src/iohmm_evac/report/plots.py

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
# SPDX-License-Identifier: AGPL-3.0-only
# Copyright (C) 2026 SWGY, Inc.
"""Diagnostic plots for a :class:`SimulationBundle`.

Every plot is a pure function: it accepts a bundle and an optional
:class:`matplotlib.axes.Axes` (or sequence of axes) and returns the axes it
drew on. Functions never call :func:`matplotlib.pyplot.show` or
:meth:`matplotlib.figure.Figure.savefig` — that is the caller's job.
"""

from __future__ import annotations

from collections.abc import Sequence
from typing import TYPE_CHECKING, cast

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd

from iohmm_evac.report.constants import (
    SCENARIO_COLORS,
    SCENARIO_ORDER,
    STATE_COLORS,
    STATE_ORDER,
)
from iohmm_evac.report.loader import SimulationBundle, load_bundle

if TYPE_CHECKING:
    from matplotlib.axes import Axes

    from iohmm_evac.bootstrap.aggregate import BandResult
    from iohmm_evac.sweep import SweepResult

__all__ = [
    "plot_bootstrap_bands",
    "plot_cumulative_departures",
    "plot_emission_summary",
    "plot_household_trajectories",
    "plot_state_occupancy",
    "plot_sweep_departures",
    "plot_sweep_network",
]


_BAND_METRIC_TITLES: dict[str, str] = {
    "failed_evacuation_count": "Failed evacuations (count)",
    "peak_enroute_share": "Peak EnRoute share",
    "total_delay_hours": "Total delay (hours)",
    "shelter_overflow_count": "Shelter overflow (count)",
}


_MAX_TRAJECTORY_HOUSEHOLDS = 6


def _state_share_panel(observations: pd.DataFrame) -> pd.DataFrame:
    """Pivot observations into a (t by state) share table aligned to STATE_ORDER."""
    counts = observations.groupby(["t", "state"]).size().unstack(fill_value=0).sort_index()
    for s in STATE_ORDER:
        if s not in counts.columns:
            counts[s] = 0
    counts = counts[list(STATE_ORDER)]
    totals = counts.sum(axis=1).replace(0, 1)
    return counts.div(totals, axis=0)


def _add_timeline_overlays(ax: Axes, bundle: SimulationBundle) -> None:
    """Draw vertical lines at voluntary, mandatory, and landfall hours."""
    timeline = bundle.timeline
    voluntary_mask = timeline["voluntary"].astype(bool).to_numpy()
    mandatory_mask = timeline["mandatory"].astype(bool).to_numpy()

    if voluntary_mask.any():
        vol_hour = int(timeline.loc[voluntary_mask, "t"].iloc[0])
        ax.axvline(vol_hour, ls="--", color="black", alpha=0.5, label="voluntary")
    if mandatory_mask.any():
        mand_hour = int(timeline.loc[mandatory_mask, "t"].iloc[0])
        ax.axvline(mand_hour, ls="--", color="firebrick", alpha=0.6, label="mandatory")
    ax.axvline(bundle.t_landfall, ls="-", color="black", alpha=0.7, label="landfall")


def plot_state_occupancy(bundle: SimulationBundle, ax: Axes | None = None) -> Axes:
    """Stacked area chart of state shares over time (Fig. 3).

    Overlays vertical dashed lines at the voluntary and mandatory order hours
    (read from the timeline DataFrame) and a solid line at landfall.
    """
    if ax is None:
        _, ax = plt.subplots(figsize=(10, 4))
    shares = _state_share_panel(bundle.observations)
    colors = [STATE_COLORS[s] for s in STATE_ORDER]
    ax.stackplot(
        shares.index.to_numpy(),
        shares.to_numpy().T,
        labels=list(STATE_ORDER),
        colors=colors,
        alpha=0.9,
    )
    _add_timeline_overlays(ax, bundle)
    ax.set_xlim(0, bundle.t_landfall)
    ax.set_ylim(0, 1)
    ax.set_xlabel("Hours from start")
    ax.set_ylabel("Population share")
    ax.set_title("State occupancy over time")
    ax.legend(loc="upper left", ncol=4, fontsize=8, framealpha=0.9)
    return ax


def plot_cumulative_departures(bundle: SimulationBundle, ax: Axes | None = None) -> Axes:
    """Cumulative share of households that have ever departed, vs time (Fig. 4).

    A household's "departure hour" is the first ``t`` where its latent
    state is ER. Households that never enter ER never depart. We plot the
    cumulative share whose departure hour is ``<= t``, against ``t``.

    The ``departure`` emission column is *not* used here: it carries
    Bernoulli noise (~3% per hour even from non-evacuating households),
    which is appropriate for IO-HMM fitting in Build 2 but produces a
    visually misleading curve for sanity-checking the underlying
    behavioral dynamics.
    """
    if ax is None:
        _, ax = plt.subplots(figsize=(10, 4))
    obs = bundle.observations
    in_er = obs[obs["state"] == "ER"]
    n_total = bundle.n_households
    timeline_t = bundle.timeline["t"].to_numpy()
    if in_er.empty:
        cum = np.zeros_like(timeline_t, dtype=float)
    else:
        first_er = in_er.groupby("household_id")["t"].min()
        counts = first_er.value_counts().sort_index()
        per_hour = pd.Series(0, index=timeline_t, dtype=np.int64)
        per_hour.loc[counts.index] = counts.to_numpy(dtype=np.int64)
        cum = per_hour.cumsum().to_numpy(dtype=float) / float(max(n_total, 1))
    ax.plot(timeline_t, cum, color="#1f77b4", lw=2.0, label="cumulative")
    _add_timeline_overlays(ax, bundle)
    ax.set_xlim(0, bundle.t_landfall)
    ax.set_ylim(0, 1)
    ax.set_xlabel("Hours from start")
    ax.set_ylabel("Cumulative departure share")
    ax.set_title("Cumulative departures")
    ax.legend(loc="upper left", fontsize=8, framealpha=0.9)
    return ax


def _state_codes(states: pd.Series) -> np.ndarray:
    """Map state-label strings to integer codes following STATE_ORDER."""
    lookup = {s: i for i, s in enumerate(STATE_ORDER)}
    out: np.ndarray = states.map(lookup).to_numpy(dtype=np.int64)
    return out


def _validate_household_ids(bundle: SimulationBundle, household_ids: Sequence[int]) -> list[int]:
    """Verify that the requested IDs exist and are not too many to plot."""
    ids = list(household_ids)
    if not ids:
        msg = "household_ids must contain at least one id"
        raise ValueError(msg)
    if len(ids) > _MAX_TRAJECTORY_HOUSEHOLDS:
        msg = (
            f"plot_household_trajectories supports at most "
            f"{_MAX_TRAJECTORY_HOUSEHOLDS} households (got {len(ids)})"
        )
        raise ValueError(msg)
    available = set(bundle.population["household_id"].astype(int).tolist())
    missing = [i for i in ids if int(i) not in available]
    if missing:
        msg = f"Households not found in bundle: {missing}"
        raise ValueError(msg)
    return [int(i) for i in ids]


def _draw_trajectory(ax: Axes, bundle: SimulationBundle, hh_id: int) -> None:
    """Render a single household's trajectory panel."""
    sub = bundle.observations[bundle.observations["household_id"] == hh_id].sort_values("t")
    t_arr = sub["t"].to_numpy()
    state_codes = _state_codes(sub["state"])
    ax.plot(
        bundle.timeline["t"].to_numpy(),
        bundle.timeline["forecast"].to_numpy(),
        color="#777777",
        lw=1.0,
        label="forecast",
    )
    ax.step(t_arr, state_codes, where="post", color="black", lw=1.5, label="state")
    # The X mark is the household's actual departure hour: the first ``t``
    # at which the latent state is ER. The noisy ``departure`` emission
    # column is intentionally ignored here (see plot_cumulative_departures
    # for rationale).
    er_hours = sub.loc[sub["state"] == "ER", "t"].to_numpy()
    if er_hours.size:
        first_er = int(er_hours.min())
        ax.scatter(
            [first_er],
            [len(STATE_ORDER) - 0.5],
            marker="x",
            color="#d73027",
            s=40,
            label="departure",
        )
    disp = sub["displacement"].to_numpy()
    if np.nanmax(disp) > 0:
        scaled = disp / max(float(np.nanmax(disp)), 1e-9) * (len(STATE_ORDER) - 1)
        ax.plot(t_arr, scaled, color="#1a9850", lw=1.0, alpha=0.6, label="displacement")
    timeline = bundle.timeline
    if timeline["voluntary"].astype(bool).any():
        vol_t = int(timeline.loc[timeline["voluntary"].astype(bool), "t"].iloc[0])
        ax.axvline(vol_t, ls="--", color="black", alpha=0.4)
    if timeline["mandatory"].astype(bool).any():
        mand_t = int(timeline.loc[timeline["mandatory"].astype(bool), "t"].iloc[0])
        ax.axvline(mand_t, ls="--", color="firebrick", alpha=0.5)
    ax.set_yticks(range(len(STATE_ORDER)))
    ax.set_yticklabels(list(STATE_ORDER))
    ax.set_xlim(0, bundle.t_landfall)
    ax.set_title(f"Household {hh_id}")
    ax.legend(loc="upper left", fontsize=7, framealpha=0.85)


def plot_household_trajectories(
    bundle: SimulationBundle,
    household_ids: Sequence[int],
    ax: Sequence[Axes] | None = None,
) -> Sequence[Axes]:
    """Multi-panel forecast / state / departure / displacement plot (Fig. 2).

    If ``ax`` is supplied, it must be a sequence of axes whose length matches
    ``household_ids``. Otherwise a new figure with one subplot per household
    is created.
    """
    ids = _validate_household_ids(bundle, household_ids)
    if ax is None:
        _, axes_obj = plt.subplots(len(ids), 1, figsize=(10, 2.5 * len(ids)), sharex=True)
        axes: list[Axes] = (
            [cast("Axes", axes_obj)] if len(ids) == 1 else list(np.atleast_1d(axes_obj))
        )
    else:
        axes = list(ax)
        if len(axes) != len(ids):
            msg = (
                f"ax must have one entry per household_id (got {len(axes)} axes for {len(ids)} ids)"
            )
            raise ValueError(msg)
    for axis, hh_id in zip(axes, ids, strict=True):
        _draw_trajectory(axis, bundle, hh_id)
    axes[-1].set_xlabel("Hours from start")
    return axes


def _emission_summary_table(observations: pd.DataFrame) -> pd.DataFrame:
    """Per-state means of departure, displacement, comm_count."""
    agg = observations.groupby("state")[["departure", "displacement", "comm_count"]].mean()
    rows = [s for s in STATE_ORDER if s in agg.index]
    return agg.loc[rows]


def plot_emission_summary(bundle: SimulationBundle, ax: Axes | None = None) -> Axes:
    """Draw a grouped bar chart of per-state emission means (sanity check)."""
    if ax is None:
        _, ax = plt.subplots(figsize=(8, 4))
    summary = _emission_summary_table(bundle.observations)
    metrics = ["departure", "displacement", "comm_count"]
    n_states = summary.shape[0]
    width = 0.25
    x = np.arange(n_states, dtype=float)
    palette = ["#4575b4", "#fdae61", "#762a83"]
    for i, metric in enumerate(metrics):
        ax.bar(
            x + (i - 1) * width,
            summary[metric].to_numpy(),
            width=width,
            label=metric,
            color=palette[i],
        )
    ax.set_xticks(x)
    ax.set_xticklabels(list(summary.index))
    ax.set_ylabel("Mean per (household, t) row")
    ax.set_title("Emission summary by state")
    ax.legend(loc="upper left", fontsize=8, framealpha=0.9)
    return ax


def _scenario_warning_hours(bundle: SimulationBundle) -> tuple[int | None, int | None]:
    """Return the first (voluntary, mandatory) warning hours from a bundle's timeline."""
    timeline = bundle.timeline
    vol_mask = timeline["voluntary"].astype(bool).to_numpy()
    mand_mask = timeline["mandatory"].astype(bool).to_numpy()
    vol = int(timeline.loc[vol_mask, "t"].iloc[0]) if vol_mask.any() else None
    mand = int(timeline.loc[mand_mask, "t"].iloc[0]) if mand_mask.any() else None
    return vol, mand


def _ordered_sweep_scenarios(sweep: SweepResult) -> list[str]:
    """Order the sweep's scenarios by SCENARIO_ORDER, appending unknowns at the end."""
    present = list(sweep.config.scenarios)
    ordered = [s for s in SCENARIO_ORDER if s in present]
    extra = [s for s in present if s not in SCENARIO_ORDER]
    return ordered + extra


def _cumulative_share(bundle: SimulationBundle) -> tuple[np.ndarray, np.ndarray]:
    """Return (timeline_t, cumulative-departure-share) for a single bundle."""
    obs = bundle.observations
    in_er = obs[obs["state"] == "ER"]
    n_total = bundle.n_households
    timeline_t = bundle.timeline["t"].to_numpy()
    if in_er.empty:
        return timeline_t, np.zeros_like(timeline_t, dtype=float)
    first_er = in_er.groupby("household_id")["t"].min()
    counts = first_er.value_counts().sort_index()
    per_hour = pd.Series(0, index=timeline_t, dtype=np.int64)
    per_hour.loc[counts.index] = counts.to_numpy(dtype=np.int64)
    cum = per_hour.cumsum().to_numpy(dtype=float) / float(max(n_total, 1))
    return timeline_t, cum


def plot_sweep_departures(sweep: SweepResult, ax: Axes | None = None) -> Axes:
    """Overlay cumulative-departure curves across scenarios (Fig. 4).

    Each scenario gets its own line in the colorblind-friendly
    :data:`SCENARIO_COLORS` palette; the legend label includes the scenario's
    voluntary / mandatory warning hours. The landfall hour is drawn as a
    single solid vertical reference; per-scenario warning verticals are
    deliberately omitted because they differ across scenarios.
    """
    if ax is None:
        _, ax = plt.subplots(figsize=(10, 4))
    ordered = _ordered_sweep_scenarios(sweep)
    landfall: int | None = None
    for scenario in ordered:
        bundle = load_bundle(sweep.bundles[scenario])
        timeline_t, cum = _cumulative_share(bundle)
        vol, mand = _scenario_warning_hours(bundle)
        landfall = bundle.t_landfall if landfall is None else landfall
        vol_txt = f"{vol}" if vol is not None else "—"
        mand_txt = f"{mand}" if mand is not None else "—"
        label = f"{scenario} (vol={vol_txt}, mand={mand_txt})"
        color = SCENARIO_COLORS.get(scenario, None)
        ax.plot(timeline_t, cum, color=color, lw=2.0, label=label)
    if landfall is not None:
        ax.axvline(landfall, ls="-", color="black", alpha=0.7, label="landfall")
        ax.set_xlim(0, landfall)
    ax.set_ylim(0, 1)
    ax.set_xlabel("Hours from start")
    ax.set_ylabel("Cumulative departure share")
    ax.set_title("Cumulative departures by scenario")
    ax.legend(loc="upper left", fontsize=8, framealpha=0.9)
    return ax


def _draw_metric_panel(
    ax: Axes,
    scenarios: list[str],
    values: list[float],
    title: str,
    labels: list[str],
) -> None:
    y_pos = np.arange(len(scenarios), dtype=float)
    colors = [SCENARIO_COLORS.get(s, "#777777") for s in scenarios]
    ax.barh(y_pos, values, color=colors, edgecolor="black", linewidth=0.5)
    ax.set_yticks(y_pos)
    ax.set_yticklabels(scenarios)
    ax.invert_yaxis()
    ax.set_title(title)
    span = max(values) if values else 0.0
    pad = max(span * 0.02, 1e-3)
    for y, v, label in zip(y_pos, values, labels, strict=True):
        ax.text(v + pad, y, label, va="center", fontsize=8)
    if span > 0:
        ax.set_xlim(0, span * 1.18)


def plot_sweep_network(
    sweep: SweepResult,
    ax: Sequence[Sequence[Axes]] | np.ndarray | None = None,
) -> np.ndarray:
    """2x2 panel of per-scenario network metrics (Fig. 5).

    Panels (top-left → bottom-right): total delay, peak EnRoute share,
    shelter overflow, failed evacuations. The peak-share panel labels each
    bar with both the share and the hour at which it is attained.

    If ``ax`` is None, a new figure with a 2x2 layout is created. Otherwise
    ``ax`` must be a 2x2 sequence/ndarray of axes.
    """
    if ax is None:
        _, axes_obj = plt.subplots(2, 2, figsize=(10, 6))
        axes_arr = np.asarray(axes_obj, dtype=object)
    else:
        axes_arr = np.asarray(ax, dtype=object)
    if axes_arr.shape != (2, 2):
        msg = f"ax must be a 2x2 array of axes (got shape {axes_arr.shape})"
        raise ValueError(msg)

    scenarios = _ordered_sweep_scenarios(sweep)
    metrics = sweep.network_metrics

    delay_values = [float(metrics[s].total_delay_hours) for s in scenarios]
    delay_labels = [f"{v:.1f}" for v in delay_values]
    _draw_metric_panel(
        cast("Axes", axes_arr[0, 0]),
        scenarios,
        delay_values,
        "Total delay (hours)",
        delay_labels,
    )

    peak_values = [float(metrics[s].peak_enroute_share) for s in scenarios]
    peak_labels = [
        f"{metrics[s].peak_enroute_share:.3f} @ t={metrics[s].peak_enroute_hour}" for s in scenarios
    ]
    _draw_metric_panel(
        cast("Axes", axes_arr[0, 1]),
        scenarios,
        peak_values,
        "Peak EnRoute share",
        peak_labels,
    )

    overflow_values = [float(metrics[s].shelter_overflow_count) for s in scenarios]
    overflow_labels = [f"{int(v)}" for v in overflow_values]
    _draw_metric_panel(
        cast("Axes", axes_arr[1, 0]),
        scenarios,
        overflow_values,
        "Shelter overflow (count)",
        overflow_labels,
    )

    failed_values = [float(metrics[s].failed_evacuation_count) for s in scenarios]
    failed_labels = [f"{int(v)}" for v in failed_values]
    _draw_metric_panel(
        cast("Axes", axes_arr[1, 1]),
        scenarios,
        failed_values,
        "Failed evacuations (count)",
        failed_labels,
    )

    return axes_arr


def plot_bootstrap_bands(
    band_result: BandResult,
    metric: str = "failed_evacuation_count",
    ax: Axes | None = None,
) -> Axes:
    """Render Fig. 6: median + 25–75% and 5–95% quantile bands across shifts.

    The solid line is the 50th-percentile, the inner shaded band is the
    25–75% range, and the outer shaded band is the 5–95% range. A vertical
    dashed line at ``δ = 0`` marks baseline timing.
    """
    if metric not in band_result.bands:
        msg = f"Metric {metric!r} not in BandResult. Known: {tuple(band_result.bands.keys())}"
        raise KeyError(msg)
    required = {5, 25, 50, 75, 95}
    missing = required.difference(band_result.percentiles)
    if missing:
        msg = (
            f"BandResult is missing percentiles {sorted(missing)} required for plot_bootstrap_bands"
        )
        raise ValueError(msg)
    if ax is None:
        _, ax = plt.subplots(figsize=(8, 4.5))
    shifts = np.asarray(band_result.shifts, dtype=np.float64)
    p5 = band_result.quantile(metric, 5)
    p25 = band_result.quantile(metric, 25)
    p50 = band_result.quantile(metric, 50)
    p75 = band_result.quantile(metric, 75)
    p95 = band_result.quantile(metric, 95)
    ax.fill_between(shifts, p5, p95, color="#4575b4", alpha=0.20, label="5–95%")
    ax.fill_between(shifts, p25, p75, color="#4575b4", alpha=0.40, label="25–75%")
    ax.plot(shifts, p50, color="#1a3a73", lw=2.0, label="median")
    ax.axvline(0.0, ls="--", color="black", alpha=0.6, label="baseline timing")
    ax.set_xlabel("Warning lead-time shift δ (hours)")
    ax.set_ylabel(_BAND_METRIC_TITLES.get(metric, metric))
    ax.set_title(f"Bootstrap bands: {_BAND_METRIC_TITLES.get(metric, metric)}")
    ax.legend(loc="best", fontsize=8, framealpha=0.9)
    return ax