DLL callback to ruby

Dan Rathbun

Dan, one last question, can the dll be compiled in 64bit or does it have to be 32bit like SK?

I would say 32bit, because the Ruby that Sketchup loads is also 32bit. There is a 64bit Ruby edition, but it may be in the 1.9.x branch, which Sketchup cannot load even 1.9.x 32bit.

@alienizer said:

Does the handle return by rb_define_module('Sketchup') correspond to the ISkpApplication interface?

I would say "not exactly". There may be some C++ methods from that interface, that are "Ruby wrapped" into the Sketchup module, as well as others from different C++ objects.

Getting a handle on the application object (and from there it's window object, and then it's UI element objects, and so on...) are methods that the current Ruby Sketchup module lacks. You'll have to use a C++ call, as shown in the SDK examples to get the app handle.

Alienizer

Thank Dan. So far, everything works great, I even got this to work...

modUI = rb_define_module('UI');
menu = rb_funcall(modUI, rb_intern('menu'), 0, nil);

but I'm stuck on this one...

rb_funcall(menu, rb_intern('add_item'), 1, rb_str_new2('File/Testing'));

It tells me "tried to create Proc object without a block"

Where do you learn all this? I can't find any docs on rb_intern or what to pass to it. I only guess!

Dan Rathbun

The API add_item method takes 2 args, a String and a Block, ... not 1 arg.

Alienizer

@dan rathbun said:

The API add_item method takes 2 args, a String and a Block, ... not 1 arg.

Indeed! I figured it out, and the rb_intern('menu') takes one, a str of the submenu name, like "Plugins" right?

When I do rb_intern('add_submenu') with 1 arg of str "Testing", it adds only "T" to the Pluigns menu, the first letter only!?

If I don't use rb_str_new2 SK crash, Am I doing it right?

For the Block, do I have to use rb_something to pass the Block, or just the reference pointer?

Sorry to be a pain in the......

Dan Rathbun

@alienizer said:

I can't find any docs on rb_intern or what to pass to it.

At the bottom of the "Extending Ruby" chapter, of the "Pick-Axe" book:
@unknownuser said:

*ID rb_intern(char name")

Returns an ID for a given name. If the name does not exist, a symbol table entry will be created for it.

I just pasted an example that used the rb_intern to create an ID... not sure if it's really needed. You can try just passing a normal string, as the method name argument.
(Using rb_intern on the C-side, is the same as using a Symbol on the Ruby side, thus:
:test_item or "test_item".intern or "test_item".to_sym

What are the arguments, for ruby C-side functions, generally?

The list at the bottom of the "Extending Ruby" chapter, is not complete, AND that edition was written for Ruby circa 1.6.x (so it's out of date.)

For a definitive list, refer to to the function prototypes, in the Ruby C source folder:
%(#000000)[**.../include/ruby-*<version>*/ruby/ruby.h**]
.. so if your using Ruby ver 1.8.6, the path would be:
%(#000000)[**.../include/ruby-1.8.6/ruby/ruby.h**]

ADD: .. and other header files as well.

Dan Rathbun

Yea, you need to use rb_str_new2 to convert the Cstring to a Ruby String.

Alienizer

Thanks Dan

Alienizer

Dan, what is the Block expected in rb_funcall(menu, rb_intern('add_submenu')...

I keep getting the error "tried to create Proc object without a block" if I use 1 arg.

If I use 2 arg, I get "wrong number of arguments (2 for 1)" but even you said it needs 2!? Confusing

Dan Rathbun

@alienizer said:

For the Block, do I have to use rb_something to pass the Block, or just the reference pointer?

Not sure... the proc and block functions are listed in intern.h

Depends on if the block / proc will be defined at runtime on the Ruby side, or before hand in your C code.

This generated doxygen website is old (v 1.8.4) but you may find it easier to find things:
http://www.ruby-doc.org/doxygen/current/globals.html#index_r

Alienizer

Thanks Dan

Dan Rathbun

@alienizer said:

Thanks Dan

no problemo ...

Also, if you have the Ruby Reference CHM, when you wish to see the C-side source for a method, you can click within the method's definition area, and a source box will unfold.

Alienizer

Sweet! But only ruby18-core.chm works out of the 3! Or is it me?

I just posted http://forums.sketchucation.com/viewtopic.php?f=180&t=39119 another challange for you

Dan Rathbun

@alienizer said:

Sweet! But only ruby18-core.chm works out of the 3! Or is it me?

ruby18.chm is a wrapper that loads both of the other two, as expandable nodes in the tree control.
CHM files use a MSIE stub named hh.exe
They a compiled HTML Help Markup files.
You can get the HTML Help Workshop app on MS Download.

Alienizer

Yes indeed, I didn't know the other 2 were loaded into the one. That's a nice ref book! Thanks Dan.

Dan Rathbun

@alienizer said:

Yes indeed, I didn't know the other 2 were loaded into the one. That's a nice ref book! Thanks Dan.

Well it's supplied with the mingGW compiled Windows Ruby one-click install packages. So I'm not responsible.

Check your Private Messages

Alienizer

@dan rathbun said:

Well it's supplied with the mingGW compiled Windows Ruby one-click install packages. So I'm not responsible.

I know, but I should have figured it out myself

driven

why would you use

Sketchup.send_action(JcB::My_Ext.initalise)

when

JcB::My_Ext.initalise

works for a loaded file?

john

Marthmatica

looks like you solved the problem.

If I want to call an rb function, from a loaded rb in SU, you could just send this function from a dll:

Sketchup.send_action my_rb_function_def_name_in_an_rb_loaded_file (no string wrap), as you can other SU commands

which rb SU command, and its function call string is just a string in a class property or in some dll function, thus called in either direction, from dll, or to dll.

Its a little clunky I notice, a bit unreliable after the stack is toasted a little.

Marthmatica

Because I didn't know. Thanks, that'll be better. Any other tips for reliably sending strings and data from DLL's easily?

Is there anyway to extend UI webdialog to contain activex controls? I have not tried it yet, I suppose it is doable. The dilemma is sending external commands to SU, in my case, and that work around is running script functions that have to be loaded, for now.

Also, if you know the browser engine for the UI webdialog, can it be set to a specific browser in SU?

@unknownuser said:

WebDialog.newSketchUp 6.0+

The new method is used to create a new webdialog.

Note that the browser which is embedded inside the dialog depends on the user's OS. On Mac, Safari is embedded, while on the PC whatever version of Internet Explorer is installed will be embedded.

I don't have explorer, thank God, but some engine is being used in SU, just wondering where that may be set.

Just trying to make some easy connectible user dialogs, and experimenting for fun. May be I'll find a hole or two as well, an accidental "in", that's always nice.

Was also wonderign why "creatObject" from the Sketchup encapsulated object model is impossible, I assume, maybe it is not programmed to be accessible. But the exe. exposes the object model, but not many useful methods, eyt I can instantiate the SU object in ActiveX anyways, but I'll keep chipping away the stone. There is more than one hole, and they'll emerge up in time, but I only have a million years here.

Thanks

Marthmatica

@driven said:

why would you use

Sketchup.send_action(JcB::My_Ext.initalise)

when

JcB::My_Ext.initalise

works for a loaded file?

john

Oh yeah. Forgot. The reason I was doing this, is to send the send_action "string" (or whatever it really is), from a DLL. And it works sometimes, soemtimes not, and so I was trying to get it to be reliable.

I don't know how to construct a "this.that" thing from a DLL, if that is required to run a ruby command. I just needed to keep the string as simple as possible. Still experimenting. I just need to call SU commands, from outside of SU data forms, even if initiated in SU, as that is all I can do at present. The reason is as usual, the data aids a real IDE may provide, versus naked scripts having to read text and Excel sheets as the "middleman". Just tryign to minimize the back and forth kind of supposed solutions I Frankenstein together.

I really don't know why SU doesn't just expose the whole OM for use by other IDEs. May be it does, but the stupid gene has yet to be naturally selected with myself, and so stupid is what stupid does, take a million years to find a four line solution that works. But, I got a million years, so, might as well do something "extraciricular" while I am here.