[Development] Touch Points Update

[Frederik Gladhorn]

Hello Francisco,

Lørdag 8. juni 2013 13.06.20 skrev Francisco Ortega:
> Hello all,
> 
> I spoke to Oliver Wolf this past Monday (6/3/2013) and a few other members
> (#qt-labs) about the qtbase/src/plugins/platforms dealing with Touch. Like
> always, everyone is extremely helpful.
> 
> I started talking about Windows having a few additional data members for a
> given touch point. Most importantly, I spoke about the unsigned long ID
> given by windows, which is not the same. Without going into much
> explanation here, I spoke in IRC that I used that unique ID to store touch
> points to a database to later reply them. The current id for a windows app
> running Qt does not does the same, since it can go from 0,1,2... and later
> 0,1,2,3 in the same moment. As most of you may guess, I could do something
> to fix this for me.

If you store them in a database you can just map your own unique IDs there, 
whenever a touch point appears, assign it a new ID and keep track of it.
Qt uses consistent IDs.
As long as  a touch event is ongoing, the ID will stay the same. When a finger 
is lifted (lets say IDs 1,2,3 were pressed) and 2 goes away and is re-added, 
it will be a new unique ID compared to the others (you might end up with 
1,3,4).
I don't see how the actualy unique ID is relevant, you just need to be able to 
identify individual points and track them from what I understand.
 
> However, I was thinking at a larger scale. Therefore, I went and check the
> three platforms that I have accessed to. Windows, Linux (Debian 7), and Mac
> OS X 10.8. Windows is the only platform that I know well.
> 
> With the search, I found that Linux and MacOSX has what is already given by
> Qt. x,y,state. Linux also has a timestamp, which I didn't no see in
> QTouchEvents::TouchPoint (in linux is under time, but I think is a
> timestamp). Windows has the timestamp as well.

What about QInputEvent::timestamp()?

> Windows also has Cx,Cy which is the contact area. Not the most useful data
> at least with my Multi-Touch monitors (3M) because it gives me two
> values... but nevertheless, a value that is no there.

This might be a good addition, it can go directly into
QTouchEventTouchPointPrivate (qevent_p.h).

> With all this preamble, I would like to get to the meet of the change that
> I will submit and see if it gets reviewed. First, let me tell you what I
> discarded. I hope your feedback about all.
> 
> (First ideas, that I decided not to go for)
> 
> 1) Returning a void * to a platform structure, it is not a good idea, so is
> out.
> 2) Changing the current id (int) to unsigned long can have multiple
> unintended consequences to current code, that is best not to do it. So is
> out.
> 3) setting the platform id (windows) to the int by moding with a "MAX_INT"
> does not feel right  to me.
> 
> What I will change in the code and submit when ready is the following:
> 
> 1) Create an interface to used a platformSpecificData() that will be the
> same across all Operating Systems. For example, say that we have
> data.cxfor mac os x. Then this will simply be 0, the data.flags will
> specify if
> the data is available.
> 
> I hope to receive feedback. I'm not sure if I can do it this month, since I
> have to submit a paper to a conference, but I will try. if not next week.
> Because the change may be more significant, is probably that even if review
> and accept it, it will take a while to make it to the stable release.
> However, I think this is the best option.
> 
> Once again, I thank you for all the feedback and continue support here and
> in Irc.
> 
> Other contributions in the future that I hope to be involved is the "Leap
> Motion" which I will be getting soon and some MEMS that I have already
> received. I like input and please consider me for any work you may need.

Help with the input events is most certainly appreciated :)
Greetings,
Frederik


> 
> Thanks,
> Francisco
> 
> 
> Francisco R. Ortega
> Ph.D. Candidate in Computer Science
> Florida International University
> http://www.FranciscoRaulOrtega.com
> "No me quieras por que gane, necesito que me quieras para ganar" -- Marcelo
> Bielsa
-- 
Best regards,
Frederik Gladhorn
Senior Software Engineer - Digia, Qt
Visit us on: http://qt.digia.com