c# - questions - understanding unicode




How do I get a consistent byte representation of strings in C# without manually specifying an encoding? (20)

How do I convert a string to a byte[] in .NET (C#) without manually specifying a specific encoding?

I'm going to encrypt the string. I can encrypt it without converting, but I'd still like to know why encoding comes to play here.

Also, why should encoding be taken into consideration? Can't I simply get what bytes the string has been stored in? Why is there a dependency on character encodings?


Also please explain why encoding should be taken into consideration. Can't I simply get what bytes the string has been stored in? Why this dependency on encoding?!!!

Because there is no such thing as "the bytes of the string".

A string (or more generically, a text) is composed of characters: letters, digits, and other symbols. That's all. Computers, however, do not know anything about characters; they can only handle bytes. Therefore, if you want to store or transmit text by using a computer, you need to transform the characters to bytes. How do you do that? Here's where encodings come to the scene.

An encoding is nothing but a convention to translate logical characters to physical bytes. The simplest and best known encoding is ASCII, and it is all you need if you write in English. For other languages you will need more complete encodings, being any of the Unicode flavours the safest choice nowadays.

So, in short, trying to "get the bytes of a string without using encodings" is as impossible as "writing a text without using any language".

By the way, I strongly recommend you (and anyone, for that matter) to read this small piece of wisdom: The Absolute Minimum Every Software Developer Absolutely, Positively Must Know About Unicode and Character Sets (No Excuses!)


How do I convert a string to a byte[] in .NET (C#) without manually specifying a specific encoding?

A string in .NET represents text as a sequence of UTF-16 code units, so the bytes are encoded in memory in UTF-16 already.

Mehrdad's Answer

You can use Mehrdad's answer, but it does actually use an encoding because chars are UTF-16. It calls ToCharArray which looking at the source creates a char[] and copies the memory to it directly. Then it copies the data to a byte array that is also allocated. So under the hood it is copying the underlying bytes twice and allocating a char array that is not used after the call.

Tom Blodget's Answer

Tom Blodget's answer is 20-30% faster than Mehrdad since it skips the intermediate step of allocating a char array and copying the bytes to it, but it requires you compile with the /unsafe option. If you absolutely do not want to use encoding, I think this is the way to go. If you put your encryption login inside the fixed block, you don't even need to allocate a separate byte array and copy the bytes to it.

Also, why should encoding be taken into consideration? Can't I simply get what bytes the string has been stored in? Why is there a dependency on character encodings?

Because that is the proper way to do it. string is an abstraction.

Using an encoding could give you trouble if you have 'strings' with invalid characters, but that shouldn't happen. If you are getting data into your string with invalid characters you are doing it wrong. You should probably be using a byte array or a Base64 encoding to start with.

If you use System.Text.Encoding.Unicode, your code will be more resilient. You don't have to worry about the endianness of the system your code will be running on. You don't need to worry if the next version of the CLR will use a different internal character encoding.

I think the question isn't why you want to worry about the encoding, but why you want to ignore it and use something else. Encoding is meant to represent the abstraction of a string in a sequence of bytes. System.Text.Encoding.Unicode will give you a little endian byte order encoding and will perform the same on every system, now and in the future.


Contrary to the answers here, you DON'T need to worry about encoding if the bytes don't need to be interpreted!

Like you mentioned, your goal is, simply, to "get what bytes the string has been stored in".
(And, of course, to be able to re-construct the string from the bytes.)

For those goals, I honestly do not understand why people keep telling you that you need the encodings. You certainly do NOT need to worry about encodings for this.

Just do this instead:

static byte[] GetBytes(string str)
{
    byte[] bytes = new byte[str.Length * sizeof(char)];
    System.Buffer.BlockCopy(str.ToCharArray(), 0, bytes, 0, bytes.Length);
    return bytes;
}

static string GetString(byte[] bytes)
{
    char[] chars = new char[bytes.Length / sizeof(char)];
    System.Buffer.BlockCopy(bytes, 0, chars, 0, bytes.Length);
    return new string(chars);
}

As long as your program (or other programs) don't try to interpret the bytes somehow, which you obviously didn't mention you intend to do, then there is nothing wrong with this approach! Worrying about encodings just makes your life more complicated for no real reason.

Additional benefit to this approach:

It doesn't matter if the string contains invalid characters, because you can still get the data and reconstruct the original string anyway!

It will be encoded and decoded just the same, because you are just looking at the bytes.

If you used a specific encoding, though, it would've given you trouble with encoding/decoding invalid characters.


Fastest way

public static byte[] GetBytes(string text)
{
    return System.Text.ASCIIEncoding.UTF8.GetBytes(text);
}

EDIT as Makotosan commented this is now the best way:

Encoding.UTF8.GetBytes(text)

C# to convert a string to a byte array:

public static byte[] StrToByteArray(string str)
{
   System.Text.UTF8Encoding  encoding=new System.Text.UTF8Encoding();
   return encoding.GetBytes(str);
}

Here is my unsafe implementation of String to Byte[] conversion:

public static unsafe Byte[] GetBytes(String s)
{
    Int32 length = s.Length * sizeof(Char);
    Byte[] bytes = new Byte[length];

    fixed (Char* pInput = s)
    fixed (Byte* pBytes = bytes)
    {
        Byte* source = (Byte*)pInput;
        Byte* destination = pBytes;

        if (length >= 16)
        {
            do
            {
                *((Int64*)destination) = *((Int64*)source);
                *((Int64*)(destination + 8)) = *((Int64*)(source + 8));

                source += 16;
                destination += 16;
            }
            while ((length -= 16) >= 16);
        }

        if (length > 0)
        {
            if ((length & 8) != 0)
            {
                *((Int64*)destination) = *((Int64*)source);

                source += 8;
                destination += 8;
            }

            if ((length & 4) != 0)
            {
                *((Int32*)destination) = *((Int32*)source);

                source += 4;
                destination += 4;
            }

            if ((length & 2) != 0)
            {
                *((Int16*)destination) = *((Int16*)source);

                source += 2;
                destination += 2;
            }

            if ((length & 1) != 0)
            {
                ++source;
                ++destination;

                destination[0] = source[0];
            }
        }
    }

    return bytes;
}

It's way faster than the accepted anwser's one, even if not as elegant as it is. Here are my Stopwatch benchmarks over 10000000 iterations:

[Second String: Length 20]
Buffer.BlockCopy: 746ms
Unsafe: 557ms

[Second String: Length 50]
Buffer.BlockCopy: 861ms
Unsafe: 753ms

[Third String: Length 100]
Buffer.BlockCopy: 1250ms
Unsafe: 1063ms

In order to use it, you have to tick "Allow Unsafe Code" in your project build properties. As per .NET Framework 3.5, this method can also be used as String extension:

public static unsafe class StringExtensions
{
    public static Byte[] ToByteArray(this String s)
    {
        // Method Code
    }
}

If you really want a copy of the underlying bytes of a string, you can use a function like the one that follows. However, you shouldn't please read on to find out why.

[DllImport(
        "msvcrt.dll",
        EntryPoint = "memcpy",
        CallingConvention = CallingConvention.Cdecl,
        SetLastError = false)]
private static extern unsafe void* UnsafeMemoryCopy(
    void* destination,
    void* source,
    uint count);

public static byte[] GetUnderlyingBytes(string source)
{
    var length = source.Length * sizeof(char);
    var result = new byte[length];
    unsafe
    {
        fixed (char* firstSourceChar = source)
        fixed (byte* firstDestination = result)
        {
            var firstSource = (byte*)firstSourceChar;
            UnsafeMemoryCopy(
                firstDestination,
                firstSource,
                (uint)length);
        }
    }

    return result;
}

This function will get you a copy of the bytes underlying your string, pretty quickly. You'll get those bytes in whatever way they are encoding on your system. This encoding is almost certainly UTF-16LE but that is an implementation detail you shouldn't have to care about.

It would be safer, simpler and more reliable to just call,

System.Text.Encoding.Unicode.GetBytes()

In all likelihood this will give the same result, is easier to type, and the bytes will always round-trip with a call to

System.Text.Encoding.Unicode.GetString()

It depends on the encoding of your string (ASCII, UTF-8, ...).

For example:

byte[] b1 = System.Text.Encoding.UTF8.GetBytes (myString);
byte[] b2 = System.Text.Encoding.ASCII.GetBytes (myString);

A small sample why encoding matters:

string pi = "\u03a0";
byte[] ascii = System.Text.Encoding.ASCII.GetBytes (pi);
byte[] utf8 = System.Text.Encoding.UTF8.GetBytes (pi);

Console.WriteLine (ascii.Length); //Will print 1
Console.WriteLine (utf8.Length); //Will print 2
Console.WriteLine (System.Text.Encoding.ASCII.GetString (ascii)); //Will print '?'

ASCII simply isn't equipped to deal with special characters.

Internally, the .NET framework uses UTF-16 to represent strings, so if you simply want to get the exact bytes that .NET uses, use System.Text.Encoding.Unicode.GetBytes (...).

See Character Encoding in the .NET Framework (MSDN) for more information.


Simply use this:

byte[] myByte= System.Text.ASCIIEncoding.Default.GetBytes(myString);

The accepted answer is very, very complicated. Use the included .NET classes for this:

const string data = "A string with international characters: Norwegian: ÆØÅæøå, Chinese: 喂 谢谢";
var bytes = System.Text.Encoding.UTF8.GetBytes(data);
var decoded = System.Text.Encoding.UTF8.GetString(bytes);

Don't reinvent the wheel if you don't have to...


The first part of your question (how to get the bytes) was already answered by others: look in the System.Text.Encoding namespace.

I will address your follow-up question: why do you need to pick an encoding? Why can't you get that from the string class itself?

The answer is in two parts.

First of all, the bytes used internally by the string class don't matter, and whenever you assume they do you're likely introducing a bug.

If your program is entirely within the .Net world then you don't need to worry about getting byte arrays for strings at all, even if you're sending data across a network. Instead, use .Net Serialization to worry about transmitting the data. You don't worry about the actual bytes any more: the Serialization formatter does it for you.

On the other hand, what if you are sending these bytes somewhere that you can't guarantee will pull in data from a .Net serialized stream? In this case you definitely do need to worry about encoding, because obviously this external system cares. So again, the internal bytes used by the string don't matter: you need to pick an encoding so you can be explicit about this encoding on the receiving end, even if it's the same encoding used internally by .Net.

I understand that in this case you might prefer to use the actual bytes stored by the string variable in memory where possible, with the idea that it might save some work creating your byte stream. However, I put it to you it's just not important compared to making sure that your output is understood at the other end, and to guarantee that you must be explicit with your encoding. Additionally, if you really want to match your internal bytes, you can already just choose the Unicode encoding, and get that performance savings.

Which brings me to the second part... picking the Unicode encoding is telling .Net to use the underlying bytes. You do need to pick this encoding, because when some new-fangled Unicode-Plus comes out the .Net runtime needs to be free to use this newer, better encoding model without breaking your program. But, for the moment (and forseeable future), just choosing the Unicode encoding gives you what you want.

It's also important to understand your string has to be re-written to wire, and that involves at least some translation of the bit-pattern even when you use a matching encoding. The computer needs to account for things like Big vs Little Endian, network byte order, packetization, session information, etc.


The key issue is that a glyph in a string takes 32 bits (16 bits for a character code) but a byte only has 8 bits to spare. A one-to-one mapping doesn't exist unless you restrict yourself to strings that only contain ASCII characters. System.Text.Encoding has lots of ways to map a string to byte[], you need to pick one that avoids loss of information and that is easy to use by your client when she needs to map the byte[] back to a string.

Utf8 is a popular encoding, it is compact and not lossy.


This is a popular question. It is important to understand what the question author is asking, and that it is different from what is likely the most common need. To discourage misuse of the code where it is not needed, I've answered the later first.

Common Need

Every string has a character set and encoding. When you convert a System.String object to an array of System.Byte you still have a character set and encoding. For most usages, you'd know which character set and encoding you need and .NET makes it simple to "copy with conversion." Just choose the appropriate Encoding class.

// using System.Text;
Encoding.UTF8.GetBytes(".NET String to byte array")

The conversion may need to handle cases where the target character set or encoding doesn't support a character that's in the source. You have some choices: exception, substitution or skipping. The default policy is to substitute a '?'.

// using System.Text;
var text = Encoding.ASCII.GetString(Encoding.ASCII.GetBytes("You win €100")); 
                                                      // -> "You win ?100"

Clearly, conversions are not necessarily lossless!

Note: For System.String the source character set is Unicode.

The only confusing thing is that .NET uses the name of a character set for the name of one particular encoding of that character set. Encoding.Unicode should be called Encoding.UTF16.

That's it for most usages. If that's what you need, stop reading here. See the fun Joel Spolsky article if you don't understand what an encoding is.

Specific Need

Now, the question author asks, "Every string is stored as an array of bytes, right? Why can't I simply have those bytes?"

He doesn't want any conversion.

From the C# spec:

Character and string processing in C# uses Unicode encoding. The char type represents a UTF-16 code unit, and the string type represents a sequence of UTF-16 code units.

So, we know that if we ask for the null conversion (i.e., from UTF-16 to UTF-16), we'll get the desired result:

Encoding.Unicode.GetBytes(".NET String to byte array")

But to avoid the mention of encodings, we must do it another way. If an intermediate data type is acceptable, there is a conceptual shortcut for this:

".NET String to byte array".ToCharArray()

That doesn't get us the desired datatype but Mehrdad's answer shows how to convert this Char array to a Byte array using BlockCopy. However, this copies the string twice! And, it too explicitly uses encoding-specific code: the datatype System.Char.

The only way to get to the actual bytes the String is stored in is to use a pointer. The fixed statement allows taking the address of values. From the C# spec:

[For] an expression of type string, ... the initializer computes the address of the first character in the string.

To do so, the compiler writes code skip over the other parts of the string object with RuntimeHelpers.OffsetToStringData. So, to get the raw bytes, just create a pointer to the string and copy the number of bytes needed.

// using System.Runtime.InteropServices
unsafe byte[] GetRawBytes(String s)
{
    if (s == null) return null;
    var codeunitCount = s.Length;
    /* We know that String is a sequence of UTF-16 codeunits 
       and such codeunits are 2 bytes */
    var byteCount = codeunitCount * 2; 
    var bytes = new byte[byteCount];
    fixed(void* pRaw = s)
    {
        Marshal.Copy((IntPtr)pRaw, bytes, 0, byteCount);
    }
    return bytes;
}

As @CodesInChaos pointed out, the result depends on the endianness of the machine. But the question author is not concerned with that.


Try this, a lot less code:

System.Text.Encoding.UTF8.GetBytes("TEST String");

Use:

    string text = "string";
    byte[] array = System.Text.Encoding.UTF8.GetBytes(text);

The result is:

[0] = 115
[1] = 116
[2] = 114
[3] = 105
[4] = 110
[5] = 103

Well, I've read all answers and they were about using encoding or one about serialization that drops unpaired surrogates.

It's bad when the string, for example, comes from SQL Server where it was built from a byte array storing, for example, a password hash. If we drop anything from it, it'll store an invalid hash, and if we want to store it in XML, we want to leave it intact (because the XML writer drops an exception on any unpaired surrogate it finds).

So I use Base64 encoding of byte arrays in such cases, but hey, on the Internet there is only one solution to this in C#, and it has bug in it and is only one way, so I've fixed the bug and written back procedure. Here you are, future googlers:

public static byte[] StringToBytes(string str)
{
    byte[] data = new byte[str.Length * 2];
    for (int i = 0; i < str.Length; ++i)
    {
        char ch = str[i];
        data[i * 2] = (byte)(ch & 0xFF);
        data[i * 2 + 1] = (byte)((ch & 0xFF00) >> 8);
    }

    return data;
}

public static string StringFromBytes(byte[] arr)
{
    char[] ch = new char[arr.Length / 2];
    for (int i = 0; i < ch.Length; ++i)
    {
        ch[i] = (char)((int)arr[i * 2] + (((int)arr[i * 2 + 1]) << 8));
    }
    return new String(ch);
}

You can use following code to convert a string to a byte array in .NET

string s_unicode = "abcéabc";
byte[] utf8Bytes = System.Text.Encoding.UTF8.GetBytes(s_unicode);

You can use the following code for conversion between string and byte array.

string s = "Hello World";

// String to Byte[]

byte[] byte1 = System.Text.Encoding.Default.GetBytes(s);

// OR

byte[] byte2 = System.Text.ASCIIEncoding.Default.GetBytes(s);

// Byte[] to string

string str = System.Text.Encoding.UTF8.GetString(byte1);

BinaryFormatter bf = new BinaryFormatter();
byte[] bytes;
MemoryStream ms = new MemoryStream();

string orig = "喂 Hello 谢谢 Thank You";
bf.Serialize(ms, orig);
ms.Seek(0, 0);
bytes = ms.ToArray();

MessageBox.Show("Original bytes Length: " + bytes.Length.ToString());

MessageBox.Show("Original string Length: " + orig.Length.ToString());

for (int i = 0; i < bytes.Length; ++i) bytes[i] ^= 168; // pseudo encrypt
for (int i = 0; i < bytes.Length; ++i) bytes[i] ^= 168; // pseudo decrypt

BinaryFormatter bfx = new BinaryFormatter();
MemoryStream msx = new MemoryStream();            
msx.Write(bytes, 0, bytes.Length);
msx.Seek(0, 0);
string sx = (string)bfx.Deserialize(msx);

MessageBox.Show("Still intact :" + sx);

MessageBox.Show("Deserialize string Length(still intact): " 
    + sx.Length.ToString());

BinaryFormatter bfy = new BinaryFormatter();
MemoryStream msy = new MemoryStream();
bfy.Serialize(msy, sx);
msy.Seek(0, 0);
byte[] bytesy = msy.ToArray();

MessageBox.Show("Deserialize bytes Length(still intact): " 
   + bytesy.Length.ToString());

byte[] strToByteArray(string str)
{
    System.Text.ASCIIEncoding enc = new System.Text.ASCIIEncoding();
    return enc.GetBytes(str);
}




character-encoding