Problem with PolyData and add_point_labels

[DomenicoGaudioso]

Hi everyone,

I'm attempting to integrate some three-dimensional visualization examples using PyVista into a Streamlit application. I'm encountering difficulties with two specific examples:

1. Creating a truss (https://docs.pyvista.org/version/stable/examples/00-load/create-truss#sphx-glr-examples-00-load-create-truss-py) I'm trying to adapt this example to display a truss in a Streamlit application. I have problem with visualization with using PolyData.

2. Calculating distance along a spline (https://docs.pyvista.org/version/stable/examples/02-plot/distance-along-spline#sphx-glr-examples-02-plot-distance-along-spline-py) I have problem with visualization when use the add_point_labels function to label points along a spline.

I would appreciate if someone could offer some advice on resolving these issues and making the examples work in Streamlit. Thank you in advance for your help!

Domenico

[edsaac]

Hi @DomenicoGaudioso, iirc, rendering lines as tubes and points as spheres are limitations from vtk-js, the library used to bring PyVista plotters to the browser: https://github.com/pyvista/pyvista/issues/5444

For example 1, my alternative would be to draw individual lines and points in the plotter:

Code:

import numpy as np
import pyvista as pv
import streamlit as st
from stpyvista import stpyvista
from matplotlib.cm import get_cmap


## Define geometry
nodes = [
    [0.0, 0.0, 0.0],
    [0.0, 1.0, 0.0],
    [4.0, 3.0, 0.0],
    [4.0, 0.0, 0.0],
    [0.0, 1.0, 2.0],
    [4.0, 1.0, 2.0],
    [4.0, 3.0, 2.0],
]

edges = [
    [0, 4],
    [1, 4],
    [3, 4],
    [5, 4],
    [6, 4],
    [3, 5],
    [2, 5],
    [5, 6],
    [2, 6],
]

## Generate colors for edges
cmap = get_cmap("hsv")
colors = map(cmap, np.linspace(0, 1, len(edges)))

## Create Plotter
plotter = pv.Plotter(window_size=(300, 300))
plotter.background_color = "salmon"

## Add nodes as points
plotter.add_points(np.array(nodes), point_size=20, color="purple")

## Add edges as lines
for edge, color in zip(edges, colors):
    i, j = edge
    line = np.array([nodes[i], nodes[j]])
    plotter.add_lines(line, color=color, width=10)

plotter.view_isometric()


## Send to streamlit
st.title("Truss")

stpyvista(
    plotter,
    panel_kwargs=dict(
        orientation_widget=True,
    ),
)

[DomenicoGaudioso]

Thanks for the reply, I am also using this alternative but I noticed that it is heavier (slowness in image processing). Thank you for the feedback

[edsaac]

Perhaps it's because I was looping through each edge element and adding it to the pyvista Plotter. We could try a couple things to improve performance:

• Decorate with st.cache_resource a function in charge of generating the Plotter
• Assembling a single array of edges/coordinates to pass to Plotter.add_lines
• Pass a key to stpyvista to avoid re-rendering of the 3D view at each interaction with the web-app

Code:

import numpy as np
import pyvista as pv
import streamlit as st
from stpyvista import stpyvista


@st.cache_resource
def mesh_to_plotter(nodes: list[list[float]], edges: list[list[int]]):
    ## Create Plotter
    plotter = pv.Plotter(window_size=(300, 300))
    plotter.background_color = "salmon"

    ## Add nodes as points
    plotter.add_points(np.array(nodes), point_size=20, color="purple")

    ## Assemble edges as single array
    coord_edges = (
        np.array([[nodes[i], nodes[j]] for (i, j) in edges])
        .flatten()
        .reshape((2 * len(edges), 3))
    )

    plotter.add_lines(coord_edges, color="k", width=10)
    plotter.view_isometric()

    return plotter


def main():
    ## Define geometry
    nodes = [
        [0.0, 0.0, 0.0],
        [0.0, 1.0, 0.0],
        [4.0, 3.0, 0.0],
        [4.0, 0.0, 0.0],
        [0.0, 1.0, 2.0],
        [4.0, 1.0, 2.0],
        [4.0, 3.0, 2.0],
    ]

    edges = [
        [0, 4],
        [1, 4],
        [3, 4],
        [5, 4],
        [6, 4],
        [3, 5],
        [2, 5],
        [5, 6],
        [2, 6],
    ]

    ## Create plotter
    plotter = mesh_to_plotter(nodes, edges)

    ## Send to streamlit
    st.title("Truss")

    stpyvista(
        plotter,
        panel_kwargs=dict(
            orientation_widget=True,
        ),
        key="stpv_truss",
    )


if __name__ == "__main__":
    main()

[DomenicoGaudioso]

Great idea. One question: How can I set the color of the line based on, say, the 'Z' values of the points? In this example, they use 'cmap'. https://docs.pyvista.org/version/stable/examples/00-load/create-truss#sphx-glr-examples-00-load-create-truss-py

[DomenicoGaudioso]

Is it possible to add a list of colors in color? I tried but with no success.

  Error:  Must be a string, rgb(a) sequence, or hex color string.  For example:
            color='white'
            color='w'
            color=[1.0, 1.0, 1.0]
            color=[255, 255, 255]
            color='#FFFFFF'

[edsaac]

I don't think that will be possible because a single lines actor is generated from a single pv.Plotter.add_lines call.