"""Define functions related to optimization and failures plotting.
.. todo::
Information on the element under the cursor (hover).
"""
import logging
from collections.abc import Sequence
import matplotlib.patches as pat
import numpy as np
from matplotlib.axes import Axes
from matplotlib.lines import Line2D
from matplotlib.typing import ColorType
from lightwin.core.elements.element import Element
from lightwin.failures.fault import Fault
from lightwin.optimisation.objective.helper import by_element
from lightwin.optimisation.objective.objective import Objective
from lightwin.visualization.helper import X_AXIS_T, create_fig_if_not_exists
from lightwin.visualization.structure import patch_kwargs
OBJECTIVE_COLOR = "blue"
[docs]
def _get_objectives(fault_scenario: list[Fault] | None) -> list[Objective]:
"""Get the objectives stored in ``fault_scenario``."""
if fault_scenario is None or len(fault_scenario) == 0:
return []
return [
objective for fault in fault_scenario for objective in fault.objectives
]
[docs]
def mark_objectives_position(
ax: Axes,
fault_scenarios: Sequence[list[Fault]] | None,
y_axis: str = "struct",
x_axis: X_AXIS_T = "z_abs",
color: ColorType = OBJECTIVE_COLOR,
alpha: float = 0.5,
) -> None:
"""Show where objectives are evaluated."""
if fault_scenarios is None:
logging.info(
"The ``fault_scenarios`` must be given to plot.factory for the "
"objectives to be displayed."
)
return
objectives_by_element: dict[Element, list[Objective]]
objectives_by_element = by_element(_get_objectives(fault_scenarios[0]))
for elt in objectives_by_element:
kwargs = patch_kwargs(elt, x_axis, color=color, alpha=alpha, pos="out")
if y_axis != "struct":
_line_objective(ax, **kwargs)
continue
ax.add_patch(_patch_objective(**kwargs))
[docs]
def _line_objective(
ax: Axes, x_0: float, color: ColorType, alpha: float, **kwargs
) -> Line2D:
"""Give a vertical line to add to a plot.
This function is mainly to intercept the kwargs axvline would not
understand, such as x_0 or width.
"""
return ax.axvline(x=x_0, color=color, alpha=alpha)
[docs]
def _patch_objective(
x_0: float, width: float, color: ColorType, alpha: float, **kwargs
) -> pat.Arrow:
"""Add a marker at the exit of provided element."""
starting_height = 0.75
ending_height = 0.05
patch = pat.Arrow(
x=x_0,
y=starting_height,
dx=0,
dy=ending_height - starting_height,
width=2 * width,
color=color,
alpha=alpha,
)
return patch
[docs]
def plot_fit_progress(hist_f, l_label, nature="Relative"):
"""Plot the evolution of the objective functions w/ each iteration."""
_, axx = create_fig_if_not_exists(1, num=32)
axx = axx[0]
scales = {
"Relative": lambda x: x / x[0],
"Absolute": lambda x: x,
}
# Number of objectives, number of evaluations
n_f = len(l_label)
n_iter = len(hist_f)
iteration = np.linspace(0, n_iter - 1, n_iter)
__f = np.empty([n_f, n_iter])
for i in range(n_iter):
__f[:, i] = scales[nature](hist_f)[i]
for j, label in enumerate(l_label):
axx.plot(iteration, __f[j], label=label)
axx.grid(True)
axx.legend()
axx.set_xlabel("Iteration #")
axx.set_ylabel(f"{nature} variation of error")
axx.set_yscale("log")