Of course, you need Alain Frisch's ulex utility first. It installs itself under the name ulex as findlib library.

Next, write your lexer, e.g. (line numbers in brackets):

[1] open Netulex
[2] let digits = lexer
[3]   | ['0'-'9']+ -> `Number(int_of_string(Ulexing.utf8_lexeme lexbuf))
[4]   | 8364       -> `Euro_sign   (* Code point #8364 in Unicode *)

This is a very trivial example. The lexer accepts sequences of digits, and returns them as `Number tokens. Furthermore, the euro sign is recognized and returned as `Euro_sign. Note that in the first case Netulex.Ulexing.utf8_lexeme is called to retrieve the current lexeme as UTF-8 string. (Well, digits are a bad example, as they are only ASCII, and UTF-8 is not really needed. Imagine you want to extend the scanner to other number systems represented in the Unicode character set.)

Line 1 is quite important. If you don't open Netulex, the generated ulex code will use the version of the Ulexing module coming with ulex, and not this one.

Call the lexer as follows (line numbers in brackets):

[5] let sample = "42543\226\130\172";;
[6] let ulb = Netulex.ULB.from_string `Enc_utf8 sample;;
[7] let lexbuf = Netulex.Ulexing.from_ulb_lexbuf ulb;;
[8] let first_token = digits lexbuf;;

Now, first_token is `Number 42543. After

[9] let second_token = digits lexbuf;;

this variable is set to `Euro_sign, because the three-byte sequence "\226\130\172" represents the euro sign in UTF-8.

In line 6, the encoding `Enc_utf8 selects that sample is an UTF-8 string. You can pass here any encoding the Netconversion module understands.

If you would like to scan from another source, just change line 6, e.g.

[6'] let ulb = Netulex.ULB.from_in_obj_channel ch

where ch is any input channel the Netchannels module supports. For example, to read from a file:

let ch = new Netchannels.input_channel (open_in "filename")

You should compile the examples with

ocamlfind ... -package ulex,netstring -syntax camlp4o ...

For the syntax of the lexer rules, see the documentation coming with ulex.