Re: Improving single click for all GTK+ apps to make it suitable as a standard

["Freddie Unpenstein" <fredderic excite com>]

From: "Jan-Christoph Borchardt", Date: 15/05/2010 06:17:

> There has been quite a discussion on Ubuntu’s usability mailing list
> (Ayatana) about making single click standard or not:
> https://lists.launchpad.net/ayatana/msg01863.html (There are some
> mockups as well.)

Urgh.  As someone who uses a laptop with one of those little finger pads for a mouse, I find myself occasionally miss-clicking or dragging when I shouldn't.  Wherever possible, I prefer to cut/copy and paste instead of dragging, simply because it's safer (accidental dropping is another problem with those things).  This business of single-click to do stuff gives me the jitters.  I'd much rather select a file, then click the open button (or more usually, select a file and hit Enter on the keyboard).  I'll quite often select a file as a place-marker which I consider whether it was in fact the one I wanted - I then have two choices, double-click, or click the dialog button.


On the toolkit side, what I would like to see, is more cooking of the input events, similar to how terminals and X itself allow access to raw keystrokes, the processed/mapped input events, down to the final activation of a widget.

In this regard, how about a consistant mechanism across all GTK widgets to intelligently process keyboard and mouse events, kind of like the three-stage cooking that goes on in commandline terminals:

1-  the existing signals for the original raw mouse and keyboard events.  Each stage's default handler then generates the signals for the next (unless the signal emission has been stopped).  This is essentially what's presently implemented in GTK.

2- recognition of the shift/click/drag operation that the user may be performing, without the spurious extra clicks and shift keydown's, also conversion of raw key codes and meta-key combinations into symbolic names.  It could also represent a letter key as both the raw character (eg. 'A' regardless of shifting), as well as the letter with shifting in effect (ie. 'A' with shift/caps pressed and 'a' without).  Dealing with these issues have been asked on the groups a few times, and having both available allows for example, the developer to easily latch onto the raw character and then test shift as a flag if they so wish, or latch onto the shifted character when a key has two different functions depending on shifting.

3- looking up the emissions on stage two in a list of symbolic mappings, and re-emitting the resultant "action" as a final "fully cooked input" signal.  For a simple button, the stage two "left click" emission, would be mapped to a stage three "activate" signal, and a simple handler on "activate" would then emit the buttons usual clicked action (button depression would still be picked up from stage 1, as they are now).  This final stage would remove the need for a lot of the existing mouse/keyboard processing code replicated and/or re-implemented through every single widget.  Supporting per-widget meta-states would allow for regions or states of the widget to be incorporated into these final-stage mappings, and multi-stage input sequences to handle some of the odder input devices.

A widget, instead of implementing its own keyboard/mouse mapping code, could in most cases simply register a set of actions and their corresponding default mappings in the third stage processing, and let the default (or theme) mappings match those actions to their input sequences.  That would making it possible for something like Ctrl-X + 'E' to be the combination to activate a button in some weird input environment where it makes sense (for example, a magic button on a custom keyboard).


Fredderic