frame_fit

A tiny backend-independent 2D viewport algebra for fitting, comparing, and tracking geometric objects, where scale and translation are independently resolvable across views.

The package does not plot or render. It returns numerical viewports that other systems can consume.

import numpy as np

from frame_fit import Target2D, fit

points = np.array([[-2.0, -1.0], [2.0, 1.0]])

view = fit(points, target=Target2D(200, 100), padding=0.1)

view.xlim
view.ylim
view.scale
view.transform

Core Ideas

geometry or bounds
    -> viewport request
    -> optional scale constraints
    -> resolved viewport
    -> caller's plotting or graphics backend

The solver preserves isotropic scale by default. It does not independently stretch x and y.

Convenience API

from frame_fit import Bounds2D, Target2D, fit, fit_shared_scale, fit_union

source = Bounds2D(-2, 2, -1, 1)
mapped = Bounds2D(-2.1, 2.1, -1.2, 1.2)
target = Target2D(600, 600)

source_view = fit(source, target=target, padding=0.05)
overlay_view = fit_union([source, mapped], target=target, padding=0.05)
source_cmp, mapped_cmp = fit_shared_scale([source, mapped], target=target, padding=0.05)

General Solver

from frame_fit import Target2D, ViewRequest, resolve_views

views = resolve_views(
    [
        ViewRequest(
            id="source-comparison",
            geometry=source_points,
            target=Target2D(600, 600),
            scale_group="comparison",
        ),
        ViewRequest(
            id="mapped-comparison",
            geometry=mapped_points,
            target=Target2D(600, 600),
            scale_group="comparison",
        ),
    ]
)

assert views["source-comparison"].scale == views["mapped-comparison"].scale

Windows

from frame_fit import ParameterWindow, Selection, bounds

window = ParameterWindow(start=10.0, stop=20.0)
selected = Selection(points, window, parameters=t)

windowed_bounds = selected.bounds()

Development Notes

The adaptive sampling layer for expensive mappings and variable-resolution curves is documented in:

docs/adaptive_sampling_layer.md

Current resume status and pending work are documented in:

docs/resume_status.md