[x3d-public] Wishlist for view3dscene

Fri Feb 10 11:19:51 PST 2023

 There should be an option to convert binary vertex and index data to
ASCII, if only during authoring.

Remember that Joe/designers like to tweak.

I find that converting JSON through JavaScript massively increases
precision, but i haven’t looked at why yet.  Maybe JSON/JavaScript
numerical libraries should be responsible for keeping precise, accurate
numbers, perhaps by storing alternate encodings for I/O and computation.

One can always convert to EXI and back, to serve ASCII purposes.

John

On Fri, Feb 10, 2023 at 12:36 PM Michalis Kamburelis <
michalis.kambi at gmail.com> wrote:

> We use optional scientific notation (with exponent) when writing
> numbers in CGE/view3dscene. To be precise, we use a formatting option
> that outputs either a scientific or decimal notation, whichever is
> shorter. The scientific notation is not that much shorter to win
> usually.
>
> As for round-trips: you must have maximum precision to have reliable
> round-trips, I believe :) Otherwise you will always discover use-cases
> when you loose some digits, and they will be important for someone.
>
> glTF does what is obvious, and I'd say X3D should follow, this is
> something I'd like to work on for X3D 4.1 (
> https://github.com/michaliskambi/x3d-tests/wiki/Binary-meshes ).
> Namely, the floats in per-vertex data are always stored as binary.
> Only the floats like material parameters are stored as text in JSON.
>
> X3D binary encoding is one solution here, but it is not wide-spread,
> e.g. I hope zero hope that Blender exporter would output X3D binary
> encoding someday (there's too many other critical issues in Blender
> X3D exporter). And X3D binary encoding doesn't make data interleaved,
> so it doesn't solve the issue of effectively delivering it to GPU.
>
> "keep numbers and arrays as strings" -- no, this really wouldn't help
> with anything. The X3D browser has to parse the text to numbers at
> some point, to process the points, upload them to GPU etc. Right now
> the bottleneck of reading real-life X3D is always, in my experience,
> "your code effectively has to make million of text-to-float
> conversions". I believe this is a bottleneck regardless of the
> programming language, 1000 * 1000 calls to text-to-float is always
> going to be much slower than reading binary data (where you say "this
> is my pointer to 1000 * 1000 * 4 bytes, just copy there the contents
> of this file").
>
> Regards,
> Michalis
>
> pt., 10 lut 2023 o 18:57 John Carlson <yottzumm at gmail.com> napisał(a):
> >
> > Not sure how feasible this is with Pascal, etc.   Perhaps consider
> scientific or engineering notation with exponents.
> >
> > In most cases i see, extending precision significantly increases file
> size, and makes it harder to do round trips.
> >
> > I don’t know if there’s a JavaScript or Python solution for this.
> >
> > Perhaps we should not implement number types in our synthesized
> programs, and keep numbers and arrays as strings.   Then we have to ask
> what is done in glTF.
> >
> > John
> >
> > On Fri, Feb 10, 2023 at 7:13 AM Michalis Kamburelis <
> michalis.kambi at gmail.com> wrote:
> >>
> >> Thanks, and sorry for a delay in answering!
> >>
> >> The "description" field has been added to texture nodes and few more,
> >> as you can see in another thread on x3d-public :)
> >>
> >> As for digits precision when saving: it's not so easy, in general it's
> >> (one of) the problems that come from X3D using text representation for
> >> floats in per-vertex data. Namely, if we cut the saved numbers to 4
> >> decimal places, I'm 90% sure that I will get a valid bugreport from
> >> someone:
> >>
> >> """
> >> Hey, my stuff no longer works! My models (in glTF, X3D, whatever) have
> >> vertexes at coordinates like Z = 0.00001, 0.00002, 0.00003. And now,
> >> after processing and saving to X3D using your tools, they are all
> >> equal zero, resulting in bugs (things start to fight in Z buffer).
> >> """
> >>
> >> Surely we could add a configuration option for it, but it feels
> >> counter-productive, when my recommendation to everyone would be still
> >> "use the maximum precision, to be safe". I would add a feature that I
> >> explicitly say should not be used.
> >>
> >> In the end, the longer text representation for numbers only hurts the
> >> file size. And I don't believe we can really fix it by outputting less
> >> digits. The right solution is storing per-vertex data in a binary
> >> blob, when each float is binary 4 bytes, so it's
> >>
> >> - efficient storage
> >> - efficient reading by X3D implementation to a "float[] array", no
> >> need for text parsing.
> >>
> >> Regards,
> >> Michalis
> >>
> >>
> >>
> >>
> >> śr., 28 gru 2022 o 21:00 Brutzman, Donald (Don) (CIV)
> >> <brutzman at nps.edu> napisał(a):
> >> >
> >> > Hi Michalis. Joe and I are really enjoying the many excellent
> features of view3dscene.
> >> >
> >> >
> >> >
> >> > Here are some suggested refinements and improvements.
> >> >
> >> >
> >> >
> >> > View3dscene converter: round off significant digits, rather than 10+
> decimal places.  (For example, suggested values in HAnim spec go to 4
> decimal places.
> >> >
> >> >
> >> >
> >> > View3dscene: note that we have added ‘description’ field to many
> nodes in X3D4, including ImageTexture.
> >> >
> >> > 1 warnings:
> >> > X3D: Unknown X3D field name (unhandled X3D XML attribute)
> "description" in node "ImageTexture"
> >> > Scene URL: "C:\x3d-code\www.web3d.org
> \x3d\content\examples\HumanoidAnimation\Skin\JoeSkeletonSkinSiteSaluteWalk.x3d".
> >> > Use "File->View Warnings" menu to view these warnings again.
> >> >
> >> >
> >> >
> >> > Coming soon: import test of three models in HAnim Motion
> specification.
> >> >
> >> >
> https://www.web3d.org/documents/specifications/19774/V2.0/MotionDataAnimation/MotionDataAnimation.html
> >> >
> >> >
> >> >
> >> > Again thanks for great work.
> >> >
> >> >
> >> >
> >> > all the best, Don
> >> >
> >> > --
> >> >
> >> > Don Brutzman  Naval Postgraduate School, Code USW/Br
> brutzman at nps.edu
> >> >
> >> > Watkins 270,  MOVES Institute, Monterey CA 93943-5000 USA
> +1.831.656.2149
> >> >
> >> > X3D graphics, virtual worlds, Navy robotics https://
> faculty.nps.edu/brutzman
> >> >
> >> >
> >>
> >> _______________________________________________
> >> x3d-public mailing list
> >> x3d-public at web3d.org
> >> http://web3d.org/mailman/listinfo/x3d-public_web3d.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://web3d.org/pipermail/x3d-public_web3d.org/attachments/20230210/da0e22f2/attachment-0001.html>