Chronological Thread 

< Chronological >      < Thread >

---

one that I forgot in the previous message is that it would be nice if
the file open dialog remembered where the last file was opened and
either provides a short cut back to previous folders or automatically
goes back to the last folder a model was loaded from.

>> - still getting the rename dialog every time I right click on a model in
>> the model tree unless I specifically choose an option or move the mouse
>> away from the pop-up menu. Similarly, a right click in the graph window
>> will always result in a zoom out unless zoom full is specifically
>> highlighted or the mouse is moved away from the pop-up menu.
>>
> I think what is happening is that you are unintentionally moving the
> mouse over the menu, selecting it, and then releasing while it is
> selected. I don't believe there is any non-standard behaviour here, but
> it could be slightly different to the rest of your system (because
> Mozilla menus are not true system menus). I don't really see what can be
> done about this.

I'm definitely not moving the mouse and there is clearly no item on the
pop up menu highlighted. If I do move the mouse onto the menu I get the
various items highlighted as expected. I'll let you know if I observe
this behaviour on other machines.

>> - once I have loaded several models and run the integrations, it is slow
>> to swap between models in the model tree.
> How slow are we talking about (i.e. how long does it take for the model
> tree view to be populated from after the switch)? Have you restarted
> cellml_corba_server since the previous release? On my system, it takes
> under 2 seconds to switch between two instances of your ten Tusscher model.

yep - I prefer not to have the cellml_corba_server running all the time
so I generally kill it after I close pcenv. On my machine its taking
more than 5 seconds to swap between slightly different versions of the
ten Tusscher model (parameter changes only). Which when you start
wanting to compare 5 or 6 variants of the model starts to be a real pain...

>> Given that all the models have
>> been run and have results, they can not be altered at all so there
>> should be no code generation going on at all. So does this confirm that
>> the RDF generation to make the variable list display is slowing down the
>> interface?
>>
> Older versions had two slowing factors (which used to take about half of
> the time each):
> 1) Code generation.
> 2) Populating the model tree view.
> The situation is now much better, because I have optimised the code
> generation, and put the model tree view populating code into C++. Note
> that we do currently re-check if the model is valid for integration
> (which requires most of the code generation to complete) after every
> switch, even if the model is frozen (we could cache this information,
> but it probably shouldn't be on the branch).

Sounds good. I guess if making it more usable will require big changes
like you mention then it makes sense to leave them until the next version.

>> - and to follow on from above, when I select a different model in the
>> model tree, the model is highlighted immediately which to me indicates
>> that the model has been successfully selected and I can continue my
>> work. However, due to the time taken to update the rest of the interface
>> I still have to wait before I can do anything. Would be nice if the
>> selected model highlighting didn't happen until after everything was
>> ready for me to work with the selected model. Maybe unhighlight the
>> previous model but don't highlight the newly selected model until
>> everything is ready.
>>
> This is non-trivial to do in Mozilla, since Mozilla decides when to
> update the UI. However, it might be worth looking into this for PCEnv 0.2.

Its a feature that would at least make pcenv more usable to me, but no
rush for the 0.1 release.

>> - by default there is no visible pane/window/whatever for the traces in
>> a graph. When I click "New" (which maybe should be relabelled "New
>> Trace"?) nothing apparently happens....I have to resize the traces pane
>> to be able to see that I have been adding traces to the graph.
>>
> This depends on the screen resolution, the size of the controls is
> determined by XUL 'flex' (most of the space for the key is often taken
> up by the header). We could only do anything about this at the expense
> of some graph space, and it would look different at every screen resolution.

that just seems really bad to me. As a new user I essentially can't draw
any graphs because of my screen resolution? I'm currently using a
1280x1024 LCD screen and no matter what size I make the pcenv window the
key size is always exactly the height of the table header.

and the initial window size I'm getting is small enough that the
variable list is also hidden, but at least when I maximise the window
that pane becomes visible.

>> - If I drag out the two graphing panes, I get to a pane that has a
>> Voltage VS Current tab - not sure what that is supposed to offer? Seems
>> any IV plots can already be done in the current graphs...
>>
> That was added by Dong when he was working on getting the
> electrophysiology layout suggested by Peter Hunter into PCEnv. The
> electrophysiology layout features will need to get looked into further
> for 0.2.
>> - is there a better way to choose variables for graphing? In a
>> reasonably complex model it is a real pain to scroll up and down the
>> component list and then choosing the right variable. Something where I
>> could start typing in variable names (ID's?) and have a completion list
>> of matching variables pop-up would be useful for me.
>>
> That could be tricky to get right without making things harder for new
> users, but I can look into that for 0.2. There are several things to
> consider here:
> 1) Do users know what component the variable is in? Is it common to want
> to find a variable without knowing the component? (I'm not sure it is,
> so perhaps we should still require users to find the component first. We
> could always provide a search for variable function, but I don't think
> that should be in the graph key. We could allow components from other
> parts of PCEnv to be dragged and dropped onto the key).
> 2) Are there a large number of components, or a few components with many
> variables? Good modelling practice would probably suggest that you
> should have many components, each with a limited number of variables,
> although users might not always follow this pattern.

ok - we can discuss this more when setting out the milestones for future
releases.

>> - and it would still be nice to have some other way to specify the
>> number of points in a graph than the point density.
>>
> I think this should definitely wait until after 0.1, as it is a
> significant change that goes all the way into the CIS.

sure.

>> - and a very minor issue, possibly due to the use of the max subscript
>> on the point density text label, all the labels (and units) in the
>> integrator settings pane are not quite aligned with the text boxes...
>>
> I think it is top-aligned, rather than bottom-aligned or centred (and
> the width of the text-box is greater than the width of text). What part
> of the label were you expecting to align with what part of the text-box?
> Or were you expecting the font-size of the label to be slightly
> increased so that the label would be the same size as the text-box? (the
> problem with this is it depends a lot on the system it is displayed on).

It'd just look nicer (to me) if it was all vertically aligned in the
middle of each row. The font size is fine.

---

• [cellml-discussion] PCEnv 0.1rc3 (release candidate for PCEnv 1.0) out, Andrew Miller, 11/16/2006
  • [cellml-discussion] PCEnv 0.1rc3 (release candidate for PCEnv 1.0)out, David Nickerson, 11/22/2006
    • [cellml-discussion] PCEnv 0.1rc3 (release candidate for PCEnv 1.0)out, Andrew Miller, 11/22/2006
      • [cellml-discussion] PCEnv 0.1rc3 (release candidate for PCEnv1.0)out, David Nickerson, 11/22/2006
  • [cellml-discussion] PCEnv 0.1rc3 (release candidate for PCEnv 1.0)out, David Nickerson, 11/22/2006
    • [cellml-discussion] PCEnv 0.1rc3 (release candidate for PCEnv 1.0)out, Andrew Miller, 11/22/2006
      • [cellml-discussion] PCEnv 0.1rc3 (release candidate for PCEnv1.0)out, David Nickerson, 11/22/2006

---

Archive powered by MHonArc 2.6.18.