Re: Another --> Re: GtkMovementStep of GtkTextView

[Pablo Saratxaga <pablo mandrakesoft com>]

Kaixo!

On Fri, May 18, 2001 at 01:54:05PM -0700, Chookij Vanatham wrote:
 
> I just make sure that the sample below "ü<Delete>" is actually the
> <baskspace> case. It should be "ü<Backspace>" produces u.

IMHO it sohuld indeed be backspace (for me backspaces delete something
before the cursor; delete deletes something after the cursor (note: not
at the left or the right, but before and after; there is RTL mode...)

> I thought that <backspace> would delete one character in front of
> the current cursor position.

in front = before ?

the problem is in fact what "a character" is.

> <Delete> will delete the current whole cluster.

as delete deletes what is *after* the cursor, it cannot delete the first
element of a conjunct (it will lead to a remaining invalid thing in thai
for example); so the whole conjunct is deleted; it is normal.

> Not quite sure about "Korean and Vietnamese" as mentioned by "Mark Leisher"
> but if it means "<backspace>" in both Korean and Vietnamese will
> delete one character in front of the current cursor position, then,
> seems to be Thai/Indic/Korean/Vietnamese behave the same for both
> <backspace> and <delete>

Again, the key is on what "a character" is...

Also, very important imho, is the visual creation of the text.
on latin, greek and cyrillic, the letter only is displayed full with all
its diacritics, eg, you type: <dead_diaeresis> and nothing displays,
and only when you type <u> it displays <udiaeresis>.

Vietnamese also has a different and popular input method called "telex";
it uses just plain ascii, and uses some letters never used in Vietnamese
to tell the kind of accents; however, those letters don't follow directly
the vowel (well, they can, but it is not the preferred typing way) but
rather the whole syllabe. Here, if you hit <Delete> after a syllabe
ending with an 'n', it will be odd to have an 'adotbelow' become a simple 'a'
instead of the 'n' being deleted.

And that leads to I think an interesting idea: the behaviour of BackSpace
should be dependent on the flow of the typing.
I explain:

If the typing method is one where the shape of conjuncts changes while
you type it (eg: like thai, indic, korean, arabic, viet-telex; but 
unlike normal use of latin, cyrillic, greek); then:

- Backspace should delete the effect of the last typed key if you didn't 
  do anything since then (no mouse focus change, no other keys hit)
- otherwise, it should delete the conjunt before the cursor.


the base for that idea is to have a non-shocking behaviour. undoing the
effect of the last keystroke can be expected when you just did it right now;
if you do other things then get back, you don't think anymore on the specific
keystroke flow, but rather you see a single visual unit.


What do you think about it; would the implementation of such a thing be
feasible without much trouble ?

What native users of the involved scripts feel about it?


-- 
Ki ça vos våye bén,
Pablo Saratxaga

http://www.srtxg.easynet.be/		PGP Key available, key ID: 0x8F0E4975