Entity Labels are not unique / Entities cannot be indentified #10

RespectableRuessel · 2020-11-03T13:06:18Z

my team is using the GLTF converter to load IFC models into Unity3d.
While working on the implementation we noticed that there can be duplicate entity labels or that some entity labels belong to the parent and not to the actual entity.
I looked at the GLTF JSON data and noticed that the issue lies in the xBimGltf converter.

{
	"mesh": 348,
	"name": "Basic Roof:Live Roof over Wood Joist Flat Roof:184483 #4229"
},
{
	"mesh": 349,
	"name": "Stair:Residential - 200mm Max Riser 250mm Tread:151086:1 #5984"
},
{
	"mesh": 350,
	"name": "Stair:Residential - 200mm Max Riser 250mm Tread:151086:1 #5984"
},
{
	"mesh": 351,
	"name": "Stair:Residential - 200mm Max Riser 250mm Tread:151086:1 #5984"
},
{
	"mesh": 352,
	"name": "Stair:Residential - 200mm Max Riser 250mm Tread:151086:1 #5984"
},
{
	"mesh": 353,
	"name": "Stair:Residential - 200mm Max Riser 250mm Tread:151086:1 #5984"
},

The name is always the same with the same entity label (#5984) only with different mesh ids.

In unity the mesh id gets added to the entity label because otherwise the name is not unique.
Our problem is that we need the actual entity ids to match them with existing IFC data.

For example we have a table with different mesh components that have different surface styles but due to the converter issue we get the same entity id for every mesh and therefore cannot match the surface styles to the entity label.

While looking at your code I also noticed some strange lines like

if (productLabel != shapeInstance.IfcProductLabel)

in Builder.cs#375 because you never change the productLabel after initializing it to zero in Builder.cs#348.

Is it possible to fix these issues?
I would be willing to fix parts of it myself, if you can tell me where exactly our issue is located.

Thanks for reading.

The text was updated successfully, but these errors were encountered:

CBenghi · 2020-11-03T14:45:59Z

Hi,
can you share a small model to be sure we are on the same page?
You can use the simplify feature of XbimExplorer to retain just a couple of elements to avoid disclosing models.
Thanks,
Claudio

RespectableRuessel · 2020-11-03T15:32:32Z

The building has a coffee table that has a inner glass and outer wood surface.
Duplex_A_20110907_optimized.zip

And in the GLTF file every mesh has the same entity label (see below) which is a problem because we are not able to match it to the entity labels from the surface styles.

 {
      "matrix": [
        -1.0,
        0.0,
        0.0,
        0.0,
        0.0,
        -1.0,
        0.0,
        0.0,
        0.0,
        0.0,
        1.0,
        0.0,
        2.61916518,
        -15.3431711,
        0.0,
        1.0
      ],
      "mesh": 288,
      "name": "M_Table-Coffee:0915 x 1830 x 0457mm:0915 x 1830 x 0457mm:168381 #498"
    },
    {
      "matrix": [
        -1.0,
        0.0,
        0.0,
        0.0,
        0.0,
        -1.0,
        0.0,
        0.0,
        0.0,
        0.0,
        1.0,
        0.0,
        2.61916518,
        -15.3431711,
        0.0,
        1.0
      ],
      "mesh": 289,
      "name": "M_Table-Coffee:0915 x 1830 x 0457mm:0915 x 1830 x 0457mm:168381 #498"
    },
    {
      "matrix": [
        -1.0,
        0.0,
        0.0,
        0.0,
        0.0,
        -1.0,
        0.0,
        0.0,
        0.0,
        0.0,
        1.0,
        0.0,
        2.61916518,
        -15.3431711,
        0.0,
        1.0
      ],
      "mesh": 290,
      "name": "M_Table-Coffee:0915 x 1830 x 0457mm:0915 x 1830 x 0457mm:168381 #498"
    },

RespectableRuessel · 2020-11-03T15:38:52Z

I think the Builder.cs#352 and Builder.cs#375 lines are on the right track.

foreach (var shapeInstance in shapeInstances.OrderBy(x => x.IfcProductLabel))

if (productLabel != shapeInstance.IfcProductLabel)

Check the product label.
A. If its the same product label (because its ordered) add the mesh to the group.
B. If the product label does not match create a new group and add the mesh to it.

A and B are not in there because in the source int productLabel = 0; is never changed. So the actual comparison is

if (0 != shapeInstance.IfcProductLabel)

which is always true.

The same issue with Builder.cs#402.

targetMesh = new gltf.Mesh
{
    Name = $"Instance {productLabel}"
};

The productLabel is always zero which leads to the string Instance 0 everytime.

CBenghi · 2020-11-03T15:46:06Z

Thanks... If you are far ahead of me a PR is welcome, otherwise I'll have a look. Best, Claudio

…

On Tue, 3 Nov 2020 at 16:39, RespectableRuessel ***@***.***> wrote: I think the Builder.cs#352 and Builder.cs#375 lines are on the right track. foreach (var shapeInstance in shapeInstances.OrderBy(x => x.IfcProductLabel)) if (productLabel != shapeInstance.IfcProductLabel) Check the product label. A. If its the same product label (because its ordered) add the mesh to the group. B. If the product label does not match create a new group and add the mesh to it. A and B are not in there because in the source int productLabel = 0; is never changed. So the actual comparison is if (productLabel != shapeInstance.IfcProductLabel) which is always true. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#10 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABJY7MNZJHDYZEUKPUNNRETSOAPZZANCNFSM4TIXDYQQ> .

RespectableRuessel · 2020-11-03T15:49:07Z

I'll see if I can fix it. I'm sure these lines (productLabel) are not correct, but I'm not sure if this will fix the problem.

CBenghi · 2020-11-03T15:52:30Z

It matters that the code can be synced to the json metadata export. If you are in doubt I'll give it a go.

…

On Tue, 3 Nov 2020 at 16:49, RespectableRuessel ***@***.***> wrote: I'll see if I can fix it. I'm sure these lines (productLabel) are not correct, but I'm not sure if this will fix the problem. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#10 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABJY7MN545TUDEVUH6FLGN3SOARAFANCNFSM4TIXDYQQ> .

RespectableRuessel · 2020-11-03T16:24:11Z

I have tested the changes and it looks like the hierarchy is now correct (single parent, one entity label) but the mesh names are now duplicates and I'm not able to get the entity labels for the child elements.
If you have time I would be thankful if you can take a look.

Here is what I've done:
RespectableRuessel@d6c4525

RespectableRuessel · 2020-11-23T20:45:00Z

@CBenghi tried a few more solutions but none worked. Would it be possible for you to look at the EntityLabel issue?

CBenghi · 2020-11-23T23:28:17Z

It might take a few days.. Deadline for me on the 3rd December. Will look into it after that. A reminder would be appreciated then... I'll have a few things piling up. thanks, Claudio

…

On Mon, 23 Nov 2020 at 21:45, RespectableRuessel ***@***.***> wrote: @CBenghiI tried a few more solutions but none worked. Would it be possible for you to look at the EntityLabel issue? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#10 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABJY7MPI7KCBOASTRPL7VSLSRLCVZANCNFSM4TIXDYQQ> .

RespectableRuessel · 2020-11-24T19:24:29Z

Will do. Big thanks!

RespectableRuessel · 2020-12-03T23:29:14Z

Reminder ⏰ if you have some time.

#10

CBenghi · 2020-12-04T10:36:02Z

Hello @RespectableRuessel,
I've looked at the code and you were clearly working in the right direction.
So I've changed the code to something much aligned to what you did.
But then I cannot understand your outstanding issue:

but the mesh names are now duplicates and I'm not able to get the entity labels for the child elements.

In the duplex that I get after conversion there are 215 meshes and they all have different instance names.
Can you clarify? Thanks.

RespectableRuessel · 2020-12-08T13:45:20Z

Sry for the late response.
The issue is that in my tests the mesh names are duplicates with an additional number added (prop. automatically) to guarantee uniqueness (see pictures below).

Also the mesh instances cannot be identified separately because they all belong to the same parent which only has one GUID. The goal is to identify every mesh by the entity label. That allows us to texturize every mesh based on the material for that specific mesh. The table for example has a glass surface and a wooden frame. But at the moment we are only able to identify the table because its the only element with a GUID / entity label.

Thanks again for looking into my problem.

RespectableRuessel · 2020-12-08T13:52:27Z

In the duplex that I get after conversion there are 215 meshes and they all have different instance names.
Can you clarify? Thanks.

Ahh sry I found the misunderstanding:
The mesh with the instance 274 has multiple primitives that are split to different objects by our converter to allow us to texture them separately.

{
      "primitives": [
        {
          "attributes": {
            "NORMAL": 355,
            "POSITION": 356
          },
          "indices": 354,
          "material": 9
        },
        {
          "attributes": {
            "NORMAL": 358,
            "POSITION": 359
          },
          "indices": 357,
          "material": 9
        },
        {
          "attributes": {
            "NORMAL": 361,
            "POSITION": 362
          },
          "indices": 360,
          "material": 9
        },
        {
          "attributes": {
            "NORMAL": 364,
            "POSITION": 365
          },
          "indices": 363,
          "material": 16
        }
      ],
      "name": "Instance 274"
    },

The xBimGltf converter should instead write out several meshes with the entity labels as names instead of one mesh with several primitives. That would allow referencing the unique parts of the table, for example.

CBenghi · 2020-12-08T14:19:25Z

I get it now.
It sounds something that I might do as a future development, but at the moment this is not high in my roadmap.

I would welcome a PR that implemented this via a configurable flag in the exporter, as I see many scenarios where breaking the sub-meshes might not be beneficial for rendering performance.
Ideally there could be a selector that made this entity-dependent or type-dependent so that only portions of the model follow this logic.

Thanks,
Claudio

RespectableRuessel · 2020-12-08T16:04:23Z

Can you point me to the things that I need to resolve the shape instance into the right entity?
Because GetShapeInstancesToRender does not return actual entities. I can only extract the product name.

CBenghi · 2020-12-08T16:43:16Z

Sorry, I don't understand what you meant here.
The code is not richly documented, but have you tried to see what comes through in debugging?
There's much more than the name in the XbimShapeInstance class.

RespectableRuessel · 2020-12-08T16:56:22Z

Sorry, I don't understand what you meant here.

Was a bad explanation by me, what I meant is the following:
The moment GetShapeInstancesToRender reads the shape instances from the geomReader.ShapeInstances the connection between entities and their shapes is lost (at least for my understanding). If I'm correct the converter just reads the data from the geometry store and writes it to GLTF.
What I don't get is how to receive the actual entity that the shape belongs to.
For example: Table (Product) -> Table surface (Entity) -> Shape
I get the shape from the geometry reader but I need to move up higher in the hierarchy to get the entity instance that the shape belongs to.
It is also possible that I simply do not understand xBim architecture because there is so little documentation.

RespectableRuessel · 2021-01-25T13:25:21Z

Closed because of #17

CBenghi added a commit that referenced this issue Dec 4, 2020

Preliminary fix for github issue #10

b924880

#10

CBenghi self-assigned this Dec 4, 2020

CBenghi added the question Further information is requested label Dec 4, 2020

CBenghi added enhancement New feature or request and removed question Further information is requested labels Dec 8, 2020

RespectableRuessel mentioned this issue Jan 25, 2021

Rework of the GLTF converter #17

Open

RespectableRuessel closed this as completed Jan 25, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Entity Labels are not unique / Entities cannot be indentified #10

Entity Labels are not unique / Entities cannot be indentified #10

RespectableRuessel commented Nov 3, 2020

CBenghi commented Nov 3, 2020 •

edited

Loading

RespectableRuessel commented Nov 3, 2020

RespectableRuessel commented Nov 3, 2020 •

edited

Loading

CBenghi commented Nov 3, 2020 via email

RespectableRuessel commented Nov 3, 2020

CBenghi commented Nov 3, 2020 via email

RespectableRuessel commented Nov 3, 2020 •

edited

Loading

RespectableRuessel commented Nov 23, 2020 •

edited

Loading

CBenghi commented Nov 23, 2020 via email

RespectableRuessel commented Nov 24, 2020

RespectableRuessel commented Dec 3, 2020

CBenghi commented Dec 4, 2020

RespectableRuessel commented Dec 8, 2020

RespectableRuessel commented Dec 8, 2020

CBenghi commented Dec 8, 2020 •

edited

Loading

RespectableRuessel commented Dec 8, 2020

CBenghi commented Dec 8, 2020

RespectableRuessel commented Dec 8, 2020

RespectableRuessel commented Jan 25, 2021

Entity Labels are not unique / Entities cannot be indentified #10

Entity Labels are not unique / Entities cannot be indentified #10

Comments

RespectableRuessel commented Nov 3, 2020

CBenghi commented Nov 3, 2020 • edited Loading

RespectableRuessel commented Nov 3, 2020

RespectableRuessel commented Nov 3, 2020 • edited Loading

CBenghi commented Nov 3, 2020 via email

RespectableRuessel commented Nov 3, 2020

CBenghi commented Nov 3, 2020 via email

RespectableRuessel commented Nov 3, 2020 • edited Loading

RespectableRuessel commented Nov 23, 2020 • edited Loading

CBenghi commented Nov 23, 2020 via email

RespectableRuessel commented Nov 24, 2020

RespectableRuessel commented Dec 3, 2020

CBenghi commented Dec 4, 2020

RespectableRuessel commented Dec 8, 2020

RespectableRuessel commented Dec 8, 2020

CBenghi commented Dec 8, 2020 • edited Loading

RespectableRuessel commented Dec 8, 2020

CBenghi commented Dec 8, 2020

RespectableRuessel commented Dec 8, 2020

RespectableRuessel commented Jan 25, 2021

CBenghi commented Nov 3, 2020 •

edited

Loading

RespectableRuessel commented Nov 3, 2020 •

edited

Loading

RespectableRuessel commented Nov 3, 2020 •

edited

Loading

RespectableRuessel commented Nov 23, 2020 •

edited

Loading

CBenghi commented Dec 8, 2020 •

edited

Loading