Feature/geometry overlay by jwasikpsnc · Pull Request #79 · iterorganization/IBEX

jwasikpsnc · 2026-06-02T06:30:26Z

New endpoint /ids_info/geometry_overlay_nodes
- returns geometry overlay nodes from IDS (pointed by uri)
- allows to show/hide structures
- allows to show/hide empty nodes (IMAS-Core >= 5.7 & IMAS-Python >= 2.2.2)
- allows to show/hide error bars

jwasikpsnc · 2026-06-02T06:30:40Z

FYI @mfiaudrin-epsyl

read-the-docs-community · 2026-06-02T06:31:35Z

Documentation build overview

📚 imas-ibex | 🛠️ Build #33112394 | 📁 Comparing 45776b8 against latest (2b772fc)

🔍 Preview build

48 files changed · ± 48 modified

± Modified

mfiaudrin-epsyl · 2026-06-02T14:34:17Z

Hi @jwasikpsnc,

Thank you for the feature. I think the geometries returned are insufficient. @olivhoenen Can you please ensure if I undestand well the geometries I have to overlay ?

I explain my point of view. Currently, the endpoint return all geometry nodes included somewhere in the given uri.
However, I don't think it's the exepected behavior checking with this example given by @olivhoenen :

As you can see, in this example the uri is : imas:hdf5?path=/home/ITER/hoeneno/public/expdata/HL3/20250515#equilibrium:0/time_slice[:]/profiles_2d[:]/psi
Since the data displayed is in equilibrium, we don't receive geometries from pf_active and wall :

In the feature I am implementing, I want to display #equilibrium:0/time_slice[:]/profiles_2d[:]/psi and to overlay geometries from pf_active and wall (maybe I have more geometries to overlay but I think I should at least overlay the mentionned ones).

Currently in my feature, the paths I overlay with psi are :

'#wall:0/description_2d[:]/limiter/unit[:]/outline',
'#pf_active/coil[:]/element[:]/geometry/rectangle',
'#pf_active/coil[:]/element[:]/geometry/oblique',

There are the overlayed geometries :

If I am not wrong, we should get all geometries having data from this uri: imas:hdf5?path=/home/ITER/hoeneno/public/expdata/HL3/20250515 but it doesn't return anything:

@olivhoenen Can you please confirm what geometries we need to overlay in in this example ?

FYI @imbeauf

jwasikpsnc · 2026-06-03T05:13:04Z

@mfiaudrin-epsyl
Current behavior returns geometries from IDS pointed by URI.
Do you want to get all geometry nodes from whole pulsefile and all IDSes, or you need some more logic to return only geometries plottable on the same plot with for instance psi?
I am not sure how we can extract this information from DD.

mfiaudrin-epsyl · 2026-06-03T06:10:58Z

@mfiaudrin-epsyl Current behavior returns geometries from IDS pointed by URI. Do you want to get all geometry nodes from whole pulsefile and all IDSes, or you need some more logic to return only geometries plottable on the same plot with for instance psi? I am not sure how we can extract this information from DD.

@jwasikpsnc I would like that the BE returns me all plottable geometries compatible with the displayed plot. To do that I think we need to get them from whole pulsefile but we certainly need to add more logic to the rule.

I don't know what is the rule so it will simplify the FE to ask to the BE what geometries I should overlay. Maybe we need @olivhoenen and @imbeauf to clarify this need.

maarten-ic · 2026-06-03T06:22:22Z

There is a lot of geometric data that can be stored in IMAS: shouldn't the user select which geometries they want to overlay, instead of IBEX plotting everything that's stored in the dataset? Maybe that could work in a similar way as that users now select which quantities they want to plot?

Sometimes you may want to plot geometry data from a machine description URI and overlay with data from another URI: will that be possible?

imbeauf · 2026-06-03T06:51:03Z

If I understood correctly, the idea of Olivier is:

The BE sends to the FE all the non-empty geometries (defined as per the generic types that we chose) of the selected data entry (i.e. all IDSs)
The paths of the non-empty geometries form a menu on the FE
The user selects from the menu which data he wants to overlay

imbeauf · 2026-06-03T06:56:52Z

Sometimes you may want to plot geometry data from a machine description URI and overlay with data from another URI: will that be possible?

I agree that this would be useful, but this may double the size of the menu for selecting which URIs to plot ... not sure that this will be very practical if we keep this menu concept.

mfiaudrin-epsyl · 2026-06-03T09:03:39Z

The fact that we want to display geometries from another URI should be possible.

Maybe we could allow the user to select the geometries he want to display in the visual customization component ?
In that case, I could add an accordion to select geometries and I could provide a parameter to choose the URI used for the geometry.

If needed, I can allow the user to add geometries from several URIs at the same time. Do you think it will be usefull or we just want to display geometries from one URI ?

imbeauf · 2026-06-03T09:16:02Z

Displaying geometries from different URIs will certainly be useful, if you find a way to present this nicely on the FE !

mfiaudrin-epsyl · 2026-06-03T09:21:20Z

@jwasikpsnc, having this conversation helps me to clarify the wanted response from BE.

I would like to always get the structures and their parameters. So I don't need the parameter show_structures, it could be always set to true.

I will use this object to easily use the data returned by your endpoint:

[
    {
        geometryNode: 'imas:hdf5?path=/home/ITER/hoeneno/public/expdata/HL3/20250515#pf_active:0/coil[:]/element[:]/geometry/rectangle/',
        parameters: ['r', 'z', 'width', 'height'],
    },
    ...
]

Is it possible for you to return the data in an object like this one ?

jwasikpsnc · 2026-06-08T11:14:22Z

@olivhoenen This PR is ready for further review

@mfiaudrin-epsyl Please check if new endpoint fulfills your needs

mfiaudrin-epsyl · 2026-06-08T11:52:56Z

Hi @jwasikpsnc,

I am testing with:
URI : imas:hdf5?path=/home/ITER/hoeneno/public/expdata/HL3/20250515#equilibrium:0/time_slice[:]/profiles_2d[:]/psi

It almost good for me. The only thing to improves is here:

Instead of element[:]/geometry list of nodes, I would like to have each nodes with their parameters:

geometry_node: element[:]/geometry/outline
- parameters: ["r", "z"]
geometry_node: element[:]/geometry/rectangle
- parameters: ["r", "z", "width", "height"]
etc.

Here is the structure from IBEX tree:

About geometry/geometry_type node I'm wondering if we need it. It's a leaf node and currently I don't know if we have to handle it and how.

jwasikpsnc · 2026-06-08T11:57:11Z

Ok, understood.
Will do it tomorrow

maarten-ic · 2026-06-08T12:17:15Z

About geometry/geometry_type node I'm wondering if we need it. It's a leaf node and currently I don't know if we have to handle it and how.

The geometry_type should be filled with a number that indicates which of the geometry nodes are filled. See for example the documentation in pf_active: https://imas-data-dictionary.readthedocs.io/en/latest/generated/ids/pf_active.html#pf_active-coil-geometry-geometry_type

Type used to describe the element shape (1:’outline’, 2:’rectangle’, 3:’oblique’, 4:’arcs of circle, 5: ‘annulus’, 6 : ‘thick line’)

It may be more efficient to parse the geometry_type so you know which of the actual geometry data you need to read?

jwasikpsnc · 2026-06-09T06:16:07Z

Current implementation cuts out geometry_type from list.
If not cutted out response will contain something like this:

{
  "geometry_node": "imas:hdf5?path=/home/ITER/hoeneno/public/expdata/HL3/20250515#pf_active:0/coil[:]/element[:]/geometry",
  "parameters": ["geometry_type"]
}

Would it be useful for you @mfiaudrin-epsyl?

mfiaudrin-epsyl · 2026-06-09T06:27:33Z

Current implementation cuts out geometry_type from list. If not cutted out response will contain something like this:
{
  "geometry_node": "imas:hdf5?path=/home/ITER/hoeneno/public/expdata/HL3/20250515#pf_active:0/coil[:]/element[:]/geometry",
  "parameters": ["geometry_type"]
}
Would it be useful for you @mfiaudrin-epsyl?

Yes, the not cutted out response will be better

mfiaudrin-epsyl · 2026-06-09T07:17:22Z

About geometry/geometry_type node I'm wondering if we need it. It's a leaf node and currently I don't know if we have to handle it and how.

The geometry_type should be filled with a number that indicates which of the geometry nodes are filled. See for example the documentation in pf_active: https://imas-data-dictionary.readthedocs.io/en/latest/generated/ids/pf_active.html#pf_active-coil-geometry-geometry_type

Type used to describe the element shape (1:’outline’, 2:’rectangle’, 3:’oblique’, 4:’arcs of circle, 5: ‘annulus’, 6 : ‘thick line’)

It may be more efficient to parse the geometry_type so you know which of the actual geometry data you need to read?

Ok thank you. I will test it once I will return on geometry overlay feature. If there is a way we can improve performances I will inform how we can do it.

mfiaudrin-epsyl · 2026-06-09T07:34:53Z

Could be done. Simplest will be to add key to ids_info/node_info/ response. For uri: imas:hdf5?path=/home/ITER/hoeneno/public/expdata/HL3/20250515#pf_active:0/coil[:]/element[:]/geometry:
{
  "name": "geometry",
  "type": "structure",
  "ndim": 0,
  "shape": [],
  "has_data": null,
  "is_geometry_strucure": True # <========= HERE
  "children": [
    {
      "name": "geometry_type",
      "type": "INT",
      "ndim": 0,
      "has_data": null
    },
    ...
    {
      "name": "thick_line",
      "type": "structure",
      "ndim": 0,
      "has_data": null
    }
  ],
  "coordinates": [
    "1...N",
    "1...N"
  ]
}
I would add "is_geometry_strucure": True key. Is it ok to have it only on structure level?:
`coil[:]/element[:]/geometry`
or do you need it also for children structures and leaf nodes?
`coil[:]/element[:]/geometry/outline`
`coil[:]/element[:]/geometry/outline/r`
`coil[:]/element[:]/geometry/outline/z`
It is much easier for me to do it only for geometry stuctures, without chilren

Which solution fits you best?

Ok the is_geometry_strucure attribut is good for me. I will prefer to also add this attribut in the leaf nodes but if it's more works for you don't worry. I will add a rule in the FE to provide the geometry information in his children

jwasikpsnc · 2026-06-09T08:35:24Z

Done with one minor change
replaced is_geometry_strucure with is_geometry_node so it won't cause confusion when assigned to leaf nodes.

Now is_geometry_node is returned in /ids_info/node_info/

Example:
imas:hdf5?path=/home/ITER/hoeneno/public/expdata/HL3/20250515#pf_active:0/coil[:]/element[:]

{
  "name": "element",
  "type": "struct_array",
  "ndim": 1,
  "shape": [],
  "is_geometry_node": false,
  "has_data": null,
  "children": [
    {
      "name": "name",
      "type": "STR",
      "ndim": 0,
      "is_geometry_node": false,
      "has_data": null
    },
    ...
    {
      "name": "geometry",
      "type": "structure",
      "ndim": 0,
      "is_geometry_node": true,
      "has_data": null
    }
  ],
  "coordinates": [
    "1...N",
    "1...N"
  ]
}

maarten-ic · 2026-06-09T09:05:59Z

@jwasikpsnc At the tree, I would like to prevent users from clicking in geometry nodes by displaying them with the disabled state. However, currently I don't know which node is used for geometry.

It may still be relevant to view the metadata associated to geometry nodes (for example, the exact values). Is that still possible when you disable the geometry nodes in the tree?

mfiaudrin-epsyl · 2026-06-09T09:14:44Z

@jwasikpsnc At the tree, I would like to prevent users from clicking in geometry nodes by displaying them with the disabled state. However, currently I don't know which node is used for geometry.

It may still be relevant to view the metadata associated to geometry nodes (for example, the exact values). Is that still possible when you disable the geometry nodes in the tree?

It can be done if needed. I propose this solution:

Disable the geometry nodes only when we update a graph (in order to prevent adding geometries from the tree).
When no graph are in edited, allow to select geometry nodes from the tree. In this case, I will add a rule to display only their metadata.

mfiaudrin-epsyl

The behavior works as expected in FE.

Once merged, I will call the new endpoint to get geometries to display and I will adapt the logic in the tree with geometry nodes.

mfiaudrin-epsyl · 2026-06-09T10:04:23Z

About show_empty_nodes parameter, it looks that /ids_info/geometry_overlay_nodes endpoint return nodes having no data even if show_empty_nodes is set to false. Is it normal ?

jwasikpsnc · 2026-06-09T10:06:33Z

It is normal for IMAS_CORE < 5.7 and IMAS-Python < 2.2.2

mfiaudrin-epsyl · 2026-06-10T07:55:41Z

Hi @jwasikpsnc,

Is it possible to also have is_geometry_node attribut in response of /ids_info/find_paths endpoint ?

I need it to get the information when we use the feature to search nodes

jwasikpsnc · 2026-06-10T07:59:57Z

Yes, it is.
I will do it

jwasikpsnc · 2026-06-11T11:10:14Z

@mfiaudrin-epsyl done

@olivhoenen
You can go further with review

jwasikpsnc · 2026-06-12T08:38:21Z

Hi @mfiaudrin-epsyl
Here's my proposition
Instead of having every geometry sub-structure separately:

    {
      "geometry_node": "imas:hdf5?path=/work/imas/shared/imasdb/ITER/3/135011/7#pf_active:0/coil[:]/element[:]/geometry/outline",
      "parameters": [
        "r",
        "z"
      ]
    },
    {
      "geometry_node": "imas:hdf5?path=/work/imas/shared/imasdb/ITER/3/135011/7#pf_active:0/coil[:]/element[:]/geometry/rectangle",
      "parameters": [
        "r",
        "z",
        "width",
        "height"
      ]
    },
    {
      "geometry_node": "imas:hdf5?path=/work/imas/shared/imasdb/ITER/3/135011/7#pf_active:0/coil[:]/element[:]/geometry/oblique",
      "parameters": [
        "r",
        "z",
        "length_alpha",
        "length_beta",
        "alpha",
        "beta"
      ]
    },

We could put them in single dict:

        {
      "geometry_node": "imas:hdf5?path=/work/imas/shared/imasdb/ITER/3/135011/7#pf_active:0/coil[:]/element[:]/geometry",
      "parameters": [
        "outline/r",
        "outline/z"
        "rectangle/r",
        "rectangle/z",
        "rectangle/width",
        "rectangle/height"
        "oblique/r",
        "oblique/z",
        "oblique/length_alpha",
        "oblique/length_beta",
        "oblique/alpha",
        "oblique/beta"
      ]
    },

Would it help?
Feel free to express any opinions, you know better what exactly do you need 🙌

mfiaudrin-epsyl · 2026-06-12T09:10:09Z

Hi @jwasikpsnc,

Thank you for the solution. It works for me. It's great because it doesn't change the structure of the response.

prasad-sawantdesai · 2026-06-12T10:11:15Z

You can also look for this implementation which covers all geometry types and its visualization.

https://github.com/iterorganization/IDStools/blob/develop/idstools/compute/pf_active.py#L32
https://github.com/iterorganization/IDStools/blob/develop/idstools/view/pf_active.py#L32

jwasikpsnc · 2026-06-12T10:33:41Z

@mfiaudrin-epsyl please check if the response is correct

mfiaudrin-epsyl · 2026-06-12T12:02:26Z

@jwasikpsnc It looks good but it doesn't works for thick_line geometry. I need all the parameters and currently I don't get the leaf nodes, just the get the nodes at 1st level:

jwasikpsnc · 2026-06-12T13:18:16Z

Should be fixed now

mfiaudrin-epsyl · 2026-06-12T14:12:25Z

Great, once it'll be merged I will finalize the FE with your updates.

jwasikpsnc added 5 commits May 26, 2026 10:32

Add geometry overlay nodes endpoint

7aa5f80

Add show_structures parameter to geometry overlay nodes endpoint

b1011a5

Geometry overlay nodes

97adfcc

Merge branch 'develop' into feature/show-only-filled-nodes_2

40a2e2d

Fix broken test

482c6a7

jwasikpsnc requested review from olivhoenen and prasad-sawantdesai as code owners June 2, 2026 06:30

jwasikpsnc marked this pull request as draft June 3, 2026 06:12

Alter geometry_nodes output

ad769d5

jwasikpsnc marked this pull request as ready for review June 8, 2026 11:13

jwasikpsnc requested a review from mfiaudrin-epsyl June 8, 2026 11:13

jwasikpsnc marked this pull request as draft June 8, 2026 11:55

Geometry overlay nodes update

77447a9

Add geometry info to ids_info/node_info/

f5809d7

mfiaudrin-epsyl approved these changes Jun 9, 2026

View reviewed changes

mfiaudrin-epsyl self-requested a review June 9, 2026 09:58

jwasikpsnc marked this pull request as ready for review June 9, 2026 10:06

mfiaudrin-epsyl approved these changes Jun 9, 2026

View reviewed changes

jwasikpsnc marked this pull request as draft June 10, 2026 08:00

Add is_geometry_node key to find_paths response

58a59e6

jwasikpsnc marked this pull request as ready for review June 11, 2026 11:09

jwasikpsnc marked this pull request as draft June 12, 2026 08:29

Alter geometry_nodes response structure

e05c426

jwasikpsnc marked this pull request as ready for review June 12, 2026 10:33

Fix

45776b8

mfiaudrin-epsyl mentioned this pull request Jun 18, 2026

Geometry overlay #74

Draft

Conversation

jwasikpsnc commented Jun 2, 2026

Uh oh!

jwasikpsnc commented Jun 2, 2026

Uh oh!

read-the-docs-community Bot commented Jun 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Documentation build overview

Uh oh!

mfiaudrin-epsyl commented Jun 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jwasikpsnc commented Jun 3, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 3, 2026

Uh oh!

maarten-ic commented Jun 3, 2026

Uh oh!

imbeauf commented Jun 3, 2026

Uh oh!

imbeauf commented Jun 3, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 3, 2026

Uh oh!

imbeauf commented Jun 3, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 3, 2026

Uh oh!

jwasikpsnc commented Jun 8, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 8, 2026

Uh oh!

jwasikpsnc commented Jun 8, 2026

Uh oh!

maarten-ic commented Jun 8, 2026

Uh oh!

jwasikpsnc commented Jun 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mfiaudrin-epsyl commented Jun 9, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 9, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 9, 2026

Uh oh!

jwasikpsnc commented Jun 9, 2026

Uh oh!

maarten-ic commented Jun 9, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 9, 2026

Uh oh!

mfiaudrin-epsyl left a comment

Choose a reason for hiding this comment

Uh oh!

mfiaudrin-epsyl commented Jun 9, 2026

Uh oh!

jwasikpsnc commented Jun 9, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 10, 2026

Uh oh!

jwasikpsnc commented Jun 10, 2026

Uh oh!

jwasikpsnc commented Jun 11, 2026

Uh oh!

jwasikpsnc commented Jun 12, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 12, 2026

Uh oh!

prasad-sawantdesai commented Jun 12, 2026

Uh oh!

jwasikpsnc commented Jun 12, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 12, 2026

Uh oh!

jwasikpsnc commented Jun 12, 2026

Uh oh!

mfiaudrin-epsyl commented Jun 12, 2026

Uh oh!

Reviewers

Assignees

read-the-docs-community Bot commented Jun 2, 2026 •

edited

Loading

mfiaudrin-epsyl commented Jun 2, 2026 •

edited

Loading

jwasikpsnc commented Jun 9, 2026 •

edited

Loading