Add uPlot Element

[NvNielen]

Motivation

This PR introduces support for uPlot, a high-performance, memory-efficient 2D plotting library. This was originally requested in #2669, but also motivated by issues related to large dataset plotting such as #3340. Compared to other plotting libraries, it has a smaller memory footprint and less overhead on rendering, which makes it ideal for processing large datasets at high refresh rates.

Implementation

The core implementation wraps uPlot as a NiceGUI element, based on uplot-vue, which exposes uPlot's config and data update mechanisms.

The design anticipates plugin support, with plans to add Wheel Zoom & Drag and legend-as-a-tooltip plugins .

The update logic should be structured to allow zoom/drag state to persist across data updates, as demonstrated in #2669 .
Some important technical considerations include keeping the API simple and extensible, down sampling support, plugin support and implementing tests.

Progress

• uPlot base element and functionality
• Persist zoom/drag across updates
• Add plugin support
• Add downsampling
• Pytests, should include rendering, updates, and plugins.
• Add documentation and provide clear, runnable demos, since uPlot's documentation is limited.

[NvNielen]

Any input and/or feedback very much appreciated!

[falkoschindler]

Thanks for this PR! uPlot is a great choice for high-performance charting, and this is a solid start.

We tested the element successfully with the following code:

from nicegui import ui
from nicegui.elements.uplot.uplot import UPlot
import math

x_data = list(range(100))
y_data = [math.sin(i * 0.1) for i in range(100)]

chart = UPlot({
    'width': 800,
    'height': 400,
    'title': 'Sine Wave',
    'series': [{}, {'stroke': 'red', 'label': 'sin(x)'}],
}, data=[x_data, y_data])

@ui.button('Add Point').on_click
def add_point():
    x_data.append(len(x_data))
    y_data.append(math.sin(len(y_data) * 0.1))
    chart.update_data([x_data, y_data])

ui.run()

A few things we noticed while looking at the code:

• The other ESM-based elements (plotly, echart, codemirror) have an __init__.py that re-exports the class, plus a ui.plotly(...) / ui.echart(...) registration. Something to keep in mind for when this gets wired up to the ui namespace.

• uplot.py imports NumPy unconditionally, but it's only used in type hints. A TYPE_CHECKING guard would avoid making NumPy a hard dependency:

  from typing import TYPE_CHECKING
  if TYPE_CHECKING:
      import numpy as np

Looking forward to seeing this evolve!