Re: Another --> Re: GtkMovementStep of GtkTextView

[Steve Underwood <steveu coppice org>]

Pablo Saratxaga wrote:
> 
> [...]
> 
> No, not Vietnamese vowels (where do you get that idea about Vietnamese? it
> is just plain latin alphabet with some accents. it just has a special
> input method very used in Vietnam, but the input method is not related
> to the text (there are three special and popular input methods; and you
> can even use standard dead keys composition, the result will be the same;

This puzzles me too. I keep seeing people describe Vietnamese like its a
complex script, and as you say, its just Romance plus some special local
accent marks. People don't seem to be separating input methods from text
at all well in these discussions.

> obviously the behaviour of BackSpace must be the same in all cases, unless
> the BackSpace knows what the input is, that is, unless the BackSpace is
> pressed during input. I mean, I can see an utility to have a special
> behaviour for the BackSpace when inputing; think of it as the behaviour
> of BackSpace in chinese or japanese XIM input method before validating:
> it deletes the phonetic kana/bopomofo/latin letter last typed on input;
> however, once validated, a backspace will delete the last ideograph,
> because the info on the input stream has been lost, there is only the
> stored char in the text.

It goes a bit beyond that with Chinese (and I think Japanese). If you
are using an intelligent input method you could enter several characters
before they become committed elements of the text. During that
composition period backspace behaviour is very different to the
character by character movements you experience when scanning the cursor
through some Chinese text. Exactly how backspacing works for the
composition buffer is very dependant on the input method itself. For
example, in some intelligent BoPoMoFo input packages the user types a
stream of BoPoMoFo, without breaks. A lexicon and some tricky algorithms
are used to find how that is likely to translate into HanZi, and
produces that translation in a area of the display. Backspacing here is
letter by letter of the BoPoMoFo alphabet, until you reach the start of
the composition buffer. With intelligent forms of the ChangJei input
method, you enter a few strokes to form a character. During that time
the backspace key will remove a single entered keystroke. Once a whole
HanZi is formed in the composition buffer, backspacing over it will
eliminate the whole HanZi.

The point I am trying to illustrate here is that for just a single
language backspacing behaviour can be very input method dependant.
However, the tricky parts of backspacing are only related to that input
method composition period. Once text is commited, it is treated in a
very simple way for cursor movement purposes. From the discussion I have
seen here, it is not obvious to me why any of these other languages are
fundamentally different from this. If the complexity is consigned to the
input method, then surely it can be eliminated from the general text
handling mechanisms. On the other hand, since cursor movement within the
compisition area is very input method dependant, the general language
handling mechanisms will never be able to support the full richness that
a new input method might need.

Regards,
Steve