Posts made by l.frisken

ah mozzie,
It's actually not too OT at all
There are a number of known problems with SU2LUX's settings atm,
and I ran into this problem when trying to fix them.
If you want I can share what I have so far, and it should work
fine on SU7 and perhaps SU8 if the problem is restricted to
my machine but I suspect it isn't... (not that there is too much
to be desired in SU8 free anyway...)

@khai said:

take a look at Luxrender. they are going to implement OpenCL GPU rendering in .8

they already have some interesting work done you can download and try out, and the SU plugin, while rough, does work.

That's what I would have suggested also

SmallLuxGPU is a good working example, and as far as I'm aware OpenCL is just essentially a layer on top of CUDA (at least in Nvidia's Implementation) so you should gain many of the same benefits.

You might need to dig around a tiny bit in the forums to get at it though.

@morisdov said:

If i disable google update service and hide associated exe, i am experiencing noticable delay on closing sketchup or opening a new model.

First i thought i had AppObservr bugs.
Even though i try to code defensively against exceptions and presume surprises.

Yes! That's probably it. I'll check before I confirm that was my problem, but if it is then this is definitely google's responsibility. I do normally close the update service because I don't like to have random processes running in the background on my computer... But this definitely shouldn't affect sketchup usage!

would fxruby or qtruby be other options?

@frederik said:

@11zulu said:

I have seen no mention of :

Guess you didn't make a search then...
You can see it mentioned in the Extensions & Applications Discussions section in the Mitsuba Physically Based Render...

Reason why you don't see it in this list is most likely because there's no exporters developed between Mitsuba render and SU...

I'm pretty sure mitsuba extensively uses the collada format for its scene intake, which I would call support for sketchup, wouldn't you? Plus it's fast becoming popular.

@tig said:

It takes a blink of an eye to get / to set literally dozens of these settings as the web-dialog opens / closes. You could also make the dialog 'modal' so that the user can't close the model with the web-dialog open and thereby loose settings - I have done this because with my tool you can do several operations with the dialog remaining open and it's only on a 'close' that the current 'fields' get saved as attributes...

So what happens when the user tries to render/export with the settings dialog still open? I would like it to be possible not to have to close the dialog every time you want to render the scene, so it will be easy to make quick changes... I'm guessing you've probably handled it similar to me in that you have a temporary attribute dictionary, that gets save synced to sketchup on demand. So that it should be possible to export your scene whether the dialog is open or not... Either way, does implementing the "syncing" action in the deactivate(view) prevent it from being registered in the undo stack? Your solution to the problem does sound like it could work for me. Could I not also do it like like this? WebDialog.set_on_close{ save_settings(blabla)}

If I do it the way you suggested at least the user will will have the option to undo the changes, I just worry about it getting in the way of the modelling workflow, because it essentially creates a point where the user can no longer undo changes they made in the model. There really should be two undo stacks running in parallel ?

@thomthom said:

But isn't it just as likely the user might want to undo these setting changes?

yes, but unfortunately, they can currently undo changes they made in the settings editor even when it's closed. They won't be able to tell that they've undone them. Perhaps, like me they might think they just mispressed the ctrl-z key, or just keep pressing it until something visual undoes (like the removal of a line), and later discover that they've just undone all the changes that they made to the settings, and there is no way to redo, because they've continued on with their work.

This is entirely plausible because I've done it myself! While it isn't too much of a loss at the moment, I can see it getting pretty bad later on, plus, any little change like ticking a box or whatever adds another undo to the stack, which means that when alternating between modelling and changing render settings, they have to delete the changes they made in the settings editor AND press undo many times just to get a visual response... (hmm I could have made that bit shorter )

@unknownuser said:

Have you tried that bare-bone example with no other plugins installed?

yes, in both sketchup 7 and 8, but I'm thinking I'd better try mucking around some more to see whether it's another problem with my sketchup installation, because it doesn't seem to be giving you guys any problems. But you are right, because it could be an observer in another script interfering with this one or something. Oh well, unfortunately the solution isn't that simple

Thanks for the help anyway thomthom (and TIG of course ). Hopefully I can get a reasonably feature complete su2lux out soon for people here to test. I'm going to be doing a bit of Arch-Vis work over the coming holidays, so I have a vested interest in this project!

edit:
oops, didn't see that last post TIG, I'd better reply to that too!

Thanks for the replies guys!

@thomthom said:

eh? he doesn't mention anything about any tool...

thomthom is right, it's not a "Tool", its an exporter for sketchup to luxrender.

Previously, settings were saved to the sketchup attribute dictionary when any change was made to the setting in the settings editor. This system does work, but it also means that after changing settings, or perhaps using/loading a preset it leaves stuff behind in the undo stack, which the user might inadvertently undo and thus lose any changes they made to the settings. I've changed it save all the settings to a temporary array until the model is saved. Which works fine in sketchup 7, but sketchup 8 seems to be having problems.

@thomthom said:

That bare-bone sample you provided cause a 5min lag on your computer?
What OS?

Not 5 minutes, but 5 seconds rather. I might do a little more testing to further narrow the problem. But for someone who likes to save fairly regularly, 5 extra seconds is kind of annoying, and I'm imagining it might annoy some users as well. The same lag occurs when opening a new model with these observers activated.
BTW: I'm using windows xp sp 2.

@thomthom said:

AFIK there isn't a way to make an operation fully transparent to the undo stack.
What type of data is it you want to store anyway?

I'm just trying to store a mapping of settings to their values.

I've run into some trouble with sketchup's application observer and model observer when using the onSaveModel, onNewModel and onOpenModel. Basically, whenever I save, open or create a new model it lags like crazy... ok, maybe not crazy, but 5 seconds is enough to get on my nerves, and I'm sure that's pretty unacceptable for any script, as it will most likely get in the way of the user... I've stripped down the problem to this code and it still occurs, so I'm thinking it probably isn't fault of mine, but if you think it is, please let me know

Interestingly enough, my code runs without a hitch in sketchup 7, so I'm thinking its a new problem. I'm surprised no-one else has run into this problem yet, or does no-one use app/model observers?! or perhaps it's just my computer? (yes I have reinstalled sketchup 8 just to make sure, and yes I have tried with a totally clean script directory.)

Any help here would be greatly appreciated, because I'm relying on this to fix/hack around another problem I'm having:

Is there any way to register an operation that will be invisible to the user? I havn't found a way, and I would like to save all my script's properties into sketchup's attribute dictionary in one hit without having to worry that the user might accidentally undo the saving of the properties. I finally decided that the only way to get around this would be to save the properties when the current model is saved.

class My_model_observer < Sketchup;;ModelObserver

  def onSaveModel(model)

    puts "SAVING MODEL"

  end

end



class MY_app_observer < Sketchup;;AppObserver

	def onNewModel(model)

    puts "NEW MODEL"

	end



	def onOpenModel(model)

    puts "OPENING MODEL"

	end

end



Sketchup.add_observer(My_app_observer.new)

Sketchup.active_model.add_observer(My_model_observer.new)

To the guy asking for sketchup -> luxrender... I'll continue work on it after finish my last high school exams (ever, yeah!)

Hey John,
Just want to say thanks for replying to some of these questions. I personally find the fact that you have bothered to take the time to reply to us much more satisfying than any new sketchup features. The fact that the sketchup team is willing to openly converse with its users is a big thing I guess, and not something I was really expecting given its track record of versions in the past few years (which is why many complain). Now I can be put to rest and just accept that sketchup will probably never do some of the things I would like it to do, and I mean that in the best way possible. I can plan ahead and make sure that I have other tools on hand to fill the gaps around sketchup. There are of course many simple core problems in sketchup, that, I'm sure will get solved given enough time, but I'm no longer left in the dark half hoping that they will get fixed soon.

Thanks

I don't think blender will be the answer to people's problem here, or at least not the one they are expecting or looking for. But for all their "arrogant" community, one has to admit that they are really reaping almost all the possible benefits to open source development. I often make it my tool of choice apart from architectural related projects due to the fact that it's development doesn't look like stopping any time soon, in fact things seen to be picking up pace as more people join in the fun. Blender, in it's core is alive, sketchup on the other hand is stagnating. There are so many small and extremely annoying issues with sketchup that have not been fixed for a very long time. That, along with the fact in the last 2-3 years, very little has changed, and there is no hopeful, public, general plan for the future practically prove the point for all to see. It's getting ridiculous. All the major 3d packages seem provide at least some future development info to get excited about. I didn't make this post to promote blender itself (which is certainly not for everyone), but rather to make an interesting contrast.

how often i find myself wishing that sketchup was like blender! Not the program itself, but the open source, enthusiastic and fast paced development community. Even if bmesh fails to fulfill my architectural modeling needs, sketchup and blender would still compliment one another nicely as sibling projects don't you think!? each filling their own niche.

why don't we just start discussing Lua while we're at it?

no, but seriously
Python is just a different language. I learned to use it before ruby, but then I discovered ruby, which I prefer for many reasons other than it's speed, which as mentioned is slow. Yes compulsory indentation is nice in some respects, but it is a complete and utter pain when mixing tabs with spaces (can't see where the problem is). Ruby allows for more creative formatting of the code (which can be very nice). I like them both, and don't we all wish the best of both could be combined?

mate,
Not to spoil your enthusiasm, but that is a bit of a lame post!
Ok, now i've got that out, i'll attempt to help. Just google something along the lines of 'sketchup render' and you will probably find an exhaustive list of results. Just to name a few of the free ones for you: kerkythea, pov-ray and perhaps luxrender in the near future (probably not really ready for inexperienced users yet, but you could give it a shot if you feel your up to it). If you have trouble installing a script, just be specific, and someone can help.

Thanks for the replies guys!

@unknownuser said:

That is correct. Most methods of classes are instance methods. You should be able to call them from inside the class (ie another method of the same class,) using it as a private method.

Yes it is an instance method, but it seems to have no effect on objects created in the sketchup ui, presumably because they have their own read-only/inaccessible templates/classe for creating objects?

And yet I still fail to see why rewriting the instance variable declaration (if that's what it's called!) for x, y, z has an effect on yaml. Have they not already been declared? I just have a funny feeling that something weird is going on under the hood.

I actually had no real purpose when extending those classes, except to try and get them to expose x, y, z values to yaml. It worked in that instnce, I have no Idea how, and was unable to do anything similar for other types. The workaround only partially worked (ie, it does not work on objects created in the sketchup ui).

@unknownuser said:

You can do anything you want FOR YOUR OWN USE.

BUT Google OWNS the Sketchup, UI, and Geom Namespaces. The Terms of use prevent you from causing other people's installations of Sketchup to be degraded or broken or such. The license agreement is at: http://sketchup.google.com/intl/en/download/license.html

At the very least, if you wanted to extend Sketchup classes (and release those extensions,) you'd need to have Google Sketchup Team consent AND Sketchup Developer Community approval.
You might join the SKX project, or monitor the SKX forum.

Normally I would have given up trying to get yaml to work and come up with my own alternative way of transmitting the object information via text through stdio, but I have thought of another interesting concept, the possibility of using other languages such as python or java with sketchup... with yaml. If there is no interest in that, and I can't find a quick fix to this problem I have with yaml, then I'll drop that Idea!

I think I need to learn how yaml works (on the inside) in order to find out why it struggles with sketchup.

I did have one Idea for a work around, and that was to deep copy the objects created in sketchup to what I have dubbed (if my theory is correct) a "True Ruby Object" and then use the .extend method to append the yaml methods on the objects. In order to do this I guess it would be easiest to make the change to the way yaml works (otherwise I may as well just copy all the attributes manually!).

By having my own copy of the objects, and then using the .extend method will I still be voiding this license aggreement?

A big problem I think I can see with that Idea is performance. Having a copy of every object doesn't seem like good, common sense to me!

Disclaimer: I'm not that experienced at ruby programming, so please excuse me if I've missed something obvious!

At the moment I'm having a play at getting a form of RPC going between 2 sketchup processes (over the network). But before I can do that I've had to face a number of problems. one being the lack of proper sockets in the embedded interpreter (i've gotten around that now by using an external process), the second being the challenge of getting YAML basics working in sketchup (my chosen form of text based serialisation). But now I think I've hit a brick wall:

As far as I can guess, yaml subclasses/extends every object and adds an automatic to_yaml method (please correct me if I'm wrong here!), that tries it's best to make sense of the objects it deals with. I got it working fine on my own simple classes/object hierarchies, but yaml seemed to draw the line at sketchup!

After much mucking about with the sketchup classes (such as face, vertex, and so on...), I found that if I did something like this:

class Sketchup;;Geom;;Point3d
  def testing
    puts "Yay the test worked"
  end
end

or this

class MyPoint < Sketchup;;Geom;;Point3d
  def testing
    puts "Yay the test worked"
  end
end

I could only get the "testing" method to work when i called it from an instance of the class (that I created) myself:

pt = Sketchup;;Geom;;Point3d.new(1, 2, 3)

or

pt = MyPoint.new(1, 2, 3)

(they both seem to be the same)

anyway, to cut a long story short I finally cracked the code on how to get yaml to work with sketchup objects (so I thought):

class Sketchup;;Geom;;Point3d
def initialize(x, y, z)
super()
@x = x
@y = y
@z = z
end
end

It seemed as though the Point3d class did not even have instance variables (weird!), I can only assume that this is because they were created from the closed side of Sketchup.

but as I said before, this only worked on objects that I created myself, not even new one created using sketchup tools. Now I'm not blaming this all on Sketchup, because I'm sure they have their reasons for preventing (or not inventing) this kind of interoperability, but If anyone can see a way for this to work (can prove me wrong please!) , or a workaround please reply.

I'm currently having similar problems after I updated the graphics driver on my laptop: ati radeon xpress series. Whenever I select a face, it is completely blue, instead of blue dots. this does not occur when I turn hardware rendering off.

The same problem and worse happens to a class I was teaching with nvidia cards in their laptops! They were also having trouble selecting faces when they were at an angle to the viewport.

I personally think it would be great to try and emulate the interface that Lux has with blender (and yes I know that this has nothing to do with the actual exporter code). It is not quite a simple as skindigo, but I think its functions are more clearly laid out. If something as good as this could be implemented in sketchup (and as nice looking) it would be a first.

I recently found a post on another forum with photo birdstrikes to planes. It's amazing, but also freaky, because it could happen to anyone I guess! poor birds http://www.mxpairport.it/forum/viewtopic.php?f=29&t=3513