIGDhISFhYWFhIOCgYGAgICBgoOEhYeIiYmJiYiGhIF+fHl3dHNyc3N1d3p9gIOGiYuMjY2NjIuKiIeFg4GAfnx7eXl4eHl5enx9fn5/gICAgYGBgYGBgYGBgoKDhISEhISEhIOCgoGAgH9/fn5+fn5/gICAgICAgICAgICAgICAf39/gICBgoOFhoeIiIiIiIeGhYOCgH99fHt6enp6e3x9foCAgoKDg4ODgoKBgH59fHt7e3t8fX5/gIKDhIWFhoaGhoaFhISDgoKBgQ==
Original data: 345B (100%),
gzip: 280B (81%),
bzip2: 289B (84%),
lzop: 415B (120%),
Are there any other methods I should try?

Answer 1

Since the data is Base64 encoded (every 3 bytes become 4 bytes), the first step would be to decode it (the compressed data will be binary anyway):
344 bytes -> 256 bytes
Then, a simple test with standard winzip here shows a compression to 170 bytes (COMP_DEFLATE block). You should get about the same with gzip/zlib.
This could probably made somewhat smaller with a higher compression factor.
Compressing the original data gives 243 bytes (data block inside the .zip file, the full zip file is 359 bytes, but you don't need all that extra data).
So, using zlib on the decoded data, should compress that to ±170 bytes.
Looking at the decoded data, an even better compression would be possible. But that depends on the other data having the same structure.
Hex dump of the decoded data (a lot of values are repeating, or only changing slightly):
66 63 62 61 63 66 6B 71 79 81 8A 92 9A A0 A4 A6
A7 A7 A4 A1 9D 98 93 8D 88 83 7F 7B 77 75 73 71
71 71 71 71 72 73 74 75 76 77 79 7A 7B 7C 7C 7D
7E 7E 7F 80 81 83 84 84 85 85 85 85 84 83 82 81
81 80 80 80 81 82 83 84 85 87 88 89 89 89 89 88
86 84 81 7E 7C 79 77 74 73 72 73 73 75 77 7A 7D
80 83 86 89 8B 8C 8D 8D 8D 8C 8B 8A 88 87 85 83
81 80 7E 7C 7B 79 79 78 78 79 79 7A 7C 7D 7E 7E
7F 80 80 80 81 81 81 81 81 81 81 81 81 82 82 83
84 84 84 84 84 84 84 83 82 82 81 80 80 7F 7F 7E
7E 7E 7E 7E 7F 80 80 80 80 80 80 80 80 80 80 80
80 80 80 80 7F 7F 7F 80 80 81 82 83 85 86 87 88
88 88 88 88 87 86 85 83 82 80 7F 7D 7C 7B 7A 7A
7A 7A 7B 7C 7D 7E 80 80 82 82 83 83 83 83 82 82
81 80 7E 7D 7C 7B 7B 7B 7B 7C 7D 7E 7F 80 82 83
84 85 85 86 86 86 86 86 85 84 84 83 82 82 81 81
Only after a quick look: it should be possible to reach about 2 to 3 bits per byte on average, resulting in 64 to 96 bytes.
A closer look at the data:
Most values don't change that much. If all data is similar to this, a high compression rate could be achieved using some custom code. For example, the differences could be stored in 1, 2, 3 or 4 bits depending on the block of data (4 bits only needed for the first data points). Another approach, instead of full custom code, would be to compress the differences (delta values) with an existing algorithm (zlib, Huffman coding, and others).
Decimal values with 2 rounds of delta-encoding:
102
99 -3
98 -1 2
97 -1 0
99 2 3
102 3 1
107 5 2
113 6 1
121 8 2
129 8 0
138 9 1
146 8 -1
154 8 0
160 6 -2
164 4 -2
166 2 -2
167 1 -1
167 0 -1
164 -3 -3
161 -3 0
157 -4 -1
152 -5 -1
147 -5 0
141 -6 -1
136 -5 1
131 -5 0
127 -4 1
123 -4 0
119 -4 0
117 -2 2
115 -2 0
113 -2 0
113 0 2
113 0 0
113 0 0
113 0 0
114 1 1
115 1 0
116 1 0
117 1 0
118 1 0
119 1 0
121 2 1
122 1 -1
123 1 0
124 1 0
124 0 -1
125 1 1
126 1 0
126 0 -1
127 1 1
128 1 0
129 1 0
131 2 1
132 1 -1
132 0 -1
133 1 1
133 0 -1
133 0 0
133 0 0
132 -1 -1
131 -1 0
130 -1 0
129 -1 0
129 0 1
128 -1 -1
128 0 1
128 0 0
129 1 1
130 1 0
131 1 0
132 1 0
133 1 0
135 2 1
136 1 -1
137 1 0
137 0 -1
137 0 0
137 0 0
136 -1 -1
134 -2 -1
132 -2 0
129 -3 -1
126 -3 0
124 -2 1
121 -3 -1
119 -2 1
116 -3 -1
115 -1 2
114 -1 0
115 1 2
115 0 -1
117 2 2
119 2 0
122 3 1
125 3 0
128 3 0
131 3 0
134 3 0
137 3 0
139 2 -1
140 1 -1
141 1 0
141 0 -1
141 0 0
140 -1 -1
139 -1 0
138 -1 0
136 -2 -1
135 -1 1
133 -2 -1
131 -2 0
129 -2 0
128 -1 1
126 -2 -1
124 -2 0
123 -1 1
121 -2 -1
121 0 2
120 -1 -1
120 0 1
121 1 1
121 0 -1
122 1 1
124 2 1
125 1 -1
126 1 0
126 0 -1
127 1 1
128 1 0
128 0 -1
128 0 0
129 1 1
129 0 -1
129 0 0
129 0 0
129 0 0
129 0 0
129 0 0
129 0 0
129 0 0
130 1 1
130 0 -1
131 1 1
132 1 0
132 0 -1
132 0 0
132 0 0
132 0 0
132 0 0
132 0 0
131 -1 -1
130 -1 0
130 0 1
129 -1 -1
128 -1 0
128 0 1
127 -1 -1
127 0 1
126 -1 -1
126 0 1
126 0 0
126 0 0
126 0 0
127 1 1
128 1 0
128 0 -1
128 0 0
128 0 0
128 0 0
128 0 0
128 0 0
128 0 0
128 0 0
128 0 0
128 0 0
128 0 0
128 0 0
128 0 0
128 0 0
127 -1 -1
127 0 1
127 0 0
128 1 1
128 0 -1
129 1 1
130 1 0
131 1 0
133 2 1
134 1 -1
135 1 0
136 1 0
136 0 -1
136 0 0
136 0 0
136 0 0
135 -1 -1
134 -1 0
133 -1 0
131 -2 -1
130 -1 1
128 -2 -1
127 -1 1
125 -2 -1
124 -1 1
123 -1 0
122 -1 0
122 0 1
122 0 0
122 0 0
123 1 1
124 1 0
125 1 0
126 1 0
128 2 1
128 0 -2
130 2 2
130 0 -2
131 1 1
131 0 -1
131 0 0
131 0 0
130 -1 -1
130 0 1
129 -1 -1
128 -1 0
126 -2 -1
125 -1 1
124 -1 0
123 -1 0
123 0 1
123 0 0
123 0 0
124 1 1
125 1 0
126 1 0
127 1 0
128 1 0
130 2 1
131 1 -1
132 1 0
133 1 0
133 0 -1
134 1 1
134 0 -1
134 0 0
134 0 0
134 0 0
133 -1 -1
132 -1 0
132 0 1
131 -1 -1
130 -1 0
130 0 1
129 -1 -1
129 0 1

Answer 2

Here is a complete set of classes that can be used to compress this data with the following approach:
First Base64 decode the string to get bytes (345 -> 256 bytes)
Then delta-encode the bytes (meaning, subtract one byte from the previous, storing the results)
Another round of delta-encoding (but 3 was worse than 2)
Then compress the deltas using Huffman compression
This approach got down to 76 bytes including the necessary overhead to decompress later on.
A full Mercurial repository with the code can be found here.
Note! The code likely contains bugs around edge-cases such as empty or near-empty inputs. A suite of unit-tests can be found in the above linked repository but more testing is likely needed in order to trust this code for production usage.
Test program:
static void Main(string[] args)
{
string inputString = "ZmNiYWNma3F5gYqSmqCkpqenpKGdmJONiIN/e3d1c3FxcXFxcnN0dXZ3eXp7fHx9fn5/gIGDhISFhYWFhIOCgYGAgICBgoOEhYeIiYmJiYiGhIF+fHl3dHNyc3N1d3p9gIOGiYuMjY2NjIuKiIeFg4GAfnx7eXl4eHl5enx9fn5/gICAgYGBgYGBgYGBgoKDhISEhISEhIOCgoGAgH9/fn5+fn5/gICAgICAgICAgICAgICAf39/gICBgoOFhoeIiIiIiIeGhYOCgH99fHt6enp6e3x9foCAgoKDg4ODgoKBgH59fHt7e3t8fX5/gIKDhIWFhoaGhoaFhISDgoKBgQ==";
byte[] original = Convert.FromBase64String(inputString);
Console.WriteLine($"Original String: {inputString.Length}");
Console.WriteLine($"Original bytes: {original.Length}");
byte[] deltaEncoded = DeltaEncoderDecoder.Encode(original, 2);
byte[] compressed = HuffmanCompression.Compress(deltaEncoded);
Console.WriteLine($"Compressed bytes: {compressed.Length}");
byte[] deltaDecoded = HuffmanCompression.Decompress(compressed);
byte[] decompressed = DeltaEncoderDecoder.Decode(deltaDecoded, 2);
Console.WriteLine($"Decompressed bytes: {decompressed.Length}");
Console.WriteLine($"Decompressed == original: {decompressed.Length == original.Length && Enumerable.Range(0, original.Length).All(index => original[index] == decompressed[index])}");
}
Output:
Original String: 344
Original bytes: 256
Compressed bytes: 76
Decompressed bytes: 256
Decompressed == original: True
Here are the necessary classes to both compress and decompress the data:
public static class HuffmanCompression
{
internal const byte CompressedSignature = 255;
internal const byte UncompressedSignature = 0;
[NotNull]
public static byte[] Compress([NotNull] byte[] input)
{
if (input == null)
throw new ArgumentNullException(nameof(input));
if (input.Length == 0)
return input;
var rootNode = GetNodesFromRawInput(input);
var bitStrings = GetBitStringsFromTree(rootNode);
var output = new MemoryStream(input.Length);
var writer = new BitStreamWriter(output);
writer.Write(CompressedSignature);
writer.Write(input.Length);
WriteNodes(writer, rootNode);
WriteStrings(writer, bitStrings, input);
writer.Flush();
if (output.Length < input.Length + 1)
return output.ToArray();
return EncodeAsUncompressed(input);
}
[NotNull]
private static byte[] EncodeAsUncompressed([NotNull] byte[] input)
{
var output = new MemoryStream();
output.WriteByte(UncompressedSignature);
output.Write(input, 0, input.Length);
return output.ToArray();
}
private static void WriteStrings([NotNull] BitStreamWriter writer, [NotNull] string[] bitStrings, [NotNull] byte[] input)
{
foreach (byte value in input)
{
Assume(bitStrings[value] != null);
foreach (char bitChar in bitStrings[value])
writer.Write(bitChar == '1');
}
}
private static void WriteNodes([NotNull] BitStreamWriter writer, [NotNull] Node node)
{
if (node.Left == null)
{
writer.Write(false);
writer.Write(node.Value);
}
else
{
Assume(node.Right != null);
writer.Write(true);
WriteNodes(writer, node.Left);
WriteNodes(writer, node.Right);
}
}
[NotNull, ItemNotNull]
private static string[] GetBitStringsFromTree([NotNull] Node node)
{
var result = new string[256];
TraverseToGetBitStringsFromTree(node, string.Empty, result);
return result;
}
private static void TraverseToGetBitStringsFromTree([NotNull] Node node, [NotNull] string prefix, [NotNull, ItemNotNull] string[] dictionary)
{
if (node.Left != null)
{
Assume(node.Right != null);
TraverseToGetBitStringsFromTree(node.Left, prefix + "0", dictionary);
TraverseToGetBitStringsFromTree(node.Right, prefix + "1", dictionary);
}
else
dictionary[node.Value] = prefix;
}
[NotNull]
private static Node GetNodesFromRawInput([NotNull] byte[] input)
{
var occurances = new int[256];
foreach (byte value in input)
occurances[value]++;
var nodes = new List<Node>(256);
for (int index = 0; index < 256; index++)
if (occurances[index] > 0)
nodes.Add(new Node
{
Occurances = occurances[index],
Value = (byte)index
});
while (nodes.Count > 1)
{
nodes.Sort((n1, n2) =>
{
Assume(n1 != null && n2 != null);
return n1.Occurances.CompareTo(n2.Occurances);
});
Assume(nodes[0] != null && nodes[1] != null);
nodes[0] = new Node
{
Left = nodes[0],
Right = nodes[1],
Occurances = nodes[0].Occurances + nodes[1].Occurances
};
nodes.RemoveAt(1);
}
Assume(nodes[0] != null);
return nodes[0];
}
[NotNull]
public static byte[] Decompress([NotNull] byte[] input)
{
if (input == null)
throw new ArgumentNullException(nameof(input));
if (input.Length == 0)
return input;
if (input[0] != CompressedSignature)
return DecodeUncompressed(input);
var reader = new BitStreamReader(new MemoryStream(input));
reader.ReadByte(); // skip signature
int length = reader.ReadInt32();
var rootNode = ReadNodes(reader);
var output = new byte[length];
for (int index = 0; index < length; index++)
output[index] = DecompressOneByte(reader, rootNode);
return output;
}
private static byte DecompressOneByte([NotNull] BitStreamReader reader, [NotNull] Node node)
{
while (node.Left != null)
{
if (reader.ReadBit())
node = node.Right;
else
node = node.Left;
Assume(node != null);
}
return node.Value;
}
[NotNull]
private static Node ReadNodes([NotNull] BitStreamReader reader)
{
if (reader.ReadBit())
return new Node
{
Left = ReadNodes(reader),
Right = ReadNodes(reader)
};
return new Node
{
Value = reader.ReadByte()
};
}
[NotNull]
private static byte[] DecodeUncompressed([NotNull] byte[] input)
{
return input.Skip(1).ToArray();
}
}
public class BitStreamReader
{
[NotNull]
private readonly MemoryStream _Source;
private byte _Buffer;
private int _InBuffer;
public BitStreamReader([NotNull] MemoryStream source)
{
if (source == null)
throw new ArgumentNullException(nameof(source));
_Source = source;
}
public bool ReadBit()
{
if (_InBuffer == 0)
FillBuffer();
return (_Buffer & BitStreamConstants.BitMasks[8 - _InBuffer--]) != 0;
}
public byte ReadByte()
{
if (_InBuffer == 8)
{
_InBuffer = 0;
return _Buffer;
}
return (byte)((ReadBit() ? 128 : 0) | (ReadBit() ? 64 : 0) | (ReadBit() ? 32 : 0) | (ReadBit() ? 16 : 0) | (ReadBit() ? 8 : 0) | (ReadBit() ? 4 : 0) | (ReadBit() ? 2 : 0) | (ReadBit() ? 1 : 0));
}
public int ReadInt32()
{
int result = 0;
if (ReadBit())
result |= ReadByte();
if (ReadBit())
result |= ReadByte() << 8;
if (ReadBit())
result |= ReadByte() << 16;
if (ReadBit())
result |= ReadByte() << 24;
return result;
}
private void FillBuffer()
{
int value = _Source.ReadByte();
if (value < 0)
throw new InvalidOperationException("Read past end of source stream");
_Buffer = (byte)value;
_InBuffer = 8;
}
}
public class BitStreamWriter
{
[NotNull]
private readonly MemoryStream _Target;
private byte _Buffer;
private int _InBuffer;
public BitStreamWriter([NotNull] MemoryStream target)
{
if (target == null)
throw new ArgumentNullException(nameof(target));
_Target = target;
}
public void Flush()
{
if (_InBuffer == 0)
return;
_Target.WriteByte(_Buffer);
_Buffer = 0;
_InBuffer = 0;
}
public void Write(bool bit)
{
unchecked
{
if (bit)
_Buffer = (byte)(_Buffer | 1 << (7 - _InBuffer));
if (++_InBuffer == 8)
Flush();
}
}
public void Write(byte value)
{
for (int index = 0; index < 8; index++)
Write((value & BitStreamConstants.BitMasks[index]) != 0);
}
public void Write(int value)
{
byte b0 = (byte)(value & 0xff);
byte b1 = (byte)((value >> 8) & 0xff);
byte b2 = (byte)((value >> 16) & 0xff);
byte b3 = (byte)((value >> 24) & 0xff);
Write(b0 != 0);
if (b0 != 0)
Write(b0);
Write(b1 != 0);
if (b1 != 0)
Write(b1);
Write(b2 != 0);
if (b2 != 0)
Write(b2);
Write(b3 != 0);
if (b3 != 0)
Write(b3);
}
}
internal static class BitStreamConstants
{
[NotNull]
public static readonly byte[] BitMasks = { 128, 64, 32, 16, 8, 4, 2, 1 };
public const byte CompressedSignature = 255;
}
public static class DeltaEncoderDecoder
{
[NotNull]
public static byte[] Encode([NotNull] byte[] input, int iterations)
{
if (input == null)
throw new ArgumentNullException(nameof(input));
var output = new byte[input.Length];
Buffer.BlockCopy(input, 0, output, 0, input.Length);
while (iterations-- > 0)
{
byte previous = 0;
for (int index = 0; index < output.Length; index++)
{
byte current = output[index];
output[index] = (byte)(current - previous);
previous = current;
}
}
return output;
}
[NotNull]
public static byte[] Decode([NotNull] byte[] input, int iterations)
{
if (input == null)
throw new ArgumentNullException(nameof(input));
var output = new byte[input.Length];
Buffer.BlockCopy(input, 0, output, 0, input.Length);
while (iterations-- > 0)
{
byte previous = 0;
for (int index = 0; index < output.Length; index++)
{
output[index] = (byte)(previous + output[index]);
previous = output[index];
}
}
return output;
}
}