"""
# top cuts

gold in four quartz veins, V1 to V4, sampled by diamond holes and underground channels. a handful of extreme assays
carry much of the metal, and one of them next to a block would lend it its grade. capping the grades at a top cut
limits their influence. you need to decide where to cut and how much metal the cut costs.
"""

# %% [hidden]
import sys
from pathlib import Path

HERE = Path(__file__).parent if "__file__" in globals() else Path.cwd()
sys.path.insert(0, str(HERE.parents[1]))

# %%
import boitata as bt
import matplotlib.pyplot as plt
import numpy as np
from common import ACCENT, save

data = bt.datasets.vein_gold_grade_control()
intervals = bt.merge_intervals(data["assays"], data["lithology"])
holes = bt.Drillholes(data["collars"], data["surveys"], intervals)
composites = holes.composite(1.0, ["AU_GPT"], domain="LITH", categories=["VEIN"])
quartz = composites.filter(composites["LITH"] == "QV").drop_null("AU_GPT")
au = quartz["AU_GPT"]
print(f"{len(quartz)} composites of 1 m in quartz vein, mean Au {au.mean():.2f} g/t, max {au.max():.0f} g/t")

# %% [markdown]
# channels crowd the developed levels, so the composites are first declustered in 20 m cells
# ([declustering](../../03-exploratory-analysis/03-declustering/README.md)).

# %%
weights = bt.cell_declustering(quartz, "AU_GPT", cell_size=20.0).weights
print(f"declustered mean Au {np.average(au, weights=weights):.2f} g/t")

# %% [markdown]
# ## how much metal sits in the tail
#
# `capping` tries caps at high quantiles and reports, for each, the share of weight above it, the metal removed, and the
# capped mean and CV.

# %%
caps = bt.capping("AU_GPT", weights=weights, data=quartz)
print(f"{'cap':>7}{'above (%)':>11}{'metal (%)':>11}{'mean':>7}{'CV':>6}")
for cap, above, metal, mean, cv in zip(*(caps[c] for c in caps.column_names), strict=True):
    print(f"{cap:>7.1f}{100 * above:>11.2f}{100 * metal:>11.1f}{mean:>7.2f}{cv:>6.2f}")

# %% [markdown]
# the top 1 % of the weight holds 11.5 % of the metal, and the top 10 % holds 36.5 %. the CV climbs from 1.07 at the
# lowest cap to 2.34 at the highest: the tail makes gold grades erratic.
#
# the log-probability plot shows where the tail breaks away from the body of the distribution. the dotted tukey fences
# sit 1.5 interquartile ranges beyond the quartiles of log Au, and the dashed line is a cap at the declustered P99 of
# the vein.

# %%
stats = bt.describe_by("AU_GPT", "VEIN", weights=weights, quantiles=[0.5, 0.99], data=quartz)
top = dict(zip(stats["category"][:-1], stats["P99"][:-1], strict=True))
vein = quartz["VEIN"]
fig, axes = plt.subplots(1, 2, figsize=(9, 3.8), layout="constrained", sharey=True)
for ax, name in zip(axes, ["V1", "V2"], strict=True):
    keep = vein == name
    bt.plot.probability(
        au[keep], weights=weights[keep], log=True, cap=top[name], fences=1.5, ax=ax, color=ACCENT, ms=2
    )
    ax.set(title=f"{name}, declustered", xlabel="Au (g/t)")
    ax.legend(loc="lower right")
axes[1].set_ylabel("")
save(fig, "probability")

# %% [markdown]
# both veins plot near a straight line (a lognormal body) up to about P99. above it the points thin out, with the 1192
# g/t channel alone at the top of V2. the upper fences, near 110 g/t, agree with a cap around P99.

# %% [markdown]
# ## one cap per vein
#
# `capping_report` applies one cap per domain and compares the declustered statistics before and after, with the metal
# removed; the last row pools the veins.

# %%
report = bt.capping_report("AU_GPT", top, domain_column="VEIN", weights=weights, data=quartz)
columns = ["domain", "cap", "n", "n_capped", "mean", "mean_capped", "cv", "cv_capped"]
print(
    f"{'vein':<5}{'cap':>7}{'n':>6}{'cut':>5}{'mean':>7}{'capped':>8}{'CV':>6}{'capped':>8}{'metal (%)':>11}"
)
for name, c, n, cut, mean, capped, cv, cv_capped in zip(*(report[k] for k in columns), strict=True):
    shown = "" if np.isnan(c) else f"{c:.1f}"
    print(
        f"{name:<5}{shown:>7}{n:>6.0f}{cut:>5.0f}{mean:>7.2f}{capped:>8.2f}{cv:>6.2f}{cv_capped:>8.2f}"
        f"{100 * (1 - capped / mean):>11.1f}"
    )

# %% [markdown]
# the caps cut 47 composites in V1 and 23 in V2, and remove 10.8 % and 15.4 % of their metal. pooled, the veins lose
# 11.1 % of the metal and the CV falls from 3.22 to 1.78. V3 and V4 hold about a hundred composites each, too few for a
# P99 to mean much: in V3 it is the maximum and cuts nothing. cap a small domain with the cap of a similar, larger one.