Programmatic Lasso/Polygon Selection #235

name: 'Enhancement'
about: 'Add programmatic lasso selection API'
labels: 'Enhancement 🎉'

Goal

Enable programmatic selection of points using polygons in data space, complementing the existing interactive lasso tool.

Use Cases

• Annotation replay: load an existing session and re-apply the saved lasso region so the same points get selected again.
• Predefined regions: select a known ROI polygon (e.g., “this cluster boundary”) as part of a workflow.
• Reproducible selections: store the polygon vertices and get consistent results across runs.

Current Limitation

scatterplot.select() only accepts point indices, requiring users to manually implement point-in-polygon algorithms:

scatterplot.select([0, 5, 12, 47]); // Requires pre-computing which points are in region

Proposed Solution

Extend select() to accept polygon vertices in data space:

// Select points within a triangle (data coordinates)
scatterplot.select([
  [10, 20],
  [50, 80],
  [90, 30]
]);

// Works with merge/remove modes
scatterplot.select(rectanglePolygon, { merge: true });

Backward Compatible: Automatically detects polygon vs indices based on input type.

Implementation Details

Reuses existing findPointsInLasso()

const isPolygon = (arg) =>
  Array.isArray(arg) && arg.length >= 3 &&
  Array.isArray(arg[0]) && arg[0].length === 2;

const select = (pointIdxsOrPolygon, options = {}) => {
  if (isPolygon(pointIdxsOrPolygon)) {
    const polygonGl = transformDataToGl(pointIdxsOrPolygon);
    const points = findPointsInLasso(polygonGl);
    select(points, options); // Recurse with indices
    return;
  }
  // Existing logic...
};

Checklist

• No duplicate issues
• Use cases described
• Backward compatible
• Working demo included
• Implementation prototyped