Determine a string's encoding in C#

Question

The code below has the following features:

1. Detection or attempted detection of UTF-7, UTF-8/16/32 (bom, no bom, little & big endian)
2. Falls back to the local default codepage if no Unicode encoding was found.
3. Detects (with high probability) unicode files with the BOM/signature missing
4. Searches for charset=xyz and encoding=xyz inside file to help determine encoding.
5. To save processing, you can 'taste' the file (definable number of bytes).
6. The encoding and decoded text file is returned.
7. Purely byte-based solution for efficiency

As others have said, no solution can be perfect (and certainly one can't easily differentiate between the various 8-bit extended ASCII encodings in use worldwide), but we can get 'good enough' especially if the developer also presents to the user a list of alternative encodings as shown here: What is the most common encoding of each language?

A full list of Encodings can be found using Encoding.GetEncodings();

// Function to detect the encoding for UTF-7, UTF-8/16/32 (bom, no bom, little
// & big endian), and local default codepage, and potentially other codepages.
// 'taster' = number of bytes to check of the file (to save processing). Higher
// value is slower, but more reliable (especially UTF-8 with special characters
// later on may appear to be ASCII initially). If taster = 0, then taster
// becomes the length of the file (for maximum reliability). 'text' is simply
// the string with the discovered encoding applied to the file.
public Encoding detectTextEncoding(string filename, out String text, int taster = 1000)
{
    byte[] b = File.ReadAllBytes(filename);

    //////////////// First check the low hanging fruit by checking if a
    //////////////// BOM/signature exists (sourced from http://www.unicode.org/faq/utf_bom.html#bom4)
    if (b.Length >= 4 && b[0] == 0x00 && b[1] == 0x00 && b[2] == 0xFE && b[3] == 0xFF) { text = Encoding.GetEncoding("utf-32BE").GetString(b, 4, b.Length - 4); return Encoding.GetEncoding("utf-32BE"); }  // UTF-32, big-endian 
    else if (b.Length >= 4 && b[0] == 0xFF && b[1] == 0xFE && b[2] == 0x00 && b[3] == 0x00) { text = Encoding.UTF32.GetString(b, 4, b.Length - 4); return Encoding.UTF32; }    // UTF-32, little-endian
    else if (b.Length >= 2 && b[0] == 0xFE && b[1] == 0xFF) { text = Encoding.BigEndianUnicode.GetString(b, 2, b.Length - 2); return Encoding.BigEndianUnicode; }     // UTF-16, big-endian
    else if (b.Length >= 2 && b[0] == 0xFF && b[1] == 0xFE) { text = Encoding.Unicode.GetString(b, 2, b.Length - 2); return Encoding.Unicode; }              // UTF-16, little-endian
    else if (b.Length >= 3 && b[0] == 0xEF && b[1] == 0xBB && b[2] == 0xBF) { text = Encoding.UTF8.GetString(b, 3, b.Length - 3); return Encoding.UTF8; } // UTF-8
    else if (b.Length >= 3 && b[0] == 0x2b && b[1] == 0x2f && b[2] == 0x76) { text = Encoding.UTF7.GetString(b,3,b.Length-3); return Encoding.UTF7; } // UTF-7

        
    //////////// If the code reaches here, no BOM/signature was found, so now
    //////////// we need to 'taste' the file to see if can manually discover
    //////////// the encoding. A high taster value is desired for UTF-8
    if (taster == 0 || taster > b.Length) taster = b.Length;    // Taster size can't be bigger than the filesize obviously.


    // Some text files are encoded in UTF8, but have no BOM/signature. Hence
    // the below manually checks for a UTF8 pattern. This code is based off
    // the top answer at: https://stackoverflow.com/questions/6555015/check-for-invalid-utf8
    // For our purposes, an unnecessarily strict (and terser/slower)
    // implementation is shown at: https://stackoverflow.com/questions/1031645/how-to-detect-utf-8-in-plain-c
    // For the below, false positives should be exceedingly rare (and would
    // be either slightly malformed UTF-8 (which would suit our purposes
    // anyway) or 8-bit extended ASCII/UTF-16/32 at a vanishingly long shot).
    int i = 0;
    bool utf8 = false;
    while (i < taster - 4)
    {
        if (b[i] <= 0x7F) { i += 1; continue; }     // If all characters are below 0x80, then it is valid UTF8, but UTF8 is not 'required' (and therefore the text is more desirable to be treated as the default codepage of the computer). Hence, there's no "utf8 = true;" code unlike the next three checks.
        if (b[i] >= 0xC2 && b[i] < 0xE0 && b[i + 1] >= 0x80 && b[i + 1] < 0xC0) { i += 2; utf8 = true; continue; }
        if (b[i] >= 0xE0 && b[i] < 0xF0 && b[i + 1] >= 0x80 && b[i + 1] < 0xC0 && b[i + 2] >= 0x80 && b[i + 2] < 0xC0) { i += 3; utf8 = true; continue; }
        if (b[i] >= 0xF0 && b[i] < 0xF5 && b[i + 1] >= 0x80 && b[i + 1] < 0xC0 && b[i + 2] >= 0x80 && b[i + 2] < 0xC0 && b[i + 3] >= 0x80 && b[i + 3] < 0xC0) { i += 4; utf8 = true; continue; }
        utf8 = false; break;
    }
    if (utf8 == true) {
        text = Encoding.UTF8.GetString(b);
        return Encoding.UTF8;
    }


    // The next check is a heuristic attempt to detect UTF-16 without a BOM.
    // We simply look for zeroes in odd or even byte places, and if a certain
    // threshold is reached, the code is 'probably' UF-16.          
    double threshold = 0.1; // proportion of chars step 2 which must be zeroed to be diagnosed as utf-16. 0.1 = 10%
    int count = 0;
    for (int n = 0; n < taster; n += 2) if (b[n] == 0) count++;
    if (((double)count) / taster > threshold) { text = Encoding.BigEndianUnicode.GetString(b); return Encoding.BigEndianUnicode; }
    count = 0;
    for (int n = 1; n < taster; n += 2) if (b[n] == 0) count++;
    if (((double)count) / taster > threshold) { text = Encoding.Unicode.GetString(b); return Encoding.Unicode; } // (little-endian)


    // Finally, a long shot - let's see if we can find "charset=xyz" or
    // "encoding=xyz" to identify the encoding:
    for (int n = 0; n < taster-9; n++)
    {
        if (
            ((b[n + 0] == 'c' || b[n + 0] == 'C') && (b[n + 1] == 'h' || b[n + 1] == 'H') && (b[n + 2] == 'a' || b[n + 2] == 'A') && (b[n + 3] == 'r' || b[n + 3] == 'R') && (b[n + 4] == 's' || b[n + 4] == 'S') && (b[n + 5] == 'e' || b[n + 5] == 'E') && (b[n + 6] == 't' || b[n + 6] == 'T') && (b[n + 7] == '=')) ||
            ((b[n + 0] == 'e' || b[n + 0] == 'E') && (b[n + 1] == 'n' || b[n + 1] == 'N') && (b[n + 2] == 'c' || b[n + 2] == 'C') && (b[n + 3] == 'o' || b[n + 3] == 'O') && (b[n + 4] == 'd' || b[n + 4] == 'D') && (b[n + 5] == 'i' || b[n + 5] == 'I') && (b[n + 6] == 'n' || b[n + 6] == 'N') && (b[n + 7] == 'g' || b[n + 7] == 'G') && (b[n + 8] == '='))
            )
        {
            if (b[n + 0] == 'c' || b[n + 0] == 'C') n += 8; else n += 9;
            if (b[n] == '"' || b[n] == '\'') n++;
            int oldn = n;
            while (n < taster && (b[n] == '_' || b[n] == '-' || (b[n] >= '0' && b[n] <= '9') || (b[n] >= 'a' && b[n] <= 'z') || (b[n] >= 'A' && b[n] <= 'Z')))
            { n++; }
            byte[] nb = new byte[n-oldn];
            Array.Copy(b, oldn, nb, 0, n-oldn);
            try {
                string internalEnc = Encoding.ASCII.GetString(nb);
                text = Encoding.GetEncoding(internalEnc).GetString(b);
                return Encoding.GetEncoding(internalEnc);
            }
            catch { break; }    // If C# doesn't recognize the name of the encoding, break.
        }
    }


    // If all else fails, the encoding is probably (though certainly not
    // definitely) the user's local codepage! One might present to the user a
    // list of alternative encodings as shown here: https://stackoverflow.com/questions/8509339/what-is-the-most-common-encoding-of-each-language
    // A full list can be found using Encoding.GetEncodings();
    text = Encoding.Default.GetString(b);
    return Encoding.Default;
}

It depends where the string 'came from'. A .NET string is Unicode (UTF-16). The only way it could be different if you, say, read the data from a database into a byte array.

This CodeProject article might be of interest: Detect Encoding for in- and outgoing text

Jon Skeet's Strings in C# and .NET is an excellent explanation of .NET strings.

Check out Utf8Checker it is simple class that does exactly this in pure managed code. http://utf8checker.codeplex.com

Notice: as already pointed out "determine encoding" makes sense only for byte streams. If you have a string it is already encoded from someone along the way who already knew or guessed the encoding to get the string in the first place.

I know this is a bit late - but to be clear:

A string doesn't really have encoding... in .NET the a string is a collection of char objects. Essentially, if it is a string, it has already been decoded.

However if you are reading the contents of a file, which is made of bytes, and wish to convert that to a string, then the file's encoding must be used.

.NET includes encoding and decoding classes for: ASCII, UTF7, UTF8, UTF32 and more.

Most of these encodings contain certain byte-order marks that can be used to distinguish which encoding type was used.

The .NET class System.IO.StreamReader is able to determine the encoding used within a stream, by reading those byte-order marks;

Here is an example:

    /// <summary>
    /// return the detected encoding and the contents of the file.
    /// </summary>
    /// <param name="fileName"></param>
    /// <param name="contents"></param>
    /// <returns></returns>
    public static Encoding DetectEncoding(String fileName, out String contents)
    {
        // open the file with the stream-reader:
        using (StreamReader reader = new StreamReader(fileName, true))
        {
            // read the contents of the file into a string
            contents = reader.ReadToEnd();

            // return the encoding.
            return reader.CurrentEncoding;
        }
    }

Another option, very late in coming, sorry:

http://www.architectshack.com/TextFileEncodingDetector.ashx

This small C#-only class uses BOMS if present, tries to auto-detect possible unicode encodings otherwise, and falls back if none of the Unicode encodings is possible or likely.

It sounds like UTF8Checker referenced above does something similar, but I think this is slightly broader in scope - instead of just UTF8, it also checks for other possible Unicode encodings (UTF-16 LE or BE) that might be missing a BOM.

Hope this helps someone!