Positions

position_beeswarm

class plotnine_extra.position_beeswarm(method='swarm', cex=1.0, side=0, priority='ascending', dodge_width=None, corral='none', corral_width=0.9)[source]

Bases: position

Jitter points using the beeswarm algorithm

Points are arranged so that they do not overlap, producing a layout that resembles a beeswarm. The resulting shape gives a good indication of the data distribution while showing every individual observation.

Parameters:

  • method (str) –

    Algorithm for arranging points.

    • "swarm" (default): place in order, shift sideways the minimum amount to avoid overlap.

    • "compactswarm": greedy strategy for tighter packing.

    • "center" / "centre": square grid, centred.

    • "hex": hexagonal grid.

    • "square": regular square grid.

  • cex (float) – Scaling factor for point spacing (1-3 recommended).

  • side (int) – 0: both sides (default), 1: right/up only, -1: left/down only.

  • priority (str) – Order in which points are placed: "ascending" (default), "descending", "density", "random", "none".

  • dodge_width (Optional[float]) – Amount of dodge between aesthetic groups.

  • corral (str) – How to handle runaway points: "none" (default), "gutter", "wrap", "random", "omit".

  • corral_width (float) – Width of the corral region.

REQUIRED_AES = {'x', 'y'}

Aesthetics required for the positioning

setup_params(data)[source]

Verify, modify & return a copy of the params.

Parameters:

data (pd.DataFrame)

Return type:

dict

classmethod compute_panel(data, scales, params)[source]

Positions must override this function

Notes

Make necessary adjustments to the columns in the dataframe.

Create the position transformation functions and use self.transform_position() do the rest.

See also

plotnine.position_jitter.compute_panel

Parameters:

  • data (pd.DataFrame)

  • scales (pos_scales)

  • params (dict)

Return type:

pd.DataFrame

position_quasirandom

class plotnine_extra.position_quasirandom(method='quasirandom', width=None, varwidth=False, bandwidth=0.5, nbins=None, dodge_width=None)[source]

Bases: position

Jitter points using quasi-random noise to reduce overplotting

Uses density-aware quasi-random jittering (van der Corput sequence) so that the point cloud reflects the underlying data distribution, similar to a violin plot but with individual points.

Parameters:

  • method (str) – "quasirandom" (default) uses a van der Corput sequence; "pseudorandom" uses uniform random jitter.

  • width (Optional[float]) – Maximum jitter width. If None, 0.4 * resolution of the data axis is used.

  • varwidth (bool) – If True, scale the width of each group proportionally to its size relative to the largest group.

  • bandwidth (float) – Bandwidth adjustment for the internal kernel density estimate. Values < 1 yield a tighter fit.

  • nbins (Optional[int]) – Number of bins for density estimation (passed through but currently unused; bandwidth controls smoothing).

  • dodge_width (Optional[float]) – Amount by which to dodge groups that share the same position. None means no dodging.

REQUIRED_AES = {'x', 'y'}

Aesthetics required for the positioning

setup_params(data)[source]

Verify, modify & return a copy of the params.

Parameters:

data (pd.DataFrame)

Return type:

dict

classmethod compute_panel(data, scales, params)[source]

Positions must override this function

Notes

Make necessary adjustments to the columns in the dataframe.

Create the position transformation functions and use self.transform_position() do the rest.

See also

plotnine.position_jitter.compute_panel

Parameters:

  • data (pd.DataFrame)

  • scales (pos_scales)

  • params (dict)

Return type:

pd.DataFrame

position_disjoint_ranges

class plotnine_extra.position_disjoint_ranges(extend=0.0, stepsize=1.0)[source]

Bases: position

Vertically stack overlapping x intervals into disjoint rows.

Parameters:

  • extend (float, default 0) – Padding added to each interval before deciding overlap. Positive values force more vertical separation.

  • stepsize (float, default 1) – Vertical spacing between rows.

REQUIRED_AES = {'xmax', 'xmin'}

Aesthetics required for the positioning

setup_params(data)[source]

Verify, modify & return a copy of the params.

classmethod compute_layer(data, params, layout)[source]

Compute position for the layer in all panels

Positions can override this function instead of compute_panel if the position computations are independent of the panel. i.e when not colliding

Parameters:

data (pd.DataFrame)

Return type:

pd.DataFrame

position_lineartrans

class plotnine_extra.position_lineartrans(scale=(1.0, 1.0), shear=(0.0, 0.0), angle=0.0, M=None)[source]

Bases: position

Apply an affine 2x2 linear transformation to x and y.

Parameters:

  • scale (tuple of float, default (1, 1)) – Scale factors (sx, sy).

  • shear (tuple of float, default (0, 0)) – Shear factors (shx, shy).

  • angle (float, default 0) – Rotation in degrees, applied after scale + shear.

  • M (array-like, optional) – A 2x2 matrix that overrides scale / shear / angle if supplied.

REQUIRED_AES = {'x', 'y'}

Aesthetics required for the positioning

setup_params(data)[source]

Verify, modify & return a copy of the params.

classmethod compute_layer(data, params, layout)[source]

Compute position for the layer in all panels

Positions can override this function instead of compute_panel if the position computations are independent of the panel. i.e when not colliding

Parameters:

data (pd.DataFrame)

Return type:

pd.DataFrame