Post-processing

optixde.post provides Matplotlib-based helpers for consistent, mask-aware scientific figures. Matplotlib is imported lazily: the numerical package can be used without it until a plotting symbol is requested.

Plot style

PlotStyle(
    cmap="jet",
    fontsize=11,
    title_size=12,
    label_size=11,
    tick_size=10,
    cbar_tick_size=10,
    figure_dpi=300,
    tight_layout=True,
    axes_aspect="equal",
)

DEFAULT_STYLE is the module default.

set_plot_style(style=DEFAULT_STYLE)

This updates global Matplotlib rcParams. Use a local Matplotlib context if other figures in the same process require different global settings.

Single field

plot_cloud(
    U,
    *,
    ax=None,
    extent=None,
    mask=None,
    title=None,
    xlabel="x",
    ylabel="y",
    unit=None,
    cmap=None,
    mode="imshow",
    interp="nearest",
    origin="lower",
    aspect="equal",
    nlevels=64,
    robust=False,
    robust_q=(1.0, 99.0),
    vmin=None,
    vmax=None,
    symmetric=False,
    center=0.0,
    cb=True,
    cb_label=None,
    cb_ticks="minmax",
    cb_nticks=None,
    cb_format="auto",
    annotate_minmax=True,
    fontsize=None,
    save=None,
    dpi=None,
    tight=None,
)

U must be a two-dimensional array. NaN and infinite values are masked automatically. If mask is supplied, True means hidden and its shape must equal U.shape.

The return value is:

fig, ax, mappable, colorbar_or_none, (vmin, vmax)

Color-scale controls:

• robust=True uses robust_q percentiles;
• symmetric=True expands the range symmetrically around center;
• explicit vmin and vmax take precedence;
• cb_ticks accepts minmax, minmidmax, min0max, linspace, or an explicit sequence.

fig, ax, _, _, limits = plot_cloud(
    error,
    extent=(0.0, Lx, 0.0, Ly),
    title="Pointwise error",
    cmap="coolwarm",
    symmetric=True,
    center=0.0,
    save="error.pdf",
)

Matplotlib selects the output format from the extension, so save="figure.pdf" creates vector PDF output.

Field panels

plot_cloud_grid(
    U_list,
    *,
    titles=None,
    extent=None,
    mask=None,
    layout=None,
    figsize=None,
    cmap=None,
    mode="imshow",
    interp="nearest",
    origin="lower",
    aspect="equal",
    nlevels=64,
    robust=False,
    robust_q=(1.0, 99.0),
    vmin=None,
    vmax=None,
    symmetric=False,
    center=0.0,
    cb_shared=True,
    cb_label=None,
    unit=None,
    cb_ticks="minmax",
    cb_nticks=None,
    cb_format="auto",
    annotate_minmax=True,
    fontsize=None,
    save=None,
    dpi=None,
    tight=None,
)

Returns:

fig, axes, mappables, colorbar_or_none, (vmin, vmax)

By default, the scale is computed across all fields and a shared colorbar is used. This is the correct choice for comparing snapshots or methods on a common physical scale. If each panel requires its own range, call plot_cloud separately on prepared axes.

Time series

plot_time_series(
    t,
    y_list,
    *,
    labels=None,
    ax=None,
    title=None,
    xlabel="t",
    ylabel="value",
    legend=True,
    grid=True,
    linewidth=1.8,
    fontsize=None,
    save=None,
    dpi=None,
    tight=None,
)

Returns (fig, ax). y_list is a sequence of one-dimensional series sharing the same time coordinate.

Mode shapes

plot_mode_shapes(
    modes,
    *,
    extent=None,
    mask=None,
    titles=None,
    normalize=True,
    layout=None,
    cmap=None,
    symmetric=True,
    center=0.0,
    robust=False,
    robust_q=(1.0, 99.0),
    cb_shared=True,
    cb_label="",
    save_prefix=None,
    dpi=None,
)

Each mode must be two-dimensional. With normalize=True, every mode is divided by its own maximum absolute finite value before the common panel scale is computed.

The return value is (fig, axes, colorbar, (vmin, vmax)). If save_prefix is provided, the current implementation saves "{save_prefix}_modes.png".

Geometry plots

plot_geometry(
    geo,
    ax=None,
    *,
    title=None,
    bounds=None,
    pad=0.05,
    sample_nx=300,
    sample_ny=300,
    cmap=None,
    facecolor="tab:blue",
    edgecolor="k",
    linewidth=1.5,
    alpha=0.35,
    show_axes=True,
)

show_geometry(geo, **kwargs)

plot_geometry uses native Matplotlib patches for recognized rectangles, disks, and polygons. Other geometries are sampled through their inside/contains method. It returns (fig, ax, info), where info records the plotting mode, bounds, primitive type, and sample shape.

show_geometry calls plt.show() and returns the same tuple. Prefer plot_geometry in scripts and test suites where blocking display behavior is undesirable.

Publication workflow

from optixde.post import PlotStyle, plot_cloud, set_plot_style

set_plot_style(PlotStyle(
    cmap="viridis",
    fontsize=10,
    figure_dpi=300,
))

fig, _, _, _, _ = plot_cloud(
    u,
    extent=(xmin, xmax, ymin, ymax),
    mask=outside_mask,
    title=r"$u(x,y)$",
    cb_label="amplitude",
    annotate_minmax=False,
)
fig.savefig("solution.pdf", bbox_inches="tight")

For a paper, set physical extents explicitly, retain equal axis scaling for spatial domains, use a common range for comparison panels, and close figures after saving them in long batch jobs.