#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
样本数据离线验证 / 渲染基准生成器。

用途：在未搭好 Qt/VTK 环境前，用 Python(matplotlib) 复现剖面散点(#17)、网格等值面(#18)
渲染效果，验证「样本解析 + colorBar 色阶映射 + 异常叠加」逻辑正确；并量化两条剖面的几何
关系，为 dd_voxel 可信度(设计 §10/§14 K-10)提供依据。

产出的 PNG 作为后续 VTK 渲染的「地面真值」对照图。

运行：
    python tools/validate_samples.py
输出：docs/_validate/ref_17_scatter.png, ref_18_grid.png（该目录已 .gitignore）

依赖：numpy, matplotlib
"""
import json
import os
import re
import sys

import numpy as np
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
from matplotlib.colors import ListedColormap, BoundaryNorm

DATA = os.path.join(os.path.dirname(__file__), "..", "docs", "剖面网格数据的色阶数据2等文件")
OUT = os.path.join(os.path.dirname(__file__), "..", "docs", "_validate")


def load(name):
    with open(os.path.join(DATA, name), encoding="utf-8") as f:
        return json.load(f)["data"]


def parse_color(c):
    """支持 #RRGGBB 与 rgba(r,g,b,a)；返回 0-1 RGB。"""
    c = c.strip()
    if c.startswith("#"):
        return (int(c[1:3], 16) / 255, int(c[3:5], 16) / 255, int(c[5:7], 16) / 255)
    nums = re.findall(r"[\d.]+", c)
    return (float(nums[0]) / 255, float(nums[1]) / 255, float(nums[2]) / 255)


def colorbar(cb):
    """colorBar: [[value, color], ...] -> (sorted values, colors)。"""
    pairs = sorted(((float(v), parse_color(c)) for v, c in cb), key=lambda p: p[0])
    return np.array([p[0] for p in pairs]), [p[1] for p in pairs]


def render_scatter():
    raw = load("剖面原数据1.txt")
    vals, cols = colorbar(load("剖面原数据的色阶数据1.txt")["properties"]["colorBar"])
    x, y, v = np.array(raw["xlist"]), np.array(raw["ylist"]), np.array(raw["vlist"])
    cmap = ListedColormap(cols)
    norm = BoundaryNorm(np.append(vals, vals[-1] * 1.2), cmap.N)
    plt.figure(figsize=(11, 3.2))
    plt.scatter(x, y, c=v, cmap=cmap, norm=norm, s=6, marker="s")
    plt.gca().invert_yaxis()
    plt.title("scatter (raw profile) ~#17")
    plt.colorbar(shrink=0.7)
    plt.tight_layout()
    plt.savefig(os.path.join(OUT, "ref_17_scatter.png"), dpi=90)
    plt.close()
    print(f"[#17] pts={len(x)} x[{x.min():.1f},{x.max():.1f}] "
          f"y[{y.min():.1f},{y.max():.1f}] v[{v.min():.1f},{v.max():.1f}]")


def render_grid():
    g = load("剖面网格数据1.txt")
    vals, cols = colorbar(load("剖面网格数据的色阶数据1.txt")["properties"]["colorBar"])
    gx, gy, gv = np.array(g["x"]), np.array(g["y"]), np.array(g["v"])  # gv: [22][100]
    X, Y = np.meshgrid(gx, gy)
    plt.figure(figsize=(11, 3.0))
    cf = plt.contourf(X, Y, gv, levels=vals, colors=cols[:len(vals) - 1])
    plt.contour(X, Y, gv, levels=vals, colors="k", linewidths=0.4)
    for a in load("剖面网格数据1——对应的异常圈定数据.txt"):
        coord = a.get("location", {}).get("coordinate", [])
        if coord and isinstance(coord[0], dict):
            plt.plot([p["x"] for p in coord], [p["y"] for p in coord], "k--", lw=1.5)
    plt.gca().invert_yaxis()
    plt.title("grid bands + contour + anomaly ~#18")
    plt.colorbar(cf, shrink=0.7)
    plt.tight_layout()
    plt.savefig(os.path.join(OUT, "ref_18_grid.png"), dpi=90)
    plt.close()
    print(f"[#18] grid V{gv.shape} vmin/max {g['vmin']:.1f}/{g['vmax']:.1f}")


def analyze_voxel_geometry():
    """量化两条剖面几何关系：判断 dd_voxel 三维插值的数据支撑充分性。"""
    def line(name):
        r = load(name)
        return np.column_stack([r["projectXList"], r["projectYList"]])

    def principal_dir(p):
        c = p - p.mean(0)
        _, _, vt = np.linalg.svd(c, full_matrices=False)
        return vt[0], p.mean(0)

    p1, p2 = line("剖面原数据1.txt"), line("剖面原数据2.txt")
    d1, c1 = principal_dir(p1)
    d2, c2 = principal_dir(p2)
    ang = np.degrees(np.arccos(abs(np.clip(d1 @ d2, -1, 1))))
    n1 = np.array([-d1[1], d1[0]])
    perp_gap = abs((c2 - c1) @ n1)
    print(f"[voxel] line1 pts={len(p1)} line2 pts={len(p2)} | "
          f"angle={ang:.1f}deg perp_gap={perp_gap:.1f}m "
          f"=> {'crossing' if ang > 30 else 'sub-parallel'}; "
          f"volume between lines is interpolated/extrapolated "
          f"(credible voxel needs >=3 lines or a 3D grid)")


def main():
    os.makedirs(OUT, exist_ok=True)
    render_scatter()
    render_grid()
    analyze_voxel_geometry()
    print("written:", sorted(os.listdir(OUT)))


if __name__ == "__main__":
    sys.stdout.reconfigure(encoding="utf-8")
    main()