UTF-8 bit by bit (2001)

Richard Suchenwirth 2001-02-28 - From a delightful debugging chat at theTcl chatroom, I was brought to write down what I think on UTF-8 analysis (cf.Unicode and UTF-8, see also).

I imagine a UTF-8 string as a railroad. It operates single-unit railcars (one-byte ASCII characters, to be known from the fact that the highest bit is 0), and trains (sequences of two or more bytes that together form a character). Each train consists of exactly one locomotive (you see I'm European) and one or more trailers . The locomotive indicates the length of the train, including itself, in the highest bits that form a consecutive row of 1's, and one 0 bit. Examples:

0xxxxxxx : I'm a railcar, just a single unit
 110xxxxx : I'm leading a train of length 2

The xxxxx bits are used for other purposes (locomotives carry some freight too ;-)

Trailers indicate that they are trailers by the initial bit sequence 10. This way, they can't be mistaken for railcars or locomotives. E.g.

10yyyyyy: I'm a trailer

The freight of the train is in the x's and y's. In that concrete case, a C program reported to have received the bytes C3 and A4. Written as binary, that's

Now for clarity we delimit the indicators with parens:

(110)00011 (10)100100

and can just remove them:

00011 100100
 11100100 => E4, the iso8859-1 value for German ä ("a umlaut").

The generalized rule for the indicator of each byte is "those bits from highest(leftmost) down, up to and including the first zero bit".

Now going the other way. In orthodox UTF-8, a NUL byte (\x00) is represented by a NUL byte. Plain enough. But in Tcl we sometimes want NUL bytes inside "binary" strings (e.g. image data), without them terminating it as a real NUL byte does. To represent a NUL byte without any physical NUL bytes, we treat it like a character above ASCII, which must be a minimum two bytes long:

(110)00000 (10)000000 => C0 80

Whoops. Took us a while, but now we can read UTF-8, bit by bit.

andrewsh 2010-03-12 - Please note that 0xc0 0x80 sequence is illegal in the "Real" UTF-8: [], []