On 11/26/2020 5:42 PM, Alan Cox wrote:

It's a sample.  Samples need to be simple.

What it demonstrates is that you can have a Python-based environment
that doesn't have to be enormously more complex to use than OpenSCAD is.

What does "truly programmatic" mean?  Conditionals?  Loops?  Functions? 
Procedures?  OpenSCAD has those, just in slightly unfamiliar forms. 
Variables?  Not exactly, but close enough that many people go a long
time before they notice the difference.

Here's something that demonstrates how you might integrate some
full-featured-language behavior into that program.  It retrieves the
parameters from a JSON file.

# Curtain rod screw base
#
#            |d4|
#           |<d3>|
#          |<-d2->|
#
#  --      |      |
#  h3      |      |
#  --      ||    ||
#  |       ||    ||
#  h2      ||    ||
#  |       ||    ||
#  --  ================
#  h1  ================
#  --  ================
#
#      |<----d1---->|

import json

params = json.load(open('CurtainRodScrewBase2.json'))

d1 = params['d1']
d2 = params['d2']
d3 = params['d3']
d4 = params['d4']
nub = params['nub']

h1 = params['h1']
h2 = params['h2']
h3 = params['h3']
h23 = h2 + h3
h123 = h1 + h2 + h3
h12 = h1 + h2

Difference(
    Union(
        Cylinder(d1/2, h1),
        Cylinder(d2/2, h123)
        # I don't have a trivial way to do a triangle yet
        # cylinder(d2/2+nub, h123)
    ),
    Cylinder(d4/2, h123*3, True),
    Cylinder(d3/2, h3*3).translate([0,0,h2])
).add()

It still looks a lot like OpenSCAD, in terms of complexity.

Sure doesn't look like it so far.  The other thing that comes
immediately to mind would be, in a model that I recently did to design a
waste basket to be made out of wood, totaling up the weight of all of
the wood involved.  That was one of the major reasons that I did the
model; in OpenSCAD I had to "echo" out the various weights and add them
by hand.  BOM applications also come to mind.  Again, that would be very
small additions to the basic modeling task.

Um... who cares?  I mean, I call OpenSCAD files "programs", and you
don't, but what difference does it make?  They both generate geometry,
and they can be of comparable complexity to write.

That's true if you spend the majority of your "render" time deriving
your CSG tree.  Except for the use/include problems in OpenSCAD, it
seems that few people do.  They spend all of their time in turning the
CSG tree into a mesh.

If your alternative-language environment generates meshes immediately,
then yeah, it can't easily get parallelism and caching.  But it doesn't
have to do that.  My prototype generates a CSG tree.  If it did mesh
generation internally, it would be able to do all sorts of good
parallelism and deduplication.  Even as it is, farming the mesh
generation out to an external program, that external program could
easily be parallel and could, with a bit of difficulty, deduplicate.

Why do you say that?  I rigged up an IDE similar to OpenSCAD based on
Carsten's AngelCAD infrastructure, where I type Python and punch a
button and out comes a rendered model in a 3D viewer.  There's no reason
why that couldn't be done in a UI that looks exactly like OpenSCAD's.

Well, we don't want that.  At least, I don't.  :)

On 11/27/20 11:12 AM, Alan Cox wrote:

On 11/27/2020 9:12 AM, Alan Cox wrote:

You must have looked at different Blender docs than I did.

I looked at them, because the idea of a Python plug-in to a tool that
also allowed interactive editing was fascinating.

I don't remember whether I ever even figured out how to drop a cube into
the scene.  I do remember quite clearly that doing so was enormously
more complex than "cube(10);".  It was a total deal killer.  There was
absolutely no way I was going to move to an environment where basic
operations were so difficult.  Maybe I just found entirely the wrong
docs to look at; if you can point me at something simpler please do.

It seems like maybe this is another case where there are disconnects
between what different people think the goals of moving to a different
language are.

It seems like the first thing that comes to your mind is the ability to
manipulate the mesh.

While that would be useful to me - in particular, to extract the
bounding box of an object - it's not the top of my mind.  The top thing
on my mind is familiarity - I don't want to have to remember yet another
subtly different syntax with subtly different semantics.  I keep five or
more programming languages "hot" in my head all the time; dropping one
or more would be nice.  Next would be data structures, so that I could
express complex parameter sets in a more structured way than nested
arrays.  After that would be the ability to have procedures that
generate geometry also return metadata about that geometry - dimensions,
attachment points, et cetera.

Those things could layer pretty easily over an OpenSCAD-like
infrastructure, if it was based on a standard full-featured language.

Blender is HARD, with a tough, steep, and long learning curve. Wish I
was fluent with it.  <sigh>

On 11/27/20 3:06 PM, Jordan Brown wrote:

JordanBrown wrote

No, I was met with, "there's no reason for libraries, everybody should just
do stuff directly like I do", "making libraries is too much work, nobody
should do it", "we shouldn't waste effort on doing stuff to support
libraries in OpenSCAD because it's pointless/unnecessary (and might take
away from useful additions)"

There were a few people who responded with "yeah, libraries" and they were
basically shut down by the anti-library crowd.      It was a case of "I
don't need libraries...do what you want", which seems to be your position.
It seemed like active hostility to the idea of libraries.

There are two kinds of libraries for OpenSCAD:  ones that supply new
geometric functionality (such as a sweep function) and ones that supply
specific objects like screws, gears, joint mechanisms, and so on.  I think
the former are more likely to be broadly useful than the latter.  I also
think that the way one designs a model is determined by the available tools.
And possibly even what models one sees as possible may be influenced by the
available tools.

--
Sent from: http://forum.openscad.org/

Alan Cox-2 wrote

Note again, I'm just commenting on OpenSCAD, not taking a position on Python
or something.

I still really don't get your point.

Note random() already exists in OpenSCAD.  I've even used it and nothing
blew up.  It worked fine.

https://en.wikibooks.org/wiki/OpenSCAD_User_Manual/Mathematical_Functions#rands

--
Sent from: http://forum.openscad.org/

On 27 Nov 2020 at 14:57, adrianv wrote:

I see a third kind of library... one that makes complex things more easily
implemented. My current favourite library is BOSL2, which has new
functionality, supplies specific objects, but most important for me, as a user
of intermediate skill, makes things easier to specify.

BOSL2 offers, for example,  things like cubes (called cuboids), that include
the ability to apply chamfers or rounding on any edges, using easily
understood parameters, like TOP, FRONT, RIGHT+REAR, etc.

These are the sorts of things that really let me do things more intuitively and
far quicker.

Sure, those sorts of things are perhaps seen as syntactic sugar and
unnecessary by folks that are far more skilled than I, but for me, they are a
wonderful addition to a very capable program.

On Fri, Nov 27, 2020 at 4:38 PM David ainut@hiwaay.net wrote:

I first fiddled with Blender in 2010.  I wanted to edit a bracket for
a Mendel printer to remove 3 holes from the STL.  I spent 2.5 hours
trying to figure out how to use Blender, how to navigate in Blender,
or how to do any single damn thing in Blender.  It was the most
non-intuitive piece of software I had ever encountered (I later found
out that using Blender from a laptop without a numeric keypad makes it
much harder to use).  At the point I gave up, I went back to
Thingiverse and grabbed the DXF file and the OpenSCAD script that made
the STL in the first place, and in seconds figured out what the
program was doing and commented out three lines that made the holes I
wanted to not be there.  F6 and export and I was done!  I have been
a happy user of OpenSCAD ever since.  Blender still pisses me off (the
last time I worked with it was to orchestrate some animation to,
effectively, drive a pre-rendered car around a virtual racetrack - it
did that fine, but it was animation, not object creation, in the
Python I was hacking).

With that, from what I'm reading in this thread, it sounds like a case
of "OpenSCAD isn't a procedural language and I don't like that" that
comes up pretty much every time someone wants to fiddle with variables
inside loops.  I've been programming (professionally) for decades.
OpenSCAD is the only Functional Language that I've learned and, yes,
it was a bit of an adjustment coming from C and Perl, etc., but it's
not like OpenSCAD is the only Functional Language ever created.  It's
a different category and in the specific case of twiddling things
inside and outside loops, it differs from a Procedural Language.

I'm not in favor of $SUBJECT.  I've been working in OpenSCAD for over
10 years at this point and I don't see Functional vs Procedural as
something that needs to be "fixed".  If someone wants to go make
Python into an object creation platform, that sounds like a different
project.

-ethan

I'm in your court, Larry!

David

On 11/27/20 4:50 PM, lar3ry@sasktel.net wrote:

Thing is, I’m not just trying to update a variable in a loop. I can do that awkwardly with C-style for().

I’m writing a library to do the hard things like make solids built from Bezier patches, or calculate path offsets, or calculate Voronoi regions so that I can do more complicated geometry. And I cannot feasibly implement some of the needed efficient algorithms because the language cannot make the data structures without being extremely inefficient.

If OpenSCAD even had a mutable dictionary type, I could make most the needed structures, but it doesn’t. And variables aren’t truly mutable.

And saying that that is not what OpenSCAD is designed for, sounds to me like saying “OpenSCAD is just a toy, and you shouldn’t try making it better.”

-Revar