Gary,
I was on an unstoppable roll (salvaged early on by Andreas' bitblt coaching),
until I needed to repeat a "complex" GUI component and wanted/insisted on doing
so with some reuse. Seaside gives us components on web pages; we need them for
GUI code too. I tried turning my presenter-like structures into factories that
would add morphs to a single shell, but it fell apart when it came time to set
the framing values. I was not particularly interested in fixing it, because if
I could do that, I would simply build proper composite widgets using the fix.
The problem appears to be in SystemWindow, which does some incredibly
complicated things, all of which (correct me if I am wrong) would be
unnecessary if only there were a good set of layout managers. I can get very
simple-minded about things, but it would make a whole lot more sense to me to
create rows and columns of widgets, adding splitters between them until the
whole things behaves as intended, rather than writing code (present only in a
top-level shell) that tries to split things up into rows using its own
judgment. Not good. I realize you did not create the situation, and have done
wonders to make it work far better than when you found it.
Some time ago, I built an emulated widget framework for Dolphin, which I needed
because the combination of Dolphin's view resources and Windows itself was too
slow for what I was trying to do. That project included a somewhat strange set
of layout classes, but the basics are present, and it is not Object Arts' IP.
I ported the classes to Pharo and did some work on stub View and Presenter
classes that I had added to Pharo largely to passify my code that I was
importing from Dolphin. The layouts think in terms of emulated widgets, and I
see no reason to change their minds: I might want to replicate the framework.
However, dynamic typing and a couple of extra methods allow them to work with
just about anything. My goals are modest. Being able to compose rows and
columns would do a lot for me. Add splitters and the ability to fix the size
of some items, and I could almost anything I would need.
View class>>example
| row column dot square out shell |
dot := ( Form dotOfSize:100 ) asFormOfDepth:32.
square := ( Form squareOfSize:100 ) asFormOfDepth:32.
out := Array writeStream.
row := ContainerView row.
2 timesRepeat:[
column := row addSubview:ContainerView column.
2 timesRepeat:[
out nextPut:(
column addSubview:ImageView new
).
].
].
out contents with:{ dot copy. square copy. square copy. dot copy. }
do:[ :view :form |
view morph image:form.
].
row rectangle:( 0...@0 extent:4...@400 ).
row layout.
shell := StandardWindow labelled:'Hello MVP'.
^shell addMorph:row morph frame:( 0...@0 extent:1...@1 ); yourself.
The above code produces an array of dots and squares, as intended. One quirk
is that the grid does not resize as the shell resizes, a consequence of my not
having hooked it up to resize events. I might get some interesting meltdowns
once I begin to do that =:0 I used your PanelMorph as the "view" associated
with ContainerView. What, ContainView isn't a view??? No. The code is biased
toward Morphic, but hopefully the same code should extend to wx, GTK, etc.
Dolphin's views have a handle instance variable to control the external
resource; these views have an instance variable pointing to their morph.
Handling of sub views works pretty much as in Dolphin: any view/morph can have
children, but adding them is "legal" only for composites.
Most systems I have seen treat coordinates relative to the parent/owner, but
not Morphic. I remember seeing plans to make the change, but nothing after
that. The view instances provide a natural place to fix things, so I took the
plunge. If we switch to some other graphical realization, we can simply remove
the the transformation.
Are the existing splitters as strange as SystemWindow? By that I mean, would
it be reasonable to add them between other morphs and look for events from
them, or will they have to be replaced? I will eventually need splitters, but
I could initially live without them if I can get reliable composition where I
need it.
There are very few view classes at present. MorphView can wrap almost
anything, so it might be better to create a rich set of presenters instead.
Dolphin's view resources are going to be interesting to replace. There are
some complexities that I suspect are in deference to Windows, and some that
might be avoidable. For our purposes, it might be enough to use SIXX to
serialize a bunch of message sends and gzip the results to save memory.
Another option might be to rely on class methods; having full closures won't
hurt; they might allow sufficient hooking that resources in the Dolphin sense
won't even be necessary. The desire for them quickly arises because views get
realized in places know nothing about how the views should be configured; at
least I think that is what happened to me after just a couple of hours. I am
far less interested in having a graphical view editor than I am in being able
to write **GOOD** code that assembles things as I want. If the result happens
to allow a graphical editor too, so much the better.
Any interest?
Bill
-----------------------------
View
MorphViev
ImageView
ContainerView
ViewGadgetLayoutAbstract
ScrollerGadgetLayout
NullViewGadgetLayout
FixedStretchFixedGadgetLayout
ProportionalGadgetLayout
PreferredExtentsGadgetLayout
GridGadgetRowsLayout
VerticalListLayout
_______________________________________________
Pharo-project mailing list
[email protected]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
