gladly-jupyter

Jupyter widget for Gladly — GPU-accelerated plotting via WebGL, driven directly from Python.

Large columns (gigabytes) are streamed to the browser over HTTP from kernel memory rather than sent through the Jupyter comm, so there is no practical data-size limit. Gladly itself is loaded from CDN — no JavaScript build step is required.

Requirements

Python ≥ 3.8
JupyterLab ≥ 4 or Jupyter Notebook ≥ 7
A WebGL-capable browser

Installation

pip install gladly-jupyter
# with optional Pint unit support:
pip install "gladly-jupyter[pint]"

From source:

pip install ./gladly_jupyter

Quick example

import gladly_jupyter as gl
import pandas as pd
import numpy as np

t  = np.linspace(0, 10, 100_000, dtype="float32")
df = pd.DataFrame({"time": t, "signal": np.sin(t)})

config = {
    "layers": [{"points": {"xData": "input.time", "yData": "input.signal"}}],
    "axes":   {
        "xaxis_bottom": {"label": "Time (s)"},
        "yaxis_left":   {"label": "Signal"},
    },
}

gl.Plot(config, df)

A single DataFrame is automatically wrapped and placed under the key input, so its columns are referenced as "input.colname" in the config. For a DataGroup you use the name you assigned each source: "raw.colname".

Multiple data sources in one plot

config = {
    "layers": [
        {"points": {"xData": "raw.time",      "yData": "raw.signal"}},
        {"points": {"xData": "filtered.time", "yData": "filtered.signal"}},
    ],
}

gl.Plot(config, gl.DataGroup({"raw": df_raw, "filtered": df_filtered}))

Linked axes across multiple plots

p1 = gl.Plot({"layers": [{"points": {"xData": "input.time", "yData": "input.depth"}}]},
             gl.Data(df1, quantity_kinds={"time": "s"}))

p2 = gl.Plot({"layers": [{"points": {"xData": "input.time", "yData": "input.velocity"}}]},
             gl.Data(df2, quantity_kinds={"time": "s"}))

gl.PlotGroup({"depth": p1, "velocity": p2})   # time axes auto-linked

Manual axis linking (cross-plot, no PlotGroup required):

link = gl.link_axes(p1.axis.xaxis_bottom, p2.axis.xaxis_bottom)
link.unlink()   # tear down later

Registering quantity kinds and layer types

# Register before creating plots — picked up at construction time
gl.register_axis_quantity_kind("frequency", label="Frequency (Hz)", scale="log")

gl.register_layer_type("myLayer", {
    "vertShader": "attribute vec2 a_position; void main() { ... }",
    "fragShader": "void main() { gl_FragColor = vec4(1.0); }",
    "schema":     gl.js("(data) => ({ type: 'object', properties: {} })"),
})

Live config and schema

config = plot.get_config()   # includes current zoom domains, blocks until JS responds
schema = plot.schema()       # JSON Schema for the config dict

# Async variants:
config = await plot.async_get_config()
schema = await plot.async_schema()

Documentation

Full API reference in docs/: