function, and reference the contents by o.a or o.b would be a big win.  It
would reduce the need to create constructor and access functions for every
aspect of a feature.

How can I help?

-Pat
Note that this reply is bottom-posted.
If using Gmail's web app, click the three dots immediately after hitting
reply, place cursor at bottom. Easy!

On 23.11.21 23:16, Patrick Callahan wrote:

Discussion about objects is here:
https://github.com/openscad/openscad/pull/3087
https://github.com/openscad/openscad/issues/3088

This is based on the OpenSCAD2 proposal Doug made some
time ago. The topic is a bit dormant, but as Jordan
wrote, the some internal parts of the objects are now
already present in the dev snapshots and used by the
text/fontmetrics calls.

Also note that other parts of the OpenSCAD2 proposal
are already released, like the function literals.

In my view, it's not meant as big bang change, like
the python2 -> python3 process. Instead it's a guide
for improvement.

ciao,
Torsten.

On Wed, Nov 24, 2021 at 12:30 PM Torsten Paul Torsten.Paul@gmx.de wrote:

On Wed, Nov 24, 2021 at 12:30 PM Torsten Paul Torsten.Paul@gmx.de wrote:

Ok Read all that.

I understand some of the software language engineering issues raised and
that parts of the OpenScad2 proposal are moving forward even if OpenSCAD2
as a whole is not.

What do function literals allow me to do that I could not do without them?
I can see what they are but can't see beyond that.
For what hat are they suitable?  I'm not clear on how to think about coding
with them.

In my view, it's not meant as big bang change, like

itself,  issue threads, and pull requests, and I'm reading more before
commenting further. Expect questions though.

Function literals allow you to store function references in valuables and
pass them as parameters. So for example you could have a quicksort routine
and pass a function to do the comparison.

On Wed, 24 Nov 2021 at 19:09, Patrick Callahan pat.callahan1@gmail.com
wrote:

On 24.11.21 20:08, Patrick Callahan wrote:

Passing functions to other functions or modules allows
for quite powerful library style functions, e.g. fold()
that give the caller influence of calculation details.

fold = function(i, v, f, off = 0) len(v) > off ? fold(f(i, v[off]), v, f, off + 1) : i;

v = [1, 2, 7, 4, -5];

f_add = function(x, y) x + y;
f_mul = function(x, y) x * y;
f_min = function(x, y) is_undef(x) ? y : min(x, y);
f_max = function(x, y) is_undef(x) ? y : max(x, y);

echo(add = fold(0, v, f_add));    // ECHO: add = 9
echo(mul = fold(1, v, f_mul));    // ECHO: mul = -280
echo(min = fold(undef, v, f_min)) // ECHO: -5
echo(max = fold(undef, v, f_max)) // ECHO: 7

More examples: https://github.com/thehans/funcutils

ciao,
Torsten.

I've looked through https://github.com/thehans/funcutils.

Now, after looking through the documentation and function definitions in a
few of the modules,  I wonder how the utility functions defined in the
funcutils library help when one uses them in an OpenSCAD program.
I know this question exhibits my lack of background in functional
programming.

-Pat
I wonder when functional programming will start to click for me.

On Wed, Nov 24, 2021 at 2:59 PM Torsten Paul Torsten.Paul@gmx.de wrote:

It took several tries for it to click. XSLT is the first functional
language I encountered and it took me forever to get it. It didn’t help
that one of the free books I looked at on Amazon did its level best to use
it as an imperative language.

On Wed, Nov 24, 2021 at 7:52 PM Patrick Callahan pat.callahan1@gmail.com
wrote:

My code looks as imperative as possible.  That For statement is my
downfall, even if it is used in a functional way.

It's been said that OpenScad is a functional language.  Is it?

Can you do this?

Array(10).fill("functional programming is not a religion")
.map((msg) => {
console.log(msg);
return msg;
});

Sorry about the impure thoughts Father Horton, see you next Saturday...

On Thu, Nov 25, 2021 at 2:17 AM Father Horton fatherhorton@gmail.com
wrote:

On 26.11.21 13:38, Patrick Callahan wrote:

Sure, but why?

array = [ for (msg = [ for ([0:9]) "blah" ]) echo(msg) msg ];

ciao,
Torsten.

Sorry, It was an attempt at humor in a programming language.  Never a good
idea.

-Pat

On Fri, Nov 26, 2021 at 8:33 AM Torsten Paul Torsten.Paul@gmx.de wrote:

I think if you want OpenSCAD programming to "click" the answer is to
write a bunch of programs where you try to do something difficult.  If
you're trying to do something and you care about efficiency then you
should try to do that thing with OpenSCAD native matrix operations
first.  For example, using the fold() function above to sum up a
simple vector of numbers is 40 times slower than using a vector
product, something like:

sum_of_test = test*[for(i=test) 1]];

The second choice for efficiency (and probably the first for
readability) is to use a list comprehension.  Neither of these two
methods is "functional programming".  List comprehensions can do a lot
of things---anything where the value at one output point is
independent of the value at other output points.  There's the business
of not being able to redefine variables, which can be an occasional
nuisance, but it's not a big issue.  The main consequence is that you
have to handle all cases all at once with the ternary operator, or you
have to introduce more variables.  But if you can't use a list
comprehension because the answer depends on previous answers, then you
need to use recursion.  That's when things look "functional".  If
you're lucky you can do this with tail recursion, and get more
efficiency.  If not you'll be in a less efficient situation.  The last
option is the C-style for loop, which in my testing I have found to be
the slowest option, slower than recursion.  It's hard to use (usually
harder than recursion).

So is OpenSCAD a "functional language"?  I'm not sure.  A lot of
OpenSCAD programming doesn't really seem functional.  If you want to
code in OpenSCAD then you need to learn OpenSCAD and its quirks and
not get hung up on terminology.  At the end of the day, if your
program works does it matter if it was "functional" or "procedural"?

On Fri, Nov 26, 2021 at 9:19 AM Patrick Callahan
pat.callahan1@gmail.com wrote:

I do have something difficult I need to do: somehow find a way to place a
marker on the rendered object interactively and locate that point. Or even
compute a distance between two markers. If I can figure out a way to do
that in openscad I don't care whether it is functional or imperative or
declarative, I like it just fine - and if I can't do that then I'll have to
move to some other system which I would find unfortunate because so far
after a few months openSCAD has always got the job done for me.

Edmund

On Fri, Nov 26, 2021 at 4:05 PM Adrian Mariano avm4@cornell.edu wrote:

I never need to do that because if a dimension is important I specify it in
my source code and make it so in the code. Do you want to measure other
people's designs?

On Fri, 26 Nov 2021 at 18:52, edmund ronald edmundronald@gmail.com wrote:

edmund ronald wrote:

When I need to check a distance/position, I add in geometry as necessary to measure, usually prefacing it w/ a # so that it renders transparent, sometimes assigning it to a different colour.

William

"I never do X" is not really an answer to a question.
My designs contain a bunch of curves, and then to be sure whether they fit
something I need to check.
I once interviewed a senior manager at IBM who told me, for quote to a
business magazine "We don't need mice and windows, we're not children".
A similar story was true of the Mac, where Steve Jobs vetoed arrow keys on
the Mac again because he didn't see a need for them.

More technically, openSCAD has a text based programming model, and a window
and mouse presentation layer. I fully agree that the programming model is
useful and powerful; I just need some added facilities in the presentation
layer. I have looked in the archives and people keep asking for this and
the answer given to them seems to be repeatedly that "real men know their
dimensions" :) I think someone has even tried to put a bounty on such a
measuring feature, which shows how desperate people are.

Edmund

Edmund

On Fri, Nov 26, 2021 at 8:07 PM nop head nop.head@gmail.com wrote:

Well OpenSCAD is a language for defining geometry in code. So if a curve
has to miss something or meet something I express that in the code. I don't
do trial and error. When you preview the geometry is just drawn as pixels,
so there isn't really a way to get dimensions from them, hence why the
requests are not implemented. Once rendered with F6  measurements could be
made like NetFabb is able to measure STLs, but it is a lot of work to
implement that, so it won't happen until a developer with the skill to do
it thinks it is worth spending time on.

[image: image.png]

This is what I would use if I had to measure something. but in this case,
and all others, I can write an expression for the distance and
OpenSCAD echos 17.9.

On Fri, 26 Nov 2021 at 20:31, edmund ronald edmundronald@gmail.com wrote:

This was a discussion about coding and it seems you're trying to divert it
to a totally different subject, since making measurements has nothing to do
with coding.  It seems to me that if this feature were implemented it would
be of very limited utility, because there's no way to unambiguously define
a point by clicking the mouse.  If you want to know that some curved
surface meets some criterion, how can you define via the mouse a point on
that curved surface that is actually well-defined?  This seems impossible,
unless you have some way to define the point such as intersection with some
other pair of surfaces.  I think I have to agree with nophead that if you
have some constraints you need to meet then you should use the programmatic
interface to build a model that meets them, not construct an ad-hoc model
and try to verify that you've met the constraints.  If this doesn't seem
adequate for your problem then I guess OpenSCAD isn't suited to your needs.

On Fri, Nov 26, 2021 at 4:29 PM nop head nop.head@gmail.com wrote:

We fully agree about the fact that making a measurement has nothing to do
with coding. It is a coding environment operation which is akin to
interactively zooming or rotating a render. To that extent it can of course
be omitted from the environment, to the same degree as it is completely
redundant to have a code editor or a visual preview or render inside
OpenScad as many other programs are available to edit code or render a
mesh.

As someone who -of course purely accidentally-  acquired a maths PhD,  I
would of course fully agree that previews are redundant and users should be
able to type in code and visualise an object in their mind as they design
it; in fact my ability to do this is what causes me to use OpenSCAD in the
first place. However I do understand that some other people would prefer
the assistance of a built-in interactive visualiser, and so I would beg you
to consider that the requests for an interactive  measurement aid in the
context not of the language but of the visualiser.

And I would like to leave you with a thought: It is easy to define an
object by a syntax tree - just as it is easy to define a program as a piece
of code. But to visualise the actual physical structure corresponding to an
OpenScad  syntax tree, bugs and all is as hard as predicting the behaviour
of a piece of conventional computer code, bugs and all. There is no
dishonor in providing debugging tools in a programming environment.

Edmund

On Fri, Nov 26, 2021 at 10:53 PM Adrian Mariano avm4@cornell.edu wrote:

On Fri, Nov 26, 2021 at 07:05:35PM +0000, nop head wrote:

Recent project: I had a central "box", some margin, and then a frame
around that. And then a box to fit that frame. So my total size is:
inner_box_x + 2margin + 2framesize.

Sure I can write down the formula for that.

But what if say right side frame-size is determined by a square tilted
at an angle. Sure, I can do the trig to calculate the right side of
that tilted square, but wouldn't it be nice to ask openscad to
calculate that for me?

Roger. 

--
** R.E.Wolff@BitWizard.nl ** https://www.BitWizard.nl/ ** +31-15-2049110 **
**    Delftechpark 11 2628 XJ  Delft, The Netherlands.  KVK: 27239233    **
f equals m times a. When your f is steady, and your m is going down
your a is going up.  -- Chris Hadfield about flying up the space shuttle.