cpt-anywidget

anywidget viewers for cone penetration tests (CPT) and geotechnical boreholes, rendered with plain D3. Built for notebook use (marimo, Jupyter). Every viewer shares a zoomable vertical axis in depth below surface or elevation, so soundings, borehole logs, and interpretations line up.

Full documentation: bedrock.engineer/docs/cpt-anywidget

Viewers

• CPTViewer — one CPT: measurement channels (qc, fs, Rf, u1/u2, inclination, …) against the vertical axis, plus optional layer columns: a nearby borehole log, read-only interpretation columns (e.g. Robertson, Lengkeek), and an editable layer column synced back to Python.
• BoreholeViewer — a single borehole log with proportional soil-composition bands and hatch patterns; its layers trait takes plain dicts from any source, and layers_from_bhrgt (brodata BHR-GT) and layers_from_bore (pygef GEF/BRO XML) convert boreholes already in the Dutch BRO soil vocabulary.
• ProfileViewer — multiple CPTs side by side along a chainage axis (a cross-section); chainage computes along-profile distances from map coordinates.

Install

From PyPI, with uv (recommended):

uv add cpt-anywidget

or with pip: pip install cpt-anywidget.

Usage

import polars as pl
from cpt_anywidget import CPTViewer, Channel

df = pl.DataFrame({
    "depth": [0.0, 0.02, 0.04],        # or "nap" for elevation
    "coneResistance": [0.1, 0.4, 0.9],  # MPa
    "localFriction": [0.01, 0.02, 0.02],
})

CPTViewer(df, vertical="depth", channels=["coneResistance", "localFriction"])

data is tidy columns — a polars or pandas DataFrame or a dict of columns (lists, tuples, numpy arrays) — one row per depth sample. The widget never parses file formats or converts units: loaders (e.g. brodata) normalize upstream. The Dutch BRO column names get built-in display defaults; any other column can be bound with Channel(key, label=…, unit=…, color=…, side=…). Custom vertical datums work the same way via Vertical(key, label=…, up=…).

Annotations, overlays, and axis limits are plain traits and track the zoom. See Interactions for the pointer and keyboard gestures.

Interactions

Zoom and pan — all viewers

• To zoom the vertical axis, hold the Ctrl key (Cmd on macOS) and turn the mouse wheel. Pinching on a trackpad or touch device zooms too, without keys.
• To pan a zoomed axis, drag in the plot area.
• To zoom to a range, hold the Shift key and drag along the axis.
• To reset the zoom, double-click in the plot area.
• Move the pointer across the plot to read the values at that depth.
• The mouse wheel without a modifier key scrolls the notebook, not the chart.

Edit layers — the CPTViewer edit column

• To move a boundary, drag it. A boundary stops at the minimum layer thickness.
• To split a layer, click in the lane beside the column. A dashed line previews where the new boundary goes.
• To merge two layers, move the pointer near their boundary in the lane and click the × it offers. The upper layer keeps its class.
• To set the class of a layer, click the layer. A pie menu opens. Click a wedge, or walk the wedges with the arrow keys and push Enter.
• As a fast path, press the layer, drag toward a wedge, and release.
• To close the pie menu, push the Escape key or click outside the menu.
• Each edit goes back to Python through the editedLayers trait.

Select a strip — ProfileViewer

• To select a CPT strip, click it. The name shows in the selected trait in Python.
• To deselect the strip, click it again.

Development

Front end (TypeScript, js/) builds into src/cpt_anywidget/static/:

npm install
npm run build     # or: npm run dev (watch mode)

Set ANYWIDGET_HMR=1 so JS edits hot-reload in the browser. Sample files (BRO XML, GEF, AGS) live in examples/.