Re: Key bindings and subclasssing

[Yevgen Muntyan <muntyan tamu edu>]

Owen Taylor wrote:

> IMO, there is no API in GtkTextView for customizing the cursor
> navigation; anything you do ... whether overriding move-cursor,
> connecting to key press events, etc, is a hack. 
>
> Now, these hacks are
> "ABI supported", in that we are very unlikely to remove the signal
> or modify it, but they are inherently fragile and are not really
> supportable at the level of behavioral details. If some new feature
> is added to GtkTextView it may have strange and unpleasant interactions
> with your hack.
>  
Sure, I do realize that the safest would be not to do anything. But what 
if I do want to
achieve certain goals? Should I write a new text widget, or copy/paste 
GtkTextView sources
and modify them? First is plain impossible, second is stupid, isn't it?
'Forward compatibility' is impossible anyway; say, I have undo/redo 
here, and it will likely
break when stock GtkTextView and GtkEntry finally get it.

> In my further opinion, the right sort of API for customization here is
> likely not at the level of user actions, but at the level of how
> GtkTextView understands its contents. 
Well, this is what I am trying to do in the subclass. But, when user 
presses ctrl-left, the widget
must do the right thing, regardless of the philosophy/design behind.

> After all, if you customize
> word navigation by overriding move-cursor, you've created a disconnect
> between the word boundaries for keynav and the word boundaries the
> users sees when they double-click to select words with the mouse.
>  
Right, I have to override mouse stuff too. Come on, changing something 
doesn't necessarily imply
creating broken inconsistent mess.

Best regards,
Yevgen