.getc won't permit completely raw input

[geekosaur]

The example code is wrong in that the getc method will always wait for additional input in order to determine whether it is a combining character or not.

To actually do raw input, you need to set the IO::Handle (e.g. $*IN) to binary mode ($*IN.encoding(Nil);) and use $*IN.read(1) to read a single byte; note that this will give you a Buf instead of a Str. (A quick hack for that is $*IN.read(1).decode("latin-1"), especially since Buf is currently not as easy to work with as Str.)

The raw input mechanism may require a somewhat recent moarvm and matching rakudo to work properly.

[AlexDaniel]

There's a rakudobug that I am rejecting with similar reasoning: RT#125828. I think something like INIT { $*IN.encoding: ‘ASCII’ } should work too.

[krunen]

I have made a module Unicode::UTF8Parser that parses UTF8 chars from a stream without waiting for combining chars. I can change the example to use this module.

I have held back doing this, as I think Rakudo should expose this as some kind of parser mode. It seems silly to reimplement a full UTF8 parser in plain perl6. Also, I am not sure how Rakudo will handle un-normalized unicode (that is, with base char and combining chars separately). I have used it myself without much problems though - the strings seems to become normalized quite quickly.

[krunen]

I have added a working example to read single chars from $*IN.