Re: predictive mode and 'ã'

[Rupert <rswarbrick@googlemail.com>]

On Jul 14, 11:14 am, Anselm Helbig <ans...@chemie.fu-berlin.de> wrote:
>
> I haven't used predictive mode yet, so I can't comment on that. But
> the usual way emacs determines if some character belongs to a word or
> not is by syntax tables; these tables are mode-specific. When I put
> point on the letter `ã', `C-u C-x =' gives me some information about
> the character (I'm using GNU Emacs 22), and it says that it's syntax
> is "w which means: word".
>
> What emacs are you using, and what mode are you in? Does M-f
> (forward-word) and M-b (backward-word) stop at accented characters?
> Maybe you can do something about it by fiddling with the syntax
> table (look them up in the elisp manual).
>
> HTH,
>
> Anselm

Could it be that the predictive mode is interpreting your hitting the
~ key as an input of "~" (which would end the word, as it's
punctuation)? Then you'd get the rubbish completion you're describing.
Once you type the following a, however, the string "~a" is squished
into just "ã" and so you end up with "não", which emacs correctly once
again decides is a single word.

The fundamental problem seems to be that predictive mode has no way of
knowing which bit of a "multi-key character" is the last bit and thus
at which point it should decide whether to expand. Maybe there's a
different hook that predictive mode could be using (does emacs send
out one when it finally decides on a character?)

A work-around might be to have a look at different input systems you
might use:

1) If you're in X, you could set, say, your Alt-Gr key to act as
compose. Then you could type something like Alt-Gr ~, a
2) Possibly faster to type with: I know that the German entry mode for
emacs has a "post" form, so you can type ä by ae. Is there a
Portuguese mode similar? Then the punctuation would happen after the
letter and maybe predictive mode wouldn't get confused.

Rupert