Calculating CRC-32 in C# and .NET
Just a few days ago I found myself needing to calculate a CRC-32 in .NET. With so many facilities available I was a little shocked that there was nothing built-in to do it so knocked up something myself.
Because unsigned ints aren’t CLS compliant it won’t play well with VB.Net and implementing the HashAlgorithm might lead people to believe it’s suitable for signing – it isn’t. CRC-32′s are only any good for checksums along the lines of WinZIP, RAR etc. and certainly shouldn’t come near a password and instead consider SHA-512 or similar.
As well as using it as a HashAlgorithm with block processing you can also access the static method Compute although there is an overhead with every call building the table with that. If neither option suits you needs cut ‘n splice it to something that does.
Updated: Fix for repeated calls, cache look-up table for the default polynomial and additional static Compute methods with defaults.
using System;
using System.Security.Cryptography;
public class Crc32 : HashAlgorithm {
public const UInt32 DefaultPolynomial = 0xedb88320;
public const UInt32 DefaultSeed = 0xffffffff;
private UInt32 hash;
private UInt32 seed;
private UInt32[] table;
private static UInt32[] defaultTable;
public Crc32() {
table = InitializeTable(DefaultPolynomial);
seed = DefaultSeed;
Initialize();
}
public Crc32(UInt32 polynomial, UInt32 seed) {
table = InitializeTable(polynomial);
this.seed = seed;
Initialize();
}
public override void Initialize() {
hash = seed;
}
protected override void HashCore(byte[] buffer, int start, int length) {
hash = CalculateHash(table, hash, buffer, start, length);
}
protected override byte[] HashFinal() {
byte[] hashBuffer = UInt32ToBigEndianBytes(~hash);
this.HashValue = hashBuffer;
return hashBuffer;
}
public override int HashSize {
get { return 32; }
}
public static UInt32 Compute(byte[] buffer) {
return ~CalculateHash(InitializeTable(DefaultPolynomial), DefaultSeed, buffer, 0, buffer.Length);
}
public static UInt32 Compute(UInt32 seed, byte[] buffer) {
return ~CalculateHash(InitializeTable(DefaultPolynomial), seed, buffer, 0, buffer.Length);
}
public static UInt32 Compute(UInt32 polynomial, UInt32 seed, byte[] buffer) {
return ~CalculateHash(InitializeTable(polynomial), seed, buffer, 0, buffer.Length);
}
private static UInt32[] InitializeTable(UInt32 polynomial) {
if (polynomial == DefaultPolynomial && defaultTable != null)
return defaultTable;
UInt32[] createTable = new UInt32[256];
for (int i = 0; i < 256; i++) {
UInt32 entry = (UInt32)i;
for (int j = 0; j < 8; j++)
if ((entry & 1) == 1)
entry = (entry >> 1) ^ polynomial;
else
entry = entry >> 1;
createTable[i] = entry;
}
if (polynomial == DefaultPolynomial)
defaultTable = createTable;
return createTable;
}
private static UInt32 CalculateHash(UInt32[] table, UInt32 seed, byte[] buffer, int start, int size) {
UInt32 crc = seed;
for (int i = start; i < size; i++)
unchecked {
crc = (crc >> 8) ^ table[buffer[i] ^ crc & 0xff];
}
return crc;
}
private byte[] UInt32ToBigEndianBytes(UInt32 x) {
return new byte[] {
(byte)((x >> 24) & 0xff),
(byte)((x >> 16) & 0xff),
(byte)((x >> 8) & 0xff),
(byte)(x & 0xff)
};
}
}To compute the hash for a file simply:
Crc32 crc32 = new Crc32();
String hash = String.Empty;
using (FileStream fs = File.Open("c:\\myfile.txt", FileMode.Open))
foreach (byte b in crc32.ComputeHash(fs)) hash += b.ToString("x2").ToLower();
Console.WriteLine("CRC-32 is {0}", hash);[)amien
18 responses
]Dude, thanks a lot, i found a lot of dumb code out there and this is the best code written for this purpose.
I wrote one my self but it was a conversion from Java to c# and i did not use at least some support which c# provides and there was some other code which generated wrong checksomes on a 64-bit processor.
Although i have not tested your code on 64 bit processor but it seems it should work fine as far as my knowledge is concerned.
Nice code, but i found a bug if you compute values for streams larger than 4096 bytes (make textfile in sample code larger than 4096). In this case the HashCore function is called more than once but the complement is build in the CalculateHash function. So you don’t get the complement for the final value but for each intermediate value which will result in a wrong value. The complement build should take place in the HashFinal function.
You’re right, multiple calls to HashCore will carry on from the final complement instead of the actual current hash value.
Removing the ~ complement from the CalculateHash function means you can’t use that static method directly… I think complementing it in Initialize and inside HashCore might be a better approach.
Will update the code once I’ve had chance to test it and check the other hashing algorithms I have up.
Thanks!
[)amien
Did you fix the error in your code now? I would like to use it, with files longer than 4096 bytes.
Hi,
I tried to execute the program under Windows XP and the CRC-hash calculated for a file of 25MB wasn’t correct, just because of in windows XP the CalculateHash was called more than once(it works with a lenght of 4096 byte at a time). I toke off the ‘~’ from “return ~crc” of CalculateHash and put it on “byte[] hashBuffer = UInt32ToBigEndianBytes(~hash)” of HashFinal, so that step was made only once at the end of cycle (without modifying the intermediate values).
In this way the algorithm worked corrctly.
I hope I gived a hand,
Alessio
I think that returning it as a BigEndian Byte array is a mistake since 98% of .NET code is running on windows based systems (which use the Little-Endian method).
also, reading every byte from the result array, formatting it to “x2″ and adding it into a string is expensive and causing you to construct 5 strings (1 empty, and another one for each byte you append the formatted output).
I think that adding something like this would benefit everyone more.. (after returning little endian byte arrays) :)
PS – thx for the kickass implementation :)
PS #2 – it works fine on Vista ultimate x64
How to speed up calculation?
On 20GB files it takes a lot…
Do you think can be done something like spot calculation?
Say 1 byte every 100 bytes?
Camillo
You can’t really do that – the whole idea of a checksum is that if a single byte changes or is transferred incorrectly then the checksum fails – if you only checksum every 1 byte in 100 then it’s not going to catch 99% of the errors.
[)amien
[...] of our asset folder) and hash values. For my "hash" I'm using a CRC32, specifically Damien Guard's C# implementation. The nice thing about CRC32 is that we end up with pretty small values. The downside is that [...]
Thanks for the code, there’s a little bug in the CalculateHash function:
for(int i = start; i < size; i++)
needs to be:
for(int i = start; i < size+start; i++)
else you wont hash the whole buffer for non-zero values of start.
Cheers,
Phooey.
That is intentional – start is an index into the buffer to start populating at. If you change the loop as you describe it will go out of bounds with a non-zero value of start.
[)amien
If you interpret size as being the size of the array, yes, but the array already has a length, why would you need to pass it in again?
When passing the array, an offset and size, size refers to the size of the data, see the parameter definition for the HashCore function of the HashAlgorithm:
cbSize:
The number of bytes in the byte array to use as data
Cheers,
Phooey.
In your example you have:
using (FileStream fs = File.Open(“c:\\myfile.txt”, FileMode.Open))
foreach (byte b in crc32.Compute(fs)) hash += b.ToString(“x2″).ToLower();
However the only Compute mehtods I see are static and none of them take a FileStream object.
You’ll see it inherits from the HashAlgorithm class which has a ComputeHash method that takes a Stream object, reads 4K blocks at a time and calls through to HashCore for each block.
[)amien
Hi,
This is a nice and good piece of code. I tried it, and it works like that,just one question in my mind. Does this piece any have any copyright?
Code on my blog is released to the public domain except where specifically marked otherwise :)
So,this is meant for public domain ryt? Hope it means so, since I never saw any copyright attached to this piece. :)
Yes.