[erlang-questions] segments of variable length

[Richard Andrews]

> I receive from several machines messages and I would like to extract data.
> My first approach was:
> 
> handler(Data)->
>     case Data of 
>         <>
>         etc...
> 
> My problem is that some segment have a fixed length (header, Status) and some 
> others a variable length (Time <= 6 bytes, UnitID <= 8 bytes).
> My approach is bad. I thought to do:      Segments = string:tokens(Data,",") 
> This separates the segments but I don't have their name 
> (header,unitid,time,status) !
> Could you advice me a better approach. My goal is to check each segment and make 
> actions according to their values.

You need to define a message format such that the required information is present to determine how to parse the rest of the message. This isn't really an erlang-specific problem but erlang is very good at supporting this.

As an example you could use erlang:term_to_binary() and binary_to_term() for the message encoding on the wire. binary_to_term() might fail if the message is incomplete. This is where the {packet, N} inet options for connect/listen come in, eg. {packet,2} will only return complete messages; the length header is added automatically on send and stripped on receive so you only need to do, eg.

gen_tcp:send( Sock, term_to_binary( MessageTerm ) )

and if you receive {tcp, Sock, Data}, then it is safe to call

MessageTerm = binary_to_term(Data)

Check out the inet module documentation. There are lots of useful options.

Hope that is useful

--
  Rich


      The new Internet Explorer 8 optimised for Yahoo!7: Faster, Safer, Easier.