Skip to content
Feeding a clanker? Grab this page as raw .md →

ScatterWidget API#

ScatterWidget is a paintable scatter canvas: pick a class color, drag the brush, and the points you paint come back as data (or as a DataFrame via data_as_pandas). It is the fastest way to invent a small labeled 2D dataset for testing a classifier, sanity-checking a clustering algorithm, or drawing the pathological case you want to explain. Up to four classes, each with its own color and label, plus a batch key per stroke.

See also: SplineDraw for the same canvas with a Python-fitted curve on top, ChartMultiSelect for labeling data you already have, and ScatterLog for accumulating computed points rather than drawn ones.

Note: ScatterWidget is provided by the drawdata package and re-exported here for convenience. You can use it via from wigglystuff import ScatterWidget or from drawdata import ScatterWidget.

Bases: AnyWidget

Interactive scatter drawing widget for painting multi-class 2D data points.

The widget renders a D3-powered SVG canvas where you can paint data points using a configurable brush. Points are grouped by class (color) and batch, making it easy to generate labeled 2D datasets interactively.

Examples:

import marimo as mo
from wigglystuff import ScatterWidget

widget = mo.ui.anywidget(ScatterWidget(n_classes=3))
widget

Create a ScatterWidget.

Parameters:

Name Type Description Default
n_classes int

Number of point classes (1-4). Each class gets a distinct color and label.

 4
brushsize int

Initial brush radius in pixels.

 40
width int

SVG viewBox width in pixels.

 800
height int

SVG viewBox height in pixels.

 400
**kwargs Any

Forwarded to anywidget.AnyWidget.

 {}

Raises:

Type Description
ValueError

If n_classes is not between 1 and 4.
Source code in wigglystuff/scatter_widget.py 
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
def __init__(
    self,
    *,
    n_classes: int = 4,
    brushsize: int = 40,
    width: int = 800,
    height: int = 400,
    **kwargs: Any,
) -> None:
    """Create a ScatterWidget.

    Args:
        n_classes: Number of point classes (1-4). Each class gets a
            distinct color and label.
        brushsize: Initial brush radius in pixels.
        width: SVG viewBox width in pixels.
        height: SVG viewBox height in pixels.
        **kwargs: Forwarded to ``anywidget.AnyWidget``.

    Raises:
        ValueError: If *n_classes* is not between 1 and 4.
    """
    if not 1 <= n_classes <= 4:
        raise ValueError("n_classes must be between 1 and 4")
    super().__init__(
        n_classes=n_classes,
        brushsize=brushsize,
        width=width,
        height=height,
        **kwargs,
    )

data_as_X_y property #

data_as_X_y

Return the drawn points as a (X, y) tuple of numpy arrays.

When configured with n_classes=1 (regression scenario), returns X with shape (n, 1) containing x values and y as the target array.

When configured with n_classes>1 (classification scenario), returns X with shape (n, 2) containing (x, y) coordinates and y as a list of color strings.

data_as_pandas property #

data_as_pandas

Return the drawn points as a :class:pandas.DataFrame.

Columns: x, y, color, label, batch.

data_as_polars property #

data_as_polars

Return the drawn points as a :class:polars.DataFrame.

Columns: x, y, color, label, batch.

Synced traitlets#

Traitlet Type Notes
data list[dict] List of drawn points with x, y, color, label, batch keys.
brushsize int Brush radius in pixels (default: 40).
width int SVG viewBox width in pixels (default: 800).
height int SVG viewBox height in pixels (default: 400).
n_classes int Number of point classes, 1-4 (default: 4).