On Jun 6, 2014 5:42 PM, "Marcus Ottosson" <[email protected]> wrote:
>>
>> If you are doing Miller columns then that should be equivalent to
multiple views of the same data. Presumably you are modeling a path
hierarchy and each column is a view onto a different level of the same data
model (presumably). Doesn’t really matter but I just thought to mention it.
>
> Not sure I’m following you here. Are you saying that:
>
> data = ['item1', 'item2', 'item3']
>
> model = MyModel()
> model.setup(data)
>
> view1 = MyView()
> view1.setModel(model)
>
> view2 = MyView()
> view2.setModel(model)
>
> Is the same as..
>
> data = ['item1', 'item2', 'item3']
>
> model1 = MyModel()
> model1.setup(data)
>
> model2 = MyModel()
> model2.setup(data)
>
> view1 = MyView()
> view1.setModel(model1)
>
> view2 = MyView()
> view2.setModel(model2)
>
> ..due to both of them using the same set of data? I assumed what you mean
with multiple views onto the same set of data involved using the same
model, and having them both update according to it, rather than meaning
that the file-system is some sort of model.
>>

No those are two different views using two different models. I am talking
about the first one, where a single model represents the state of things
and different views can root at different levels of it and respond to the
same changes in one place.  This would be opposed to loading two models
with duplicate data. In a Miller column I figured you were showing the
hierarchy of a filesystem or file content meaning it is all the same data
source which could use the same model, but with different list views
showing different parts of the model

>> Event filtering means you can tell one object to receive all events
first for a given QObject subclass. It can choose to do something based on
the event and/or decide whether the event should continue to reach the
original target. That means you could have logic all in one class that
chooses to do things for other objects like when the mouse is pressed.
>
> Yes, in my example this is what is happening, but without event filters.
My question is, what would event filters do differently?
>
>

In the event filtering approach you would need a custom subclass to use as
a baseclass for all your widgets. You would have one class that can look at
an event and decide what to do with the object and then install it as an
event handler on any widget that needs it.

​
>
>
> On 5 June 2014 23:12, Justin Israel <[email protected]> wrote:
>>
>> If you are doing Miller columns then that should be equivalent to
multiple views of the same data. Presumably you are modeling a path
hierarchy and each column is a view onto a different level of the same data
model (presumably).  Doesn't really matter but I just thought to mention it.
>>
>> Event filtering means you can tell one object to receive all events
first for a given QObject subclass. It can choose to do something based on
the event and/or decide whether the event should continue to reach the
original target. That means you could have logic all in one class that
chooses to do things for other objects like when the mouse is pressed.
>>
>> On Jun 6, 2014 9:48 AM, "Marcus Ottosson" <[email protected]> wrote:
>>>
>>> Thanks Justin. The code example above is using custom events, and I
think it should do what I'm looking for, but I won't know for sure until
tomorrow once i've integrated it.
>>>
>>> As for event filters, don't have much experience with those, but do you
think they would allow for the same thing? Without or with custom events?
>>>
>>> And finally, multiple views is fancy but are not important in this
situation. If that was a requirement, I would definitely go for separating
the model.
>>>
>>> Ultimately, I believe that my requirements are low, that there is a
straightforward solution, and that throwing pre-packaged widgets or
QAbstractItem* at the problem is only escalating things.
>>>
>>> Looking forward to your thoughts on the example, I find it nimble and
complete, more sophisticated than what I've got going on currently and
mainly simple and easy to communicate to others with minimal knowledge of
Python.
>>>
>>> Best,
>>> Marcus
>>>
>>> On Thursday, June 5, 2014, Justin Israel <[email protected]> wrote:
>>>>
>>>> I wouldn't call it a magic bullet either. They are more complicated to
set up for sure. But they do work well in situations where you have
multiple views on the same data. I've done the Miller Column thing both
using the built in QColumnView, and also setups that use the vertical
layout with custom widgets, like you are doing. The QColumnView is pretty
much a bunch of list views in a horizontal scroll that have their root item
set to different levels of the same model. So multiple views share the same
data. In the case of the custom widgets in vertical layouts, when I have
done stuff like that, I have usually given the "run" button equivalent the
context it needs to use when the click happens. Like giving it a prewrapped
callback that it does not have to know about. But you obviously know at the
time of its creation that it has a context.
>>>>
>>>> But ya lets just leave the whole model thing aside because it is a
drastic mental shift than what you are already doing. Always something that
can be played with later. I can try and take a more focused look at your
code example tonight, but it does seem like something could be done with an
event filter or customEvents. Ultimately as long as you get some sort of
mechanism that pushes events up the chain, you would be good. I wasn't
clear on the part where you said you didn't want to overload the event()
method since you weren't sure of its logic. But I would think you could
still do that and perform save operations, and then call the original event
handler at the end.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> On Fri, Jun 6, 2014 at 7:51 AM, Marcus Ottosson <[email protected]>
wrote:
>>>>>
>>>>> Well, the general issue I see with the QAbstractItem* family is its
complexity and that it seems better suited for massive datasets, for which
I suspect it was designed.
>>>>>
>>>>> I believe it can be simpler if datasets aren't as large. (less than
hundreds of items)
>>>>>
>>>>> If you have a look at the example I just posted, can you see where
such a model would fail and where a QAbstractItem* approach would make it
easier to maintain? Even if we disregard the low count in items, and
disregard performance, can you see anywhere where this model would stop
making sense?
>>>>>
>>>>> I'm trying hard to KISS, QAbstractItem* isn't simple enough. I really
don't think it's as much of a silver bullet as it you're making it out to
be.
>>>>>
>>>>>
>>>>> On 5 June 2014 20:44, Justin Israel <[email protected]> wrote:
>>>>>>
>>>>>> Yea sorry. Apparently I don't get it either. I was following Tony's
reply and it made sense to me but I suppose was not applicable to you? I
also saw it as a situation where child items in a model know their own
context and their parent item. So when representing the "run" item, it
knows the context of it being a "run command for snake within the stats
location" because it all shares the same underlying data to know that. The
fact that it is a button with a signal is part of the view/delegate aspect.
A custom data role could give you back the "context" object that you need
for the view or delegate to perform the action in response to the click.
>>>>>> Again,  like Tony, I apologize if this sounds complicated or
abstract. It still doesn't click for me that it needs to be a massive
signal passing situation. Only so if you are purely working with widgets
that are nested and don't know about each others logic as opposed to a
model that already knows all the data in one place.
>>>>>> The performance and data aspects of the model aren't just about what
is currently visible. It is also about all the data that is know but not
visible. Like if you are modeling a filesystem, you may not be in a
location viewing a ton of files and directories, but that doesn't mean the
model couldn't resolve sibling or child directory listing ahead of time and
have extra data available. Or that you could resolve info about parent
locations that may not be displaying the data in a view, but have the data
available in the model (like the listing of a directory for which you had
just browsed through while navigating down).
>>>>>> I can see why in this situation you would need to pass a lot of
dynamic signals, when widgets are completely autonomous. If you are
sticking with that approach then it might be good to keep trying that
customEvent propagation technique that people list online,  and seeing if
you can get it to work. Your custom widgets would probably have to have a
customEvent method implemented to handle the event if they want. The manual
propagation method you found gets around that need by continuing to drive
it up the chain until someone handles it. I'm not sure off hand but maybe
the customEvent default implementation accepts the event.
>>>>>>
>>>>>> On Jun 6, 2014 3:40 AM, "Marcus Ottosson" <[email protected]>
wrote:
>>>>>>>
>>>>>>> I'm making an appearance here too.
>>>>>>>
http://stackoverflow.com/questions/24064422/traverse-events-through-hierarchy-of-unique-widgets
>>>>>>>
>>>>>>>
>>>>>>> On 5 June 2014 16:38, Marcus Ottosson <[email protected]>
wrote:
>>>>>>>>
>>>>>>>> Aw. :'( Ok, thanks again for sticking with me this long!
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> Marcus Ottosson
>>>>>>> [email protected]
>>>>>>>
>>>>>>> --
>>>>>>> You received this message because you are subscribed to the Google
Groups "Python Programmi
>>>>>
>>>>> To view this discussion on the web visit
https://groups.google.com/d/msgid/python_inside_maya/CAFRtmOCLF5dCi%3DPmE1ZKGOttM5dn3HqHCX0PXww5HHRdoFMBXg%40mail.gmail.com
.
>>>>>
>>>>> For more options, visit https://groups.google.com/d/optout.
>>>>
>>>>
>>>> --
>>>> You received this message because you are subscribed to the Google
Groups "Python Programming for Autodesk Maya" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send
an email to [email protected].
>>>> To view this discussion on the web visit
https://groups.google.com/d/msgid/python_inside_maya/CAPGFgA3gR-mwQBc7Q1L25GAcajvoJ2VeRA6oQAXkiP8ZrNABmA%40mail.gmail.com
.
>>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>>
>>>
>>> --
>>> Marcus Ottosson
>>> [email protected]
>>>
>>> --
>>> You received this message because you are subscribed to the Google
Groups "Python Programming for Autodesk Maya" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
an email to [email protected].
>>> To view this discussion on the web visit
https://groups.google.com/d/msgid/python_inside_maya/CAFRtmOBAYw%2BLNYEqUsoFEtttMyEKctNoCsPXv9Vbyyp-wrbaiA%40mail.gmail.com
.
>>>
>>> For more options, visit https://groups.google.com/d/optout.
>>
>> --
>> You received this message because you are subscribed to the Google
Groups "Python Programming for Autodesk Maya" group.
>> To unsubscribe from this group and stop receiving emails from it, send
an email to [email protected].
>> To view this discussion on the web visit
https://groups.google.com/d/msgid/python_inside_maya/CAPGFgA2Qk6wJ41qzUymzxLTTJQuSp%2BXxLJ7W9VwkFk8mUmPemw%40mail.gmail.com
.
>>
>> For more options, visit https://groups.google.com/d/optout.
>
>
>
>
> --
> Marcus Ottosson
> [email protected]
>
> --
> You received this message because you are subscribed to the Google Groups
"Python Programming for Autodesk Maya" group.
> To unsubscribe from this group and stop receiving emails from it, send an
email to [email protected].
> To view this discussion on the web visit
https://groups.google.com/d/msgid/python_inside_maya/CAFRtmOC4mxs67k6HMYGJ5AOPKORo4C1jDbfofgLFQbcXnFoY2Q%40mail.gmail.com
.
>
> For more options, visit https://groups.google.com/d/optout.

-- 
You received this message because you are subscribed to the Google Groups 
"Python Programming for Autodesk Maya" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/python_inside_maya/CAPGFgA209ZfdtcpKvYD859-NksH1F8vs9DS%3D%3DoW9Yi2PuoYeBQ%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to