Re: gtk_editable_select_region and cursor position

[Owen Taylor <otaylor redhat com>]

On Thu, 2003-04-24 at 06:47, Matt Keenan wrote:
> Hi,
>
> Came across this problem in gtk and not sure if solveable...
>
> Scenario :
>
> Trying to implement command history within a gtkentry. As you are entering 
> text I want to be able to display a match from a list of previously entered
> strings. I monitor the key_press_event on the GtkEntry widget, and get the
> current text position, attempt to make a match and if there is a match
> disply this match with the extra text highlighted, however when the user
> enters the next key the current position returned from
> gdk_editable_get_position()
> is the end of the string as opposed to where the next char is typed.
>
> e.g.
>       List of Previously entered strings :
>               string1
>               sring1
>               sing1
>
>       Order of entry
>
>       1. type "s"
>               Within the key_press_event callback
>       
>               gtk_editable_get_position -> 0
>               
>          I match the string "sing1"
>
>               gtk_entry_set_text("sing1");
>               gtk_editable_set_position(current_position+1);
>               gtk_editable_select_region(current_position+1, -1);
>
>          This displays the text "s" unhighlighted and the rest "ing1" highlighted.
>
>       2, type "i"
>
>          The "i" appears directly after the previously typed "s" which is as
>          i expect however
>
>               gtk_editable_get_position -> 5
>          surely it should return 1....
>
> Can anyone answer me as to why this is happening, or am I doing something
> wrong..

Hard to say what's going on without seeing a small standalone test case;
I would suggest that you shouldn't be doing stuff off of key_press_event
(that won't work when you have internationalization and input methods),
but rather off of ::insert_text.

If you look at gtkfilechooser/gtkfilechooser.c, you'll find something
similar. (Though it's certainly a little complex for an example...)
What I ended up doing there was to install an idle handler in
insert_text, where the idle handler checked for completions.

Regards,
                                               Owen