import osimport matplotlib.pyplot as pltimport numpy as npimport pandas as pdfrom scipy.stats import gaussian_kdefrom matplotlib.patches import Patch# 1. 自动创建图表文件夹output_dir = "图表"if not os.path.exists(output_dir):    os.makedirs(output_dir)# 2. 从 Excel 中读取数据if not os.path.exists("data.xlsx"):    raise FileNotFoundError("未找到 data.xlsx 文件，请先运行数据生成脚本。")df_shot = pd.read_excel("data.xlsx", sheet_name="Shot_Latency")df_subshot = pd.read_excel("data.xlsx", sheet_name="Subshot_Latency")# 获取所有的循环周期cycles = sorted(df_shot["Cycles"].unique())# 3. 符合主流期刊标准的专业科研配色方案COLOR_PALETTES = {    1:  {"shot": "#1E77F4", "subshot": "#F2A900"},  # Nature 风格标准科技蓝与高对比橙黄}# 选择方案 1 作为绘图颜色selected_palette = COLOR_PALETTES[1]color_shot = selected_palette["shot"]color_subshot = selected_palette["subshot"]# 4. 初始化画布fig, ax = plt.subplots(figsize=(10, 4.5), dpi=300)# 5. 核心绘图逻辑for x_idx, cycle in enumerate(cycles):    shot_vals = df_shot[df_shot["Cycles"] == cycle]["Latency"].values    subshot_vals = df_subshot[df_subshot["Cycles"] == cycle]["Latency"].values    x_pos = x_idx    # --- 左侧：Shot Latency (半边小提琴) ---    if len(shot_vals) > 0:        shot_pts = shot_vals[:10]        log_shot_pts = np.log10(shot_pts)        kde = gaussian_kde(log_shot_pts)        y_log_range = np.linspace(np.log10(20), np.log10(350), 300)        kde_vals = kde(y_log_range)        kde_vals = (kde_vals / kde_vals.max()) * 0.24        y_range = 10 ** y_log_range        mask = (y_range >= shot_pts.min() * 0.85) & (y_range <= shot_pts.max() * 1.15)        ax.fill_betweenx(y_range[mask], x_pos - kde_vals[mask], x_pos, color=color_shot, alpha=0.2, lw=0)        ax.plot([x_pos, x_pos], [shot_pts.min(), shot_pts.max()], color=color_shot, lw=1.2, zorder=2)        median_shot = np.median(shot_pts)        ax.plot([x_pos - 0.12, x_pos + 0.12], [median_shot, median_shot], color=color_shot, lw=2.5, zorder=4)        ax.scatter([x_pos] * len(shot_pts), shot_pts, color=color_shot, s=18, alpha=0.9, zorder=3, edgecolors='none')    # --- 右侧：Subshot Latency (高密度精细直方图) ---    if len(subshot_vals) > 0:        log_subshot = np.log10(subshot_vals)        counts, log_bins = np.histogram(log_subshot, bins=70, range=(np.log10(25), np.log10(250)))        bins = 10 ** log_bins        counts = (counts / counts.max()) * 0.42        for j in range(len(counts)):            if counts[j] > 0:                ax.fill_betweenx([bins[j], bins[j + 1]], x_pos, x_pos + counts[j],                                 color=color_subshot, alpha=0.9, edgecolor=None, lw=0)                ax.plot([x_pos + counts[j], x_pos + counts[j]], [bins[j], bins[j + 1]], color=color_subshot, lw=0.6)                if j < len(counts) - 1 and counts[j + 1] > 0:                    ax.plot([x_pos + counts[j], x_pos + counts[j + 1]], [bins[j + 1], bins[j + 1]], color=color_subshot, lw=0.6)# 6. 坐标轴与图表美化ax.set_yscale('log')ax.set_ylim(20, 300)ax.set_yticks([30, 100, 300])ax.set_yticklabels(['30', '100', '300'])ax.get_yaxis().set_major_formatter(plt.ScalarFormatter())ax.set_ylabel(r'Latency ($\mu$s)', fontsize=13)target_indices = [idx for idx, c in enumerate(cycles) if np.log10(c).is_integer()]target_labels = [f"$10^{int(np.log10(cycles[idx]))}$" for idx in target_indices]ax.set_xticks(target_indices)ax.set_xticklabels(target_labels, fontsize=12)ax.set_xlabel('Number of cycles', fontsize=13)ax.spines['top'].set_visible(False)ax.spines['right'].set_visible(False)ax.spines['left'].set_linewidth(0.8)ax.spines['bottom'].set_linewidth(0.8)ax.tick_params(direction='out', length=6, width=0.8)legend_elements = [    Patch(facecolor=color_shot, label='Shot latency'),    Patch(facecolor=color_subshot, label='Subshot latency')]ax.legend(handles=legend_elements, loc='upper right', frameon=False, ncol=2, fontsize=11, handlelength=1, handleheight=1)ax.text(-0.06, 1.02, 'c', transform=ax.transAxes, fontsize=16, fontweight='bold', va='top', ha='right')plt.tight_layout()# 7. 保存高清图表output_path = os.path.join(output_dir, "latency_chart.png")plt.savefig(output_path, bbox_inches='tight')plt.close()print(f"绘制完成，应用配色方案: {output_path}")