Language selection in speechd-el

[Milan Zamazal]

>>>>> "PL" == Pierre Lorenzon <devel at pollock-nageoire.net> writes:

    PL> 1. If I'am not wrong current-input-method is something that must
    PL> be selected manually by the user. Simply tell me if I read emacs
    PL> documentation too fast !!! 

I personally set input method mostly automatically in my mode hooks,
based (indeed) on the buffer coding, file names etc.

    PL> But for instance coding systems is something that is guessed by
    PL> emacs with respect to the buffer content and more or less
    PL> reflecting the language you are using. I would consider as
    PL> natural then to offer a mapping between coding systems and
    PL> languages in speechd-el. I think that I is not something very
    PL> complicated to implement, simply creating an alis just like
    PL> speechd-speak-input-method-languages but mapping coding systems
    PL> on languages and that must be checked at the appropriate
    PL> location in the code.

I think that although this would be easy to implement it would be out of
speechd-el scope.  There are more things you may want to set based on
buffer coding (language), for example an input method or a spell checker
dictionary.  So IMHO it would be better to provide a general mechanism,
not specific to speechd-el, for language guessing.  Then you could set
speechd-language easily within this mechanism.

    PL> 2. More surprisingly I tried to use mapping between faces and
    PL> voices via speechd-face-voices variable. When I change
    PL> (manually) current-input-method in a buffer, language is changed
    PL> accordingly speechd-speak-input-method-languages but not for
    PL> pieces of text displayed whith faces mapped to special voices !
    PL> Why ? I must confess that I do not understand this priority
    PL> schedule. In my opinion, face does not depend on the language as
    PL> input-method do : hence language determined by input-method
    PL> should override that determined by face.

The idea is that it may help handling multilingual buffers.  For
instance Emacs doesn't provide general mechanism for message
translations so you may encounter English pieces of text in non-English
buffers.  A typical examples might be Info or w3m-el.  In such cases you
can map faces known to contain English texts to English.  speechd-el
does something similar with echo area where only English texts are
expected.

But if you'd prefer to override face language settings with
speechd-speak-input-method-languages settings, I can add an option for
it.  Just tell me.

Regards,

Milan Zamazal