My framework can be turned into a PyCharm plugin with code completion/viewing/refactoring of the parameter hierarchy and Part library management. It can be extended to document parameters visually as annotations over renderings of each Part. It can already generate complex extrusions, can handle Bezier paths and some more - but none of that will be published soon if I don't get help. I'm still working out the details of how to handle aliases (happens a lot that different parts with similar shape want to name their parameters differently), orientation/alignment/attaching parts in an assembly. I contemplate getting rid of the openscad translate/rotate functions and replace them with versions which are stored on each part, that would allow automated kinematic chain calculations. There's a lot of work which would radically change what I currently have, and I don't have the motivation to do all that alone in a short term... I will eventually do it for my own sake, but gradually as I need it for my own modeling. Cheers, Csaba On Tue, 2021-12-14 at 18:48 +0100, edmund ronald wrote: > Here is a Medium post where John Clark Craig describes his wrapper. > He seems to have a userbase of sorts, and to be a smart cookie so > there is probably a need for something like this, and I expect sooner > or later someone will make an integrated 3D system of their own > similar to this as the openScad crowd don't like the idea. > > Edmund > > https://jccraig.medium.com/yes-you-can-use-python-for-3d-design-and-printing-508b1791c863 > > > Edmund > > > On Tuesday, December 14, 2021, Csaba Nagy <ncsaba@javampire.com> > wrote: > > Hi David, > > > > at first I thought I was looking for the same, a python 3D > > programming > > language. So I started using SolidPython: > > > > https://github.com/SolidCode/SolidPython > > > > It is a fairly straightforward wrapper around OpenSCAD, allowing > > you to > > write your 3D model in python, which is then translated to > > OpenSCAD. > > But soon I realized that what I needed is a better way to express > > parameters of a complex model. > > > > So I wrote my own python framework to process JSON descriptors, > > with > > meta-class magic that merges the parameters across the inheritance > > hierarchy. That way I can define sub-classes which basically > > inherit > > the parameters of their base class, and can override just that > > single > > one they actually change. The framework messes with the > > setters/getters > > too so parameters defined in that decriptor are available with the > > attribute access syntax as well as dictionary entries (the "Part" > > class > > extends an ordered dictionary). Class instantiation can override > > any of > > the parameters, the ones not set are left at the default value. > > That > > overall makes for a hierarchical access of all parameters - in > > simple > > cases simplifies the code a lot, in complex ones makes for > > spaghetti- > > like names, but I think it is still the better way to handle 100+ > > variables. > > > > Working with PyCharm I also have some magic built in the Part class > > which allows me to render the current file and open it in OpenSCAD > > view > > with the press of a key-combination. Same can be done for any file > > in > > the project view, or for a class highlighted in the editor. That > > makes > > for fairly streamlined work-flow where coding and preview > > alternate. > > > > The catch with all this is that at the moment it is tailored for > > exactly what I needed to model and it is less then pre-alpha > > quality > > for public release. I also can't find the motivation to spend the > > time > > necessary to improve it for the sake of publishing, real life and > > my > > own modeling tasks are what finally win. > > > > I will include an OpenSCAD file and the top level python class > > which > > generated it for a sample of what can be done. I will not publish > > the > > code as I'm not satisfied with it and will surely radically change > > it > > still in the future - but if somebody is interested enough in it to > > help preparing a publishable version, I'm ready to do a few on-line > > sessions to explain what I have and discuss what and how to > > publish. > > Without help, it will take me a few more years with the speed I'm > > progressing now. > > > > The project is hosted on gitlab.com, at the moment private and it > > seems > > gitlab will not allow me to change that (bug I assume). Anybody > > wanting > > to check the code, ask me and I will give you access. > > > > Cheers, > > Csaba > > > > P.S. the scad file is a table-tennis ball machine I'm working on > > right > > now. All elements you see are fully configurable, almost no hard- > > coded > > number outside the CONFIG_DESCRIPTOR of each used Part/Assembly. > > The > > iter_components method in the python file will generate a list of > > 3D- > > printable components, processing is bound to a PyCharm key > > combination > > too. Any class name extending Part/Assembly in the python file can > > be > > highlighted and previewed using a key-press too (of course only in > > my > > local setup, I would need to package the key shortcuts in a PyCharm > > plugin). > > > > On Tue, 2021-12-14 at 10:14 -0600, David Schooley wrote: > > > Thank you everyone for the comments up to the point. Even the > > “no, > > > that’s not going to happen” type of feedback is useful. > > > > > > To put things in context: I was a contributor to gnuplot years > > ago. > > > (I did a really nice Mac version.) I’ve made a lot of 2D plots > > using > > > Perl-generated gnuplot scripts. After playing with flex and bison > > a > > > bit, I started thinking about a new plotting program using a more > > > modern interpreter. Python was still relatively new, I hadn’t > > used > > > Python at all at the time, but the syntax for this never-written > > > plotting software would have been very Python-like. I started > > using > > > Python a short time later, abandoned by software project, and now > > > things that would have been Perl+gnuplot are Python with > > matplotlib. > > > Life is good. > > > > > > When I started 3D printing, I started with FreeCAD (which has a > > > Python workbench) and then moved to OpenSCAD. I have a large > > number > > > of OpenSCAD scripts laying around. > > > > > > OpenSCAD meets my needs for almost everything I do. (It does fall > > > down when it comes to things such as fillets around complex > > shapes.) > > > There are times when I need real variables and there are times > > when > > > the design needs to change depending on the size or something > > else, > > > i.e., the hole pattern changes depending on the size of the > > object. > > > The next step would be data-driven designs, which I can visualize > > but > > > can’t really do at this point without using Python to create > > OpenSCAD > > > scripts. > > > > > > An example might be a 3D printed map using GIS data. You could > > > perhaps do this with the OpenSCAD language by extruding a bitmap > > > generated using Python. It would be cleaner, and you would have > > more > > > control, if you did the whole thing in Python. > > > > > > Thinking of OpenSCAD (the software) as a wrapper for OpenSCAD > > (the > > > language) is perfectly reasonable. It’s not the answer I wanted, > > but > > > it’s the answer I got. > > > > > > So that’s where I am at. I’ll take a look at the book that Harm > > > suggested. I think I can get the Kindle version for free. > > > > > > > On Dec 14, 2021, at 1:44 AM, edmund ronald < > > edmundronald@gmail.com> > > > > wrote: > > > > > > > > Everybody and their dog seems to have done Python covers of > > > > OpenSCAD. > > > > Which seems to indicate that there is a perceived need in the > > 3d > > > > Print userbase. > > > > > > > > My guess is sooner or later another project will mature enough > > to > > > > serve that population, at which point the question will go > > away. > > > > > > > > Edmund > > > > > > > > > > > > > > > > On Mon, Dec 13, 2021 at 11:34 AM <info@hjcreations.nl> wrote: > > > > > Hi David, > > > > > > > > > > Do you know this book : > > > > > Python for 3D Printing openSCAD by Craig, John Clark > > > > > > > > > > I got it and seems interesting, but had no time read it at > > all. > > > > > If you'reinterested I can give you more info. > > > > > > > > > > Regards, > > > > > > > > > > Harm Jeurink > > > > > > > > > > David Schooley schreef op 2021-12-12 10:43: > > > > > > I have a feeling this has come up before… > > > > > > > > > > > > I’m curious about any thoughts/previous discussions have > > > > > > > > > > happened > > > > > > regarding incorporating Python as an embedded scripting > > > > > > > > > > language to > > > > > > OpenSCAD. Some work has been done with Python and CSG, so > > this > > > > > > > > > > isn’t > > > > > > entirely impossible, but it could be a lot of work; I get > > that. > > > > > > > > > > > > The OpenSCAD language isn’t bad, but it’s no great shakes > > > > > > > > > > either. It’s > > > > > > interesting to see work happening to make structures a > > reality, > > > > > > > > > > but > > > > > > maybe it would be better to incorporate a language that has > > > > > > > > > > real > > > > > > variables and objects. There was a recent discussion around > > > > > > calculating parameters on the fly, which is easy with > > Python > > > > > > > > > > but is > > > > > > more difficult or impossible using the OpenSCAD language. > > > > > > > > > > > > I’m not at all interested in generating OpenSCAD code from > > > > > > > > > > Python that > > > > > > just adds another layer of “Why doesn’t this work?” I think > > > > > > > > > > being able > > > > > > to write some Python code that generates an object that > > > > > > > > > > OpenSCAD can > > > > > > display and turn into something I can 3D print or cut using > > a > > > > > > > > > > laser > > > > > > would be amazing. > > > > > > > > > > > > _______________________________________________ > > > > > > OpenSCAD mailing list > > > > > > To unsubscribe send an email to > > > > > > > > > > discuss-leave@lists.openscad.org > > > > > _______________________________________________ > > > > > OpenSCAD mailing list > > > > > To unsubscribe send an email to > > discuss-leave@lists.openscad.org > > > > > > > > _______________________________________________ > > > > OpenSCAD mailing list > > > > To unsubscribe send an email to > > discuss-leave@lists.openscad.org > > > > > > _______________________________________________ > > > OpenSCAD mailing list > > > To unsubscribe send an email to discuss-leave@lists.openscad.org > > _______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org

cu = Cube(size=45,center=True);
sp = Sphere(r=30)
cusp = cu.intersect(sp)

## union of 3 cylinders
cy = Cylinder(r=17,h=50,center=True)
cy2 = cy.union(cy.rotatex(90))
cy3 = cy2.union(cy.rotatey(90))

## difference of the 2 above
cusp.subtract(cy3).add()

I looked into this a bit a year or two ago, more from the perspective of wanting to reduce the number of languages that I keep in my head, and to make it easier for newcomers, than from any specific problem with OpenSCAD functionality.  Why reinvent the wheel, but a little differently? I got far enough to have a toolchain based on Carsten's AngelCAD environment.  You could type a Python program like cu = Cube(size=45,center=True); sp = Sphere(r=30) cusp = cu.intersect(sp) ## union of 3 cylinders cy = Cylinder(r=17,h=50,center=True) cy2 = cy.union(cy.rotatex(90)) cy3 = cy2.union(cy.rotatey(90)) ## difference of the 2 above cusp.subtract(cy3).add() into the IDE and my toolchain would turn that into xcsg that could then be viewed using Carsten's xcsg viewer - basically the same flow as Carsten's AngelScript flow.  No OpenSCAD in the process - not specifically to avoid it, but to directly emit a lower-level result. I stopped working on it for a couple of reasons: * I was offended by the need to explicitly add objects to the model (the .add() method above), and couldn't figure out how to get rid of it.  The minimum OpenSCAD program is "cube();" and I wanted to get down to that level of simplicity. * Python is not sandboxable.  Python programs can do anything.  You could not take somebody's Python-based SCAD program and run it, without risking that it would trash your system. Note also that designing in a "real language" doesn't automatically solve some of the OpenSCAD complaints.  In particular, just because you've got a real language and real objects doesn't mean that you've got access to the geometry.  My toolchain operated by generating an xcsg <https://github.com/arnholm/xcsg> file to be fed into a geometry engine.  It has no idea what, for instance, a sphere is; it just knows that it's a geometric object that can be generated and operated on by the geometry engine. It was really pretty easy to write - it's only about 300 lines of Python plus a one line batch file to glue it together.  I'd be happy to forward it to anybody who wants to play with it.

 solid  = Cube(100) + Sphere(50);

On 14.12.2021 19:44, Jordan Brown wrote: > I stopped working on it for a couple of reasons: > > * I was offended by the need to explicitly add objects to the model > (the .add() method above), and couldn't figure out how to get rid of > it.  The minimum OpenSCAD program is "cube();" and I wanted to get > down to that level of simplicity. You could probably use Python operator overloading https://www.pythontutorial.net/python-oop/python-operator-overloading/ solid = Cube(100) + Sphere(50); > My toolchain operated by generating an xcsg > <https://github.com/arnholm/xcsg> file to be fed into a geometry > engine.  It has no idea what, for instance, a sphere is; it just knows > that it's a geometric object that can be generated and operated on by > the geometry engine. Correct. This implies one could reimplement the engine to use CGAL or OpenCascade instead of Carve, Carsten Arnholm

cube();

cube()

On 12/16/2021 8:17 AM, Carsten Arnholm wrote: > On 14.12.2021 19:44, Jordan Brown wrote: >> I stopped working on it for a couple of reasons: >> >>   * I was offended by the need to explicitly add objects to the model >>     (the .add() method above), and couldn't figure out how to get rid of >>     it.  The minimum OpenSCAD program is "cube();" and I wanted to get >>     down to that level of simplicity. > > You could probably use Python operator overloading > https://www.pythontutorial.net/python-oop/python-operator-overloading/ > >     solid  = Cube(100) + Sphere(50); Yes, I could do that, but that's not what I was offended by. In OpenSCAD, you can say cube(); and that's a complete and valid program that yields a model of a cube. In Python, if cube() is a function that returns a cube, saying cube() will create a geometric object that's a cube... and then will drop it on the floor, because nothing consumes it. You have to have a method or function that consumes the cube and adds it to the model.  In my library, that's the .add() method, but it could equally well be an add(obj) function.  Or maybe the program is required to return the object (which also requires you to explicitly union the various components of the model, another thing that OpenSCAD doesn't require you to do explicitly).  In any event, the required minimum program is more than just c u b e ( ). I was hoping that I could find a hook akin to what the REPL <https://en.wikipedia.org/wiki/Read%E2%80%93eval%E2%80%93print_loop> does, so that if the result of a statement was a geometric object that would otherwise be dropped on the floor, it would instead be added to the model, in the same way that the REPL would print it, but I couldn't find such a hook. I also looked at JavaScript as the base for a 3D modeling environment, and ran into similar issues.  I think there was something else that I didn't like about JavaScript for the purpose, but I don't remember what.  Perhaps it was simply that command-line JavaScript interpreters are not as ubiquitous as command-line Python interpreters.  (Yes, I know about Node.js and use it extensively, but it's not ubiquitous.  Yes, every browser has a JavaScript interpreter, but you can't hook them into a toolchain, at least not without *also* creating an xscg processor that runs in a browser.)