03:43 <InPhase> We really should improve the performance of typos.

03:47 <InPhase> I was testing my BBox on some real world examples. Worked fine except for the one that I ran on pointarras instead of pointarrays, which rather than 1 second of bounding box calculation for 163,172 points, it launched into 100% cpu for like 10 minutes before I killed it.

03:47 <InPhase> And unfortunately, this happens a lot.

03:48 <InPhase> Drop in some typo, you get undef, you propagate that undef through a bunch of involved calculations, and suddenly the runtime increases by more than a factor of 1000.

04:12 <peepsalot> InPhase: stop on first warning?

04:21 <peepsalot> really instead of a checkbox I think we should just have a number entry "Allow up to __ warnings before halting."

04:22 <peepsalot> the error log gets increasingly slow after thousands of entries

04:23 <peepsalot> that was the biggest slowdown I observed anyways

04:31 <peepsalot> i remember profiling that when I was working on Console optimizations

05:21 <J1A8479> afaik if calculating points you always get a list of all points that failed to calc (even with stop on first warning)

05:42 <InPhase> peepsalot: Yeah, a numerical entry would probably do it.

05:42 <InPhase> One is not enough for me. 100 would be plenty.

05:43 <InPhase> I probably had closer to 2 million being generated in that model, which would never be useful for anyone.

06:43 <lf94> is smooth union in openscad yet

12:15 <gbruno> [github] Parkinbot opened issue #4321 (Feature request: simple view port animation without rerender) https://github.com/openscad/openscad/issues/4321

17:28 <teepee> ah, peepsalot is right, https://github.com/openscad/openscad/issues/4002 is a duplicate. J1A8479 on what operating system is that happening for you?

17:29 <J1A8479> windows

17:37 <gbruno> [github] UBaer21 edited issue #4002 (inconsistent render of transparent object in export png) https://github.com/openscad/openscad/issues/4002

17:41 <teepee> so that matches the other issue too

17:42 <teepee> I'll close this issue and copy the note over to the other one

17:44 <gbruno> [github] t-paul closed issue #4002 (inconsistent render of transparent object in export png) https://github.com/openscad/openscad/issues/4002

17:45 <peepsalot> weird i thought i closed it when i marked as duplicate

17:48 <teepee> github is acting strange a bit, I had to reload 3 times to actually comment

18:54 <joseph_> teepee InPhase: Yesterday and today I've been focusing on investigating the failing tests. I'll let you know when I have produced some fixes or run into a substantial block that I need advice about. I have not been focusing much more on the WASM build issue yet because I didn't hear whether I'll be able to find the dummy class elsewhere in the codebase already, or if it requires special knowledge from a community member

18:57 <InPhase> joseph_: Ok. Feel free to discuss intermediate bits of the work with more detail as well, so that you do not get stuck on confusing parts for long.

18:58 <InPhase> joseph_: It's not guaranteed anyone will know the answer to a specific question, but the price of asking is smaller than the price of long confusion.

19:05 <teepee> joseph_: looking at CMakeList.txt it's probably tied to the NULLGL define in code, which I can see in Render.cc for example

19:05 <teepee> with the dummy stuff at the bottom of that file

19:06 <teepee> from the error message I was guessing that class has a new method and this is not mentioned in that NULLGL case

19:36 <gbruno> [github] t-paul pushed 1 modifications (Update news ticker.) https://github.com/openscad/openscad.github.com/commit/ef0ed66f30f56d5bd95de0e654b38ff9608ed015

20:16 <linext> any recommendations for things to print on a resin printer?

20:17 <Scopeuk> linext can't recommend specific models but anything small and high/ultra high detail is what they excel at, jewelry mock ups, figurines that sort of thing

20:17 <Friithian> highly accurate model of a heart

20:19 <Friithian> minis, as always

20:20 <linext> i made the gear cube

20:23 <InPhase> linext: Inquiring minds want to know: https://www.thingiverse.com/thing:1695879

20:24 <linext> depends what you're nailing

20:24 <Friithian> if the answer is wood, it's a solid lmao no

20:24 <InPhase> Well the ideal is softwood. :)

20:25 <InPhase> We just need good enough filament!

20:25 <Friithian> I tried ABS, PLA, PETG

20:26 <linext> maybe if you predrill the hole

20:26 <linext> slightly smaller than the nail

20:34 <J1A8479> make the pilot hole bigger and give the nail some ratchets

20:35 <J1A8479> so it acts as dowel

20:35 <Friithian> give the nail a spiral groove on the outside and use that to push itself in

20:35 <Friithian> waitaminute…

20:36 <J1A8479> replace the plastic nail with a steel pin

20:36 <InPhase> Friithian: There's no need, just add a spiral groove and use them like ratchets when you hammer it in.

20:39 <J1A8479> https://eaglefloorcoveringsupplies.com/wp-content/uploads/2017/05/Spiral-Nails-Pewter.jpg

20:39 <Friithian> pewter? interesting

20:40 <J1A8479> they also exist as steel

20:44 <J1A8479> guess the pewter are deco nails

20:58 <pa> i suppose it's not possible to transform children with a function, instead of a matrix? like y = 0.2x + 1.5

20:59 <pa> or y_out = 0.2x + y

21:32 <gbruno> [github] jbinvnt synchronize pull request #4305 (Select shader location) https://github.com/openscad/openscad/pull/4305

21:34 <J1A8479> pa sure  you just put the function into  a translate() command to move your child  or use multmatrix()

21:35 <pa> J1A8479: well i actually meant to, for example, scale a child with such a function

21:35 <pa> so that y wouldn't be scaled , say, 0.2, but proportionaly to x position of each point

21:37 <J1A8479> a child(ren)  is  an object not a point  .. you can use scale or multmatrix to independently scale xyz

21:39 <J1A8479> you would use a loop to move  (translate(p[i]) scale([i*.2+1.5, 1,1]) children();     you can also put a function into the scale that gets x of the point as input

21:41 <J1A8479> for(p=points) translate(p)scale ([p.x*.2+1.5,1,1])children();

21:48 <linext> anyone have a ideas for the domain name of a customizer site?

21:52 <teepee> oof, names

21:53 <J1A8479> oof.cloud

21:54 <J1A8479> oof.name Ü

21:56 <teepee> haha, good idea, leave it to the users to explain :)

22:00 <teepee> transform.cloud

22:02 <J1A8479> https://en.wikipedia.org/wiki/List_of_Internet_top-level_domains

22:07 <teepee> 3d.design

22:12 <InPhase> pa: $fn=24; module Foo() { multmatrix([[1,0,0,0],[0.2,1,0],[0,0,1,0]]) children(); } Foo() translate([2, 1, 0.5]) sphere(2);

22:12 <InPhase> pa: You just need to embrace the matrix math.

22:13 <InPhase> pa: Oops, that should have been $fn=24; module Foo() { multmatrix([[1,0,0,0],[0.2,1,0,0],[0,0,1,0]]) children(); } Foo() translate([2, 1, 0.5]) sphere(2);

22:13 <InPhase> Can't count when writing that compactly.

22:17 <J1A8479> hope i am not the bad influence that your code become unreadable like mine Ü

22:17 <InPhase> pa: I had flushed out that multmatrix section of the manual with some info and examples a while ago, so I think it will give a good explanation of that. Although if you aren't familiar with matrix multiplication on a vector, looking up that might help.

22:18 <InPhase> J1A8479: Hahah. Well I knew it was going to be a little long and wanted to one-line it for the paste. ;)

22:18 <linext> 2nd for 3d.design

22:18 <J1A8479> can't belive that 3d.design doesn't exist .. probably *.design are one of these 100k domains ..

22:19 <J1A8479> s/are/is

22:19 <InPhase> 3d.design is super memorable if you could get it. Can't beat that.

22:22 <J1A8479> ah 2 letter .. need to be 3  ..   http://ddd.design/   exist . .. and about 3 to 6 k per year

22:23 <linext> i want to find an unused domain

22:24 <InPhase> 3dp.design

22:25 <InPhase> Oh, which is taken

22:25 <J1A8479> p?

22:26 <J1A8479> ah print

22:26 <J1A8479> oh already tomorow .. Ü

22:29 <J1A8479> hmm if i make an app that translate domain names into TOR adresses ..  i can sell every domain name Ü

22:29 <linext> .net or a .org seems applicable if users can allow their PCs to render other jobs

22:29 <J1A8479> *.DNS

22:30 <J1A8479> render.net

22:31 <linext> custom3d.design is $40 year

22:33 <J1A8479> https://en.wikipedia.org/wiki/.sucks_(registry)

22:33 <linext> form3d.org $12/year

22:34 <JordanBrown> I keep thinking that it should be possible to write a skew() function that is more directly comprehensible than multmatrix(), but I suspect that to be general it would need to take a 3x3 matrix input, at which point it's almost multmatrix().

22:35 <teepee> isn't form3d the name of one of those first sla printers?

22:35 <teepee> and company?

22:35 <J1A8479> seems length to costs is anti exponential

22:35 <J1A8479> c3d.design ..

22:36 <teepee> ah, no formlabs and form 3 - close :)

22:37 <J1A8479> JordanBrown  i have skew in my lib .. as  M(skewxz=1)

22:37 <JordanBrown> I don't know what you mean by M(skewxz=1).

22:38 <JordanBrown> I have at least one skew* function in my library, but I don't remember what its semantics are. I just know it wasn't a general solution.

22:38 <J1A8479> M(skewxz=1)cube();   skews a cube along xz

22:39 <J1A8479> you sure can also have skewzx

22:39 <JordanBrown> It seems like having skewx(y,z), skewy(x, z), and skewz(x,y) would be more or less complete. All of them meaning "as you move in X, scale Y and Z", and so on.

22:44 <J1A8479> the most cases will be  Z axis skewed in X or Y  ..  for the rest you could rotate your object

22:50 <JordanBrown> Well, sure, but that's like saying that you only need scale([x,1,1]), since you can always rotate your object into the right orientation.

22:51 <J1A8479> you mean like rotate_extrude is always around Z ?

22:52 <JordanBrown> Yeah, true, and linear_extrude is always into Z.

22:52 <J1A8479> and you sure want to scale in 3D  but you rarely skew in 2 or 3 dimensions

22:53 <JordanBrown> https://pastebin.com/kuNtiNmU has skewx, skewy, and skewz and seem to do sensible things.

22:54 <JordanBrown> Skewing in three dimensions hurts my head.

22:54 <J1A8479> as said i have all 6 in one module, so you can combine them if wanted

22:55 <JordanBrown> I haven't truly wrapped my head around what skewing deeply *means*, so I'm not sure I'm really covering all of it.

22:55 <J1A8479> the problem is you need a 3D monitor as all this skewing lead to optical illusions

22:55 <JordanBrown> Yes, I was noticing that simple skewed cubes didn't look right.

22:58 <J1A8479> would be fun if like skew you could scale  along a dimension like  a sphere would become an egg

22:59 <J1A8479> so having a transformation boundingBox  that can also twist and bend

23:13 <JordanBrown> Wearing my "not a mathematician" hat and using terms that I'm not 100% sure of the definition of, I think that's not an "affine" transformation and can't be achieved with a 4x4 matrix multiply.

23:14 <JordanBrown> But yes, it would be nice if you could say

23:14 <JordanBrown> newpoints = [ for (p = points) myfunction(p) ];

23:14 <JordanBrown> and have it apply to geometry.

23:15 <JordanBrown> Well, hmm. 4x4 can do perspective transforms, and I think that's related. But I don't think OpenSCAD does 4x4; I think it only does 3x4.

23:34 <InPhase> J1A8479: Honestly I find "skewing" a fuzzy concept, but the matrix math is intuitive to me, so my mental mapping is "skewing is what happens when the number is there in the matrix". :)

23:35 <InPhase> I'm not really sure where people would go and pick up an intuitive mental model of skewing without it involving these matrices.