why doesn't byte[] to string and back work as expected

Question

I have this code:

Int32 i1 = 14000000;
byte[] b = BitConverter.GetBytes(i1);
string s = System.Text.Encoding.UTF8.GetString(b);
byte[] b2 = System.Text.Encoding.UTF8.GetBytes(s);
Int32 i2 = BitConverter.ToInt32(b2,0);;

i2 is equal to -272777233. Why isn't it the input value? (14000000) ?

EDIT: what I am trying to do is append it to another string which I'm then writing to file using WriteAllText

Answer 1

Because an Encoding class is not going to just work for anything. If a "character" (possibly a few bytes in case of UTF-8) is not a valid character in that particular character set (in your case UTF-8), it will use a replacement character.

a single QUESTION MARK (U+003F)

(Source: http://msdn.microsoft.com/en-us/library/ms404377.aspx#FallbackStrategy)

Some case it is just a ?, for example in ASCII/CP437/ISO 8859-1, but there is a way for you to choose what to do with it. (See the link above)

For example if you try to convert (byte)128 to ASCII:

string s = System.Text.Encoding.ASCII.GetString(new byte[] { 48, 128 }); // s = "0?"

Then convert it back:

byte[] b = System.Text.Encoding.ASCII.GetBytes(s); // b = new byte[] { 48, 63 }

You will not get the original byte array.

This can be a reference: Check if character exists in encoding

---

I can't imagine why you would need to convert a byte array to a string. It obviously doesn't make any sense. Let's say you're going to write to a stream, you could just directly write byte[]. If you need to use it in some text representation, it makes perfect sense to just convert it to a string by yourIntegerVar.ToString() and use int.TryParse to get it back.

---

Edit:

You can write a byte array to a file, but you are not going to "concatenate" the byte array to a string and use the lazy method File.WriteAllText because it is going to handle the encoding conversion and you will probably end up having question marks ? all over your file. Instead, Open a FileStream and use FileStream.Write to directly write the byte array. Alternatively, you can use a BinaryWriter to directly write an integer in its binary form (and also a string) and use its counterpart BinaryReader to read it back.

Example:

FileStream fs;

fs = File.OpenWrite(@"C:\blah.dat");
BinaryWriter bw = new BinaryWriter(fs, Encoding.UTF8);
bw.Write((int)12345678);
bw.Write("This is a string in UTF-8 :)"); // Note that the binaryWriter also prefix the string with its length...
bw.Close();

fs = File.OpenRead(@"C:\blah.dat");
BinaryReader br = new BinaryReader(fs, Encoding.UTF8);
int myInt = br.ReadInt32();
string blah = br.ReadString(); // ...so that it can read it back.
br.Close();

This example code will result in a file which matches the following hexdump:

00  4e 61 bc 00 1c 54 68 69 73 20 69 73 20 61 20 73  Na¼..This is a s  
10  74 72 69 6e 67 20 69 6e 20 55 54 46 2d 38 20 3a  tring in UTF-8 :  
20  29                                               )   

Note that BinaryWriter.Write(string) also prefix the string with its length and it depends on it when reading back, so it is not appropriate to use a text editor to edit the resulting file. (Well you are writing an integer in its binary form so I expect this is acceptable?)

Answer 2

To make a long story short:

You need a encoding that maps each bytevalue to a unique char and vice versa. A UTF8 Character can be from 1 to 4 bytes long so you wont archive that mapping, you need a more basic encoding like ASCII. Unfortunaly the original ASCII doesnt do that, it is just a 7-bit encoding and only defines the lower 128 Codes, the upper half (extended codes) is codepage specific. To get the full range translation, you just need a complete 8-bit encoding like in codepage 437 or 850 or whatever:

Int32 i1 = 14000000;
byte[] b = BitConverter.GetBytes(i1);
string s = System.Text.Encoding.GetEncoding(437).GetString(b);
byte[] b2 = System.Text.Encoding.GetEncoding(437).GetBytes(s);
Int32 i2 = BitConverter.ToInt32(b2,0);