nbimplot 0.1.4

ImPlot-powered Jupyter plotting widget for interactive, million-point visualizations.

nbimplot

Jupyter-native plotting focused on high interactivity and large time-series support.

Current implementation includes:

• Notebook-only widget canvas rendering
• Binary transport (numpy -> bytes -> wasm memory)
• WASM core for plot state, autoscale, and min/max LOD generation
• Strict WASM runtime requirement (JS fallback disabled)
• ImPlot-required mode (prefer_implot=True, strict_wasm=True only)
• Lifecycle-safe cleanup for repeated cell execution
• Subplots via nbimplot.Subplots
• Expanded primitive APIs: scatter, bubbles, stairs, stems, digital, bars, bar_groups, bars_h, shaded, error_bars, error_bars_h, inf_lines, vlines, hlines, histogram, histogram2d, heatmap, image, pie_chart, text, annotation, dummy

Quick start

import numpy as np
import nbimplot as ip

y = np.sin(np.linspace(0, 100, 1_000_000, dtype=np.float32))

p = ip.Plot(width=900, height=450, title="Signal")
h = p.line("mid", y)
p.show()

h.set_data(y * 0.8)
p.render()

API

import nbimplot as ip

p = ip.Plot(width=900, height=450, title="Signal")
h = p.line("mid", y, x_axis="x1", y_axis="y1", subplot_index=0)
h.set_data(y_new)
p.render()
p.show()

prefer_implot=False and strict_wasm=False are not supported.

Additional controls:

• p.set_view(x_min, x_max, y_min, y_max)
• p.autoscale()
• p.set_axis_scale(x="linear|log", y="linear|log")
• p.set_plot_flags(...) for native ImPlot flags (no_legend, no_menus, etc.)
• p.set_axis_state("x2|x3|y2|y3", enabled=True|False, scale="linear|log|time")
• p.set_axis_label(...), p.set_axis_format(...), p.set_axis_ticks(...), p.clear_axis_ticks(...)
• p.set_axis_limits_constraints(...), p.set_axis_zoom_constraints(...), p.set_axis_link(...)
• p.set_secondary_axes(x2=..., x3=..., y2=..., y3=...)
• p.set_time_axis("x1|x2|x3|y1|y2|y3")
• p.set_colormap("Deep|Dark|Pastel|Paired|Viridis|Plasma|Hot|Cool|Pink|Jet|Twilight|RdBu|BrBG|PiYG|Spectral|Greys")
• Most primitive APIs support x_axis="x1|x2|x3" and y_axis="y1|y2|y3"
• p.line(..., color=\"#3b82f6\", line_weight=2.0, marker=\"circle\", marker_size=5.0) and handle.set_style(...)
• p.stream_line(..., capacity=N) and handle.append(...) for in-place ring-buffer style updates
• p.hide_next_item() to hide the next added series/primitive without removing it
• p.heatmap(..., label_fmt="%.2f", scale_min=None, scale_max=None) and label_fmt="" to hide cell-value text
• p.heatmap(..., show_colorbar=True, colorbar_label="Intensity", colorbar_format="%g", colorbar_flags=0)
• p.histogram2d(..., show_colorbar=True, colorbar_label="Count", colorbar_format="%g", colorbar_flags=0)
• p.image(..., bounds=((x0, y0), (x1, y1)), uv0=(0, 0), uv1=(1, 1)) for true ImPlot::PlotImage rendering (z can be 2D grayscale or H x W x 3/4 RGB(A))
• p.tag_x(...), p.tag_y(...), p.colormap_slider(...), p.colormap_button(...), p.colormap_selector(...)
• p.drag_drop_plot(...), p.drag_drop_axis(...), p.drag_drop_legend(...)
• p.on_perf_stats(callback, interval_ms=500) for live FPS/WASM timings
• p.on_tool_change(callback) for drag tool updates
• p.on_selection_change(callback) for box-selection/query updates

Interactive tools:

• p.drag_line_x(...)
• p.drag_line_y(...)
• p.drag_point(...)
• p.drag_rect(...)

Subplots (Native ImPlot)

import nbimplot as ip
import numpy as np

sp = ip.Subplots(
    2,
    2,
    title="Dashboard",
    width=1000,
    height=700,
    link_rows=True,
    link_cols=True,
    share_items=True,
)
sp.subplot(0, 0).line("line", np.sin(np.linspace(0, 20, 5000, dtype=np.float32)))
sp.subplot(0, 1).scatter("pts", np.random.randn(2000).astype(np.float32))
sp.subplot(1, 0).bars("bars", np.abs(np.random.randn(120)).astype(np.float32))
sp.subplot(1, 1).histogram("hist", np.random.randn(20_000).astype(np.float32), bins=40)
sp.show()

Subplots(...) supports axis-linking controls:

• link_rows: link y-axis within each row
• link_cols: link x-axis within each column
• link_all_x: link x-axis for all cells
• link_all_y: link y-axis for all cells
• Additional compatibility flags: share_items, no_legend, no_menus, no_resize, no_align, col_major

Aligned-group helper:

• ip.AlignedPlots(rows, cols, group_id="aligned", vertical=True, ...)

Example Notebook

• notebooks/nbimplot_examples.ipynb
• notebooks/nbimplot_api_gallery.ipynb
• notebooks/nbimplot_benchmarks.ipynb

Troubleshooting

If JupyterLab shows:

Failed to load model class 'AnyModel' from module 'anywidget'

this is typically an environment/frontend mismatch (server env vs kernel env) or an old/stale anywidget install. Fix with:

python -m pip install -U "nbimplot>=0.1.2" "anywidget>=0.9.21" ipywidgets jupyterlab_widgets

Then fully restart the JupyterLab server (not just kernel restart).

Quick verification in a notebook cell:

import nbimplot as ip, anywidget, sys
print("python:", sys.executable)
print("anywidget:", anywidget.__version__)
print("has Plot:", hasattr(ip, "Plot"))

Build WASM core

Prerequisites

• Emscripten SDK (emcmake, emcc)
• CMake >= 3.20

Build (WASM core only)

scripts/build_wasm.sh

Outputs are written to nbimplot/wasm/:

• nbimplot_wasm.js
• nbimplot_wasm.wasm

If outputs are missing or invalid, the widget reports a runtime error.

Build with ImPlot sources

ImPlot is required for runtime. Build with Dear ImGui + ImPlot sources:

NBIMPLOT_WITH_IMPLOT=ON \
NBIMPLOT_IMGUI_DIR=/path/to/imgui \
NBIMPLOT_IMPLOT_DIR=/path/to/implot \
scripts/build_wasm.sh

If you cloned deps into third_party/imgui and third_party/implot, this also works:

NBIMPLOT_WITH_IMPLOT=ON scripts/build_wasm.sh

Notes:

• wasm/core/nbimplot_implot_layer.cpp creates ImGui/ImPlot contexts when compiled with NBIMPLOT_WITH_IMPLOT=1.
• The final WebGL backend draw hookup is intentionally isolated behind the ImPlotLayer::render(...) seam.

Performance behavior

• Raw rendering path when visible points <= 3 * pixel_width
• Min/max bucket LOD path when visible points > 3 * pixel_width
• LOD decision and output generation happen in WASM