diff --git a/batch_ssh/batch_ssh.py b/batch_ssh/batch_ssh.py
index 927ddbf..6e29134 100755
--- a/batch_ssh/batch_ssh.py
+++ b/batch_ssh/batch_ssh.py
@@ -54,7 +54,7 @@ def load_config(path):
 
 def ssh(host: Host, command: str, timeout: int) -> Result:
     ssh_cmd = [
-        "ssh", 
+        "ssh",
         "-p", host.port,
         "-o", f"ConnectTimeout={timeout}",
         "-o", "StrictHostKeyChecking=no",
@@ -113,19 +113,20 @@ def main():
             max_prefix_len = max(max_prefix_len, len(prefix))
             entry.append(prefix)
 
-        for i, entry in enumerate(res):
-            host, future, prefix = entry
-
+        for host, future, prefix in res:
             future_res = future.result()
             output = future_res.output.strip()
             if future_res.returncode is None:
                 assert("\n" not in output)
                 output = f"exception: {output}"
+                suffix = "---"
+            else:
+                suffix = future_res.returncode
             
             output_lines = output.splitlines(keepends=True)
             # 处理缩进
-            output = output_lines[0] + "".join(textwrap.indent("".join(output_lines[1:]), " " * (max_prefix_len + 13)))
-            print(f"[{prefix:<{max_prefix_len}}][ret: {future_res.returncode:>3}] {output}")
+            output = output_lines[0] + "".join(textwrap.indent("".join(output_lines[1:]), " " * (max_prefix_len + 8)))
+            print(f"[{prefix:<{max_prefix_len}}][{suffix:>3}] {output}")
             
 if __name__ == "__main__":
     main()
diff --git a/gantt/big_input.png b/gantt/big_input.png
new file mode 100644
index 0000000..594123f
Binary files /dev/null and b/gantt/big_input.png differ
diff --git a/gantt/big_input.txt b/gantt/big_input.txt
new file mode 100644
index 0000000..1b2f457
--- /dev/null
+++ b/gantt/big_input.txt
@@ -0,0 +1,324 @@
+timestamp: 2026-09-08 19:38:07.431900
+
+gantt results:
+#, begin, duration, task, size
+io-thread-00, 0.000, 0.000, sort, medium
+io-thread-00, 0.000, 0.200, write, tiny
+io-thread-00, 0.700, 0.005, dispatch, large
+io-thread-00, 0.405, 0.000, sync, small
+io-thread-00, 0.405, 8.000, poll, tiny
+io-thread-00, 8.905, 0.200, merge, xlarge
+io-thread-00, 9.205, 0.200, link, xlarge
+io-thread-00, 9.405, 0.000, fsync, large
+io-thread-00, 9.405, 0.500, write, small
+io-thread-00, 9.905, 0.005, sync, large
+worker-01, 0.000, 1.000, alloc, xlarge
+worker-01, 1.000, 0.000, sort, large
+worker-01, 1.500, 0.005, free, tiny
+worker-01, 2.005, 0.500, flush, xlarge
+worker-01, 2.505, 0.200, merge, tiny
+worker-01, 2.405, 0.200, send, tiny
+worker-01, 2.605, 0.005, free, medium
+worker-01, 2.710, 1.000, fsync, medium
+worker-01, 3.710, 0.200, checkpoint, medium
+worker-01, 3.610, 0.050, serialize, small
+worker-02, 0.000, 5.000, free, medium
+worker-02, 5.500, 0.200, checkpoint, medium
+worker-02, 5.400, 0.200, compute, medium
+worker-02, 5.700, 0.500, sync, small
+worker-02, 6.700, 1.000, kernel-A, large
+worker-02, 7.800, 5.000, write, medium
+worker-02, 12.800, 0.200, sort, xlarge
+worker-02, 13.500, 0.500, sort, xlarge
+worker-02, 14.100, 2.500, alloc, small
+worker-02, 16.600, 8.000, optimize, tiny
+io-thread-03, 0.000, 0.005, write, small
+io-thread-03, 0.105, 0.005, free, large
+io-thread-03, 0.610, 5.000, parse, medium
+io-thread-03, 6.110, 0.000, checkpoint, tiny
+io-thread-03, 6.610, 0.500, flush, medium
+io-thread-03, 6.810, 0.500, compile, small
+io-thread-03, 7.410, 0.000, sort, medium
+io-thread-03, 7.910, 0.050, parse, tiny
+io-thread-03, 7.960, 8.000, poll, small
+io-thread-03, 15.960, 1.000, fsync, xlarge
+worker-04, 0.500, 0.000, poll, medium
+worker-04, 0.600, 0.000, read, medium
+worker-04, 0.600, 0.200, compute, small
+worker-04, 1.300, 0.005, sync, large
+worker-04, 1.005, 8.000, write, small
+worker-04, 9.105, 8.000, fsync, medium
+worker-04, 17.605, 2.500, kernel-A, xlarge
+worker-04, 20.105, 0.500, free, medium
+worker-04, 20.705, 8.000, link, tiny
+worker-04, 28.705, 0.200, sync, medium
+worker-05, 0.000, 8.000, kernel-B, xlarge
+worker-05, 8.000, 0.000, sync, tiny
+worker-05, 8.000, 0.005, compute, medium
+worker-05, 7.705, 8.000, kernel-B, medium
+worker-05, 15.805, 0.200, serialize, small
+worker-05, 16.505, 5.000, kernel-B, large
+worker-05, 21.605, 0.200, read, tiny
+worker-05, 21.905, 1.000, compile, large
+worker-05, 23.005, 0.000, checkpoint, tiny
+worker-05, 22.705, 2.500, sort, medium
+io-thread-06, 0.000, 0.200, kernel-A, small
+io-thread-06, 0.700, 5.000, write, large
+io-thread-06, 5.700, 0.500, link, small
+io-thread-06, 5.900, 5.000, serialize, tiny
+io-thread-06, 10.600, 8.000, init, tiny
+io-thread-06, 18.600, 0.050, compile, large
+io-thread-06, 18.750, 0.200, free, tiny
+io-thread-06, 18.950, 2.500, init, large
+io-thread-06, 21.450, 5.000, send, large
+io-thread-06, 26.950, 5.000, write, small
+worker-07, 0.000, 0.200, compute, xlarge
+worker-07, 0.700, 0.000, sort, medium
+worker-07, 0.400, 0.000, dispatch, large
+worker-07, 0.900, 8.000, fsync, tiny
+worker-07, 9.400, 0.005, fsync, tiny
+worker-07, 9.905, 0.005, checkpoint, small
+worker-07, 10.010, 0.005, dispatch, small
+worker-07, 10.515, 0.000, poll, tiny
+worker-07, 10.615, 8.000, ping, medium
+worker-07, 18.615, 1.000, kernel-B, medium
+worker-08, 0.100, 0.050, checkpoint, medium
+worker-08, 0.250, 1.000, sync, tiny
+worker-08, 1.350, 0.005, sync, xlarge
+worker-08, 1.355, 8.000, recv, small
+worker-08, 9.355, 0.005, kernel-B, medium
+worker-08, 9.360, 0.050, link, xlarge
+worker-08, 9.410, 8.000, init, xlarge
+worker-08, 17.410, 0.005, write, medium
+worker-08, 17.115, 0.005, sort, xlarge
+worker-08, 17.120, 0.500, send, xlarge
+io-thread-09, 0.000, 1.000, kernel-A, medium
+io-thread-09, 1.500, 5.000, recv, tiny
+io-thread-09, 6.200, 2.500, recv, tiny
+io-thread-09, 8.400, 1.000, write, medium
+io-thread-09, 9.400, 5.000, serialize, large
+io-thread-09, 14.900, 0.000, read, tiny
+io-thread-09, 14.900, 8.000, compute, medium
+io-thread-09, 23.400, 8.000, write, large
+io-thread-09, 31.400, 0.000, send, medium
+io-thread-09, 31.100, 1.000, kernel-A, small
+worker-10, 0.000, 1.000, serialize, large
+worker-10, 1.500, 0.050, kernel-B, small
+worker-10, 1.550, 2.500, init, small
+worker-10, 4.050, 2.500, checkpoint, small
+worker-10, 6.550, 0.050, merge, tiny
+worker-10, 6.700, 0.000, optimize, small
+worker-10, 6.700, 5.000, alloc, medium
+worker-10, 11.700, 0.200, init, small
+worker-10, 12.000, 1.000, recv, tiny
+worker-10, 13.000, 1.000, flush, xlarge
+worker-11, 0.100, 8.000, ping, tiny
+worker-11, 7.800, 0.500, fsync, xlarge
+worker-11, 8.300, 0.000, read, xlarge
+worker-11, 8.400, 1.000, sort, medium
+worker-11, 9.500, 8.000, read, large
+worker-11, 18.000, 0.200, recv, tiny
+worker-11, 18.300, 0.000, parse, xlarge
+worker-11, 18.300, 1.000, compile, tiny
+worker-11, 19.300, 1.000, checkpoint, tiny
+worker-11, 20.300, 8.000, send, large
+io-thread-12, 0.000, 2.500, merge, medium
+io-thread-12, 3.000, 0.050, kernel-A, large
+io-thread-12, 3.150, 0.050, poll, xlarge
+io-thread-12, 3.200, 2.500, serialize, tiny
+io-thread-12, 5.700, 0.500, kernel-A, large
+io-thread-12, 6.700, 1.000, link, large
+io-thread-12, 7.800, 0.200, parse, large
+io-thread-12, 8.000, 0.005, send, xlarge
+io-thread-12, 8.505, 1.000, sync, small
+io-thread-12, 9.505, 0.200, kernel-A, small
+worker-13, 0.000, 0.000, kernel-B, large
+worker-13, 0.500, 0.005, link, large
+worker-13, 1.005, 0.200, merge, large
+worker-13, 1.305, 2.500, kernel-B, small
+worker-13, 3.505, 0.005, compile, small
+worker-13, 3.510, 8.000, link, tiny
+worker-13, 12.010, 0.200, read, large
+worker-13, 12.210, 5.000, checkpoint, xlarge
+worker-13, 17.710, 1.000, link, xlarge
+worker-13, 19.210, 2.500, serialize, large
+worker-14, 0.000, 5.000, link, medium
+worker-14, 5.000, 0.500, parse, large
+worker-14, 5.500, 0.500, link, tiny
+worker-14, 6.000, 0.200, recv, medium
+worker-14, 6.200, 8.000, sync, small
+worker-14, 14.200, 0.200, free, small
+worker-14, 14.400, 0.005, compile, large
+worker-14, 14.405, 8.000, link, large
+worker-14, 22.105, 0.200, compile, large
+worker-14, 22.805, 0.000, dispatch, large
+io-thread-15, 0.100, 0.000, alloc, medium
+io-thread-15, 0.200, 2.500, serialize, xlarge
+io-thread-15, 3.200, 0.200, optimize, small
+io-thread-15, 3.400, 2.500, optimize, tiny
+io-thread-15, 6.000, 1.000, checkpoint, large
+io-thread-15, 7.000, 5.000, write, xlarge
+io-thread-15, 12.500, 0.000, free, xlarge
+io-thread-15, 13.000, 0.000, sync, large
+io-thread-15, 13.000, 5.000, fsync, tiny
+io-thread-15, 18.000, 2.500, ping, small
+worker-16, 0.100, 1.000, ping, large
+worker-16, 1.100, 2.500, recv, tiny
+worker-16, 3.700, 0.000, sort, xlarge
+worker-16, 3.400, 1.000, kernel-B, tiny
+worker-16, 4.100, 0.000, kernel-B, small
+worker-16, 3.800, 0.050, kernel-B, small
+worker-16, 3.950, 0.005, dispatch, small
+worker-16, 4.055, 0.500, alloc, small
+worker-16, 5.055, 0.005, fsync, medium
+worker-16, 4.760, 0.000, send, xlarge
+worker-17, 0.100, 2.500, merge, small
+worker-17, 2.300, 0.200, read, medium
+worker-17, 3.000, 0.005, dispatch, tiny
+worker-17, 3.005, 8.000, compile, medium
+worker-17, 10.705, 8.000, flush, medium
+worker-17, 18.405, 2.500, optimize, tiny
+worker-17, 21.005, 1.000, flush, large
+worker-17, 22.005, 2.500, fsync, xlarge
+worker-17, 24.505, 8.000, optimize, large
+worker-17, 32.605, 0.500, ping, small
+io-thread-18, 0.000, 0.500, link, small
+io-thread-18, 0.600, 2.500, ping, tiny
+io-thread-18, 3.200, 1.000, fsync, large
+io-thread-18, 4.200, 1.000, recv, medium
+io-thread-18, 5.200, 0.500, serialize, tiny
+io-thread-18, 6.200, 0.200, sync, small
+io-thread-18, 6.500, 5.000, serialize, small
+io-thread-18, 11.600, 5.000, link, tiny
+io-thread-18, 16.300, 0.500, kernel-B, large
+io-thread-18, 16.800, 0.500, checkpoint, xlarge
+worker-19, 0.000, 5.000, serialize, xlarge
+worker-19, 5.000, 2.500, sort, xlarge
+worker-19, 7.500, 1.000, merge, large
+worker-19, 8.500, 0.500, recv, small
+worker-19, 8.700, 0.200, ping, tiny
+worker-19, 9.400, 0.050, kernel-A, small
+worker-19, 9.550, 0.500, sort, xlarge
+worker-19, 10.550, 0.500, read, small
+worker-19, 11.050, 0.200, alloc, small
+worker-19, 11.250, 0.000, merge, xlarge
+worker-20, 0.000, 0.500, compute, tiny
+worker-20, 1.000, 0.500, merge, small
+worker-20, 1.600, 0.005, init, xlarge
+worker-20, 1.605, 5.000, optimize, large
+worker-20, 6.605, 0.050, compute, medium
+worker-20, 6.755, 0.005, sync, large
+worker-20, 6.860, 0.005, dispatch, tiny
+worker-20, 6.865, 0.050, dispatch, medium
+worker-20, 6.615, 0.200, read, xlarge
+worker-20, 6.915, 0.200, parse, large
+io-thread-21, 0.100, 5.000, send, xlarge
+io-thread-21, 5.200, 0.500, dispatch, xlarge
+io-thread-21, 5.400, 0.005, kernel-A, small
+io-thread-21, 5.405, 0.005, fsync, small
+io-thread-21, 5.410, 8.000, sync, small
+io-thread-21, 13.110, 2.500, link, xlarge
+io-thread-21, 15.710, 0.500, compute, small
+io-thread-21, 16.210, 0.500, merge, large
+io-thread-21, 16.410, 0.200, recv, xlarge
+io-thread-21, 16.610, 2.500, read, xlarge
+worker-22, 0.000, 0.050, recv, small
+worker-22, 0.000, 0.000, fsync, medium
+worker-22, 0.500, 0.500, link, tiny
+worker-22, 1.100, 0.500, merge, large
+worker-22, 1.600, 8.000, serialize, large
+worker-22, 9.700, 0.005, poll, tiny
+worker-22, 9.805, 1.000, poll, medium
+worker-22, 10.505, 0.005, kernel-B, xlarge
+worker-22, 11.010, 0.000, checkpoint, medium
+worker-22, 11.510, 0.000, fsync, large
+worker-23, 0.500, 5.000, recv, small
+worker-23, 6.000, 2.500, flush, large
+worker-23, 8.200, 5.000, alloc, large
+worker-23, 13.200, 1.000, checkpoint, tiny
+worker-23, 14.200, 1.000, compile, large
+worker-23, 15.200, 2.500, serialize, tiny
+worker-23, 17.800, 0.005, ping, medium
+worker-23, 17.805, 0.005, free, xlarge
+worker-23, 17.510, 8.000, link, large
+worker-23, 25.210, 0.200, parse, medium
+io-thread-24, 0.500, 5.000, flush, large
+io-thread-24, 5.200, 0.200, recv, xlarge
+io-thread-24, 5.400, 0.500, link, large
+io-thread-24, 5.600, 0.000, flush, xlarge
+io-thread-24, 5.600, 0.050, send, xlarge
+io-thread-24, 5.350, 8.000, compile, tiny
+io-thread-24, 13.350, 0.005, link, tiny
+io-thread-24, 13.355, 5.000, merge, medium
+io-thread-24, 18.855, 0.500, compile, large
+io-thread-24, 19.455, 0.200, link, xlarge
+worker-25, 0.000, 2.500, kernel-A, xlarge
+worker-25, 3.000, 0.050, sync, medium
+worker-25, 3.150, 1.000, parse, medium
+worker-25, 3.850, 0.500, sort, medium
+worker-25, 4.850, 5.000, write, large
+worker-25, 10.350, 5.000, alloc, medium
+worker-25, 15.850, 8.000, free, large
+worker-25, 23.850, 0.200, merge, small
+worker-25, 24.550, 2.500, kernel-B, large
+worker-25, 26.750, 1.000, sort, large
+worker-26, 0.100, 2.500, checkpoint, small
+worker-26, 2.700, 0.000, write, xlarge
+worker-26, 3.200, 1.000, read, large
+worker-26, 3.900, 8.000, link, tiny
+worker-26, 11.900, 2.500, flush, small
+worker-26, 14.100, 5.000, recv, medium
+worker-26, 19.600, 2.500, flush, tiny
+worker-26, 22.100, 8.000, free, medium
+worker-26, 30.200, 8.000, compute, xlarge
+worker-26, 37.900, 0.200, flush, medium
+io-thread-27, 0.000, 0.005, compile, tiny
+io-thread-27, 0.000, 5.000, fsync, medium
+io-thread-27, 4.700, 0.000, ping, tiny
+io-thread-27, 4.700, 1.000, alloc, large
+io-thread-27, 5.700, 0.200, parse, large
+io-thread-27, 6.400, 0.050, fsync, small
+io-thread-27, 6.150, 2.500, poll, small
+io-thread-27, 8.750, 0.050, kernel-B, large
+io-thread-27, 8.800, 5.000, recv, tiny
+io-thread-27, 13.900, 0.500, checkpoint, xlarge
+worker-28, 0.000, 0.005, link, medium
+worker-28, 0.505, 0.500, flush, large
+worker-28, 1.005, 5.000, send, small
+worker-28, 6.105, 5.000, read, small
+worker-28, 11.205, 1.000, dispatch, medium
+worker-28, 12.205, 0.000, checkpoint, large
+worker-28, 12.205, 0.000, dispatch, tiny
+worker-28, 12.705, 5.000, send, small
+worker-28, 18.205, 1.000, kernel-B, small
+worker-28, 19.705, 0.500, checkpoint, small
+worker-29, 0.000, 0.000, send, large
+worker-29, 0.000, 1.000, sort, small
+worker-29, 0.700, 0.500, ping, large
+worker-29, 1.200, 0.200, write, xlarge
+worker-29, 1.400, 8.000, parse, medium
+worker-29, 9.400, 0.500, optimize, medium
+worker-29, 9.900, 0.005, link, tiny
+worker-29, 9.905, 0.200, checkpoint, large
+worker-29, 10.605, 1.000, sync, large
+worker-29, 11.305, 0.500, serialize, tiny
+io-thread-30, 0.100, 1.000, checkpoint, medium
+io-thread-30, 1.600, 2.500, flush, medium
+io-thread-30, 3.800, 0.200, optimize, tiny
+io-thread-30, 4.500, 8.000, ping, xlarge
+io-thread-30, 12.500, 0.005, flush, large
+io-thread-30, 12.505, 2.500, read, small
+io-thread-30, 14.705, 0.000, send, large
+io-thread-30, 14.405, 0.005, kernel-B, xlarge
+io-thread-30, 14.410, 2.500, link, medium
+io-thread-30, 17.410, 2.500, dispatch, small
+worker-31, 0.100, 0.005, optimize, xlarge
+worker-31, 0.205, 0.500, compute, medium
+worker-31, 0.705, 5.000, link, small
+worker-31, 5.705, 0.005, checkpoint, medium
+worker-31, 6.210, 1.000, compute, large
+worker-31, 7.210, 0.200, read, large
+worker-31, 7.110, 0.200, flush, xlarge
+worker-31, 7.010, 0.000, ping, small
+worker-31, 7.010, 0.200, sync, xlarge
+worker-31, 7.210, 0.200, kernel-B, medium
diff --git a/gantt/gantt.py b/gantt/gantt.py
new file mode 100755
index 0000000..71bca9e
--- /dev/null
+++ b/gantt/gantt.py
@@ -0,0 +1,313 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""gantt.py — Gantt chart for program task timing analysis
+
+Input format (see input.txt in the same directory):
+    timestamp: <YYYY-mm-dd HH:MM:SS.ffffff>
+
+    gantt results:
+    #, begin, duration, <tag1>, <tag2>, ...
+    <worker>, <begin>, <duration>, <tag1 value>, <tag2 value>, ...
+
+  - worker (the # column) is a string deciding which row the task is drawn on;
+  - begin / duration are seconds relative to timestamp (float);
+  - remaining columns are custom tags, rendered onto bars via --label format string.
+
+Usage examples:
+    python3 gantt.py input.txt
+    python3 gantt.py input.txt -o result.png --label '{task} ({duration}s)'
+    python3 gantt.py input.txt --color-by task --label '{task}'
+
+Placeholders available in the --label format string:
+  {tag_name}     tag names defined in the header, e.g. {task}, {size}
+  {c<N>} / {N}   raw CSV column index (0=worker, 1=begin, 2=duration, tags from 3)
+  {worker} {begin} {duration} {end}   built-in fields
+"""
+import argparse
+import re
+import sys
+from dataclasses import dataclass
+from pathlib import Path
+
+import matplotlib
+# 使用Anti-Grain Geometry图像渲染后端，避免在无GUI环境运行时出现错误
+matplotlib.use("Agg")
+import matplotlib.colors
+import matplotlib.pyplot as plt
+
+# 全局配置
+BASE_COLORMAP = "Set3"  # 基础配色方案（matplotlib定性色条）
+ALT_LIGHTEN = 0.35      # 交替明度提升参数，控制同一行相邻任务深浅差异
+FIG_W = 16.0            # 输出图像宽度，单位：英寸
+FIG_ASPECT = 9 / 16     # 默认高度:宽度
+BASE_ROWS = 32          # 基准行数：不超过时维持默认比例，超过后保持每行高度按比例增高
+
+# 由BASE_COLORMAP锚点插值构建连续渐变COLORMAP
+COLORMAP = matplotlib.colors.LinearSegmentedColormap.from_list(
+    "gradient", matplotlib.colormaps[BASE_COLORMAP].colors)
+
+
+# 公共函数
+# 从COLORMAP上均匀采样n个颜色
+def sample_colors(n):
+    if n <= 1:
+        return [COLORMAP(0.5)]
+    return [COLORMAP(i / (n - 1)) for i in range(n)]
+
+
+# 提升颜色rgb明度，amount表示向白色混合的百分比，取值为0-1
+def lighten(rgb, amount):
+    """向白色混合 amount，略微提高明度。"""
+    return tuple(c + (1 - c) * amount for c in rgb[:3])
+
+
+# 自然排序：划分数字和文本段，每段逐次比较，数字段优先文本段，数字段按数值排序，文本段按字典序排序
+def natural_key(s):
+    return [(0, int(p)) if p.isdigit() else (1, p) for p in re.split(r"(\d+)", s) if p]
+
+
+@dataclass(frozen=True)
+class Task:
+    worker: str
+    begin: float
+    duration: float
+    tags: tuple
+
+
+@dataclass(frozen=True)
+class GanttData:
+    """输入文件解析结果。
+
+    meta: 头部 key: value（如 timestamp）
+    columns: tag 列名（无表头时为 c3/c4/... 形式的兜底命名）
+    tasks: 解析出的任务列表
+    """
+    meta: dict
+    columns: tuple
+    tasks: tuple
+
+
+# ---------------------------------------------------------------- 解析输入
+
+def _warn(lineno, reason, raw):
+    print(f"warning: line {lineno} {reason}, skipped: {raw!r}", file=sys.stderr)
+
+
+def _parse_row(fields, lineno, raw):
+    """解析一条数据行为 Task，字段不足或数值非法时告警并返回 None。"""
+    if len(fields) < 3:
+        _warn(lineno, "has too few fields", raw)
+        return None
+    try:
+        begin, duration = float(fields[1]), float(fields[2])
+    except ValueError:
+        _warn(lineno, "begin/duration not numeric", raw)
+        return None
+    return Task(fields[0], begin, duration, tuple(fields[3:]))
+
+
+def parse_input(path):
+    """Parse the input file, returning a GanttData."""
+    meta, columns, tasks = {}, [], []
+    in_results = False
+    for lineno, raw in enumerate(path.read_text(encoding="utf-8").splitlines(), 1):
+        line = raw.strip()
+        if not line:
+            continue
+        if not in_results:
+            if line == "gantt results:":
+                in_results = True
+            else:
+                key, sep, value = line.partition(":")
+                if sep:
+                    meta[key.strip()] = value.strip()
+            continue
+        fields = [f.strip() for f in line.split(",")]
+        if not columns:
+            # 表头行：第二列为 begin 则认为是表头，否则首行即数据，tag 按列号命名
+            if len(fields) >= 3 and fields[1].lower() == "begin":
+                columns = fields[3:]
+                continue
+            columns = [f"c{3 + i}" for i in range(max(len(fields) - 3, 0))]
+        task = _parse_row(fields, lineno, raw)
+        if task:
+            tasks.append(task)
+    # 统一 tag 数量与列数一致
+    if columns:
+        tasks = [Task(t.worker, t.begin, t.duration,
+                      (t.tags + ("",) * len(columns))[: len(columns)]) for t in tasks]
+    return GanttData(meta, tuple(columns), tuple(tasks))
+
+
+def fmt_num(value):
+    return f"{value:.6g}"
+
+
+def make_label(fmt, task, columns):
+    """Render the --label format string onto a bar. Unknown placeholders become empty."""
+    ctx = {f"c{i + 3}": tag for i, tag in enumerate(task.tags)}
+    for i, name in enumerate(columns):
+        ctx[name] = task.tags[i] if i < len(task.tags) else ""
+    ctx.update(
+        c0=task.worker, c1=fmt_num(task.begin), c2=fmt_num(task.duration),
+        worker=task.worker, begin=fmt_num(task.begin),
+        duration=fmt_num(task.duration), end=fmt_num(task.begin + task.duration),
+    )
+
+    class _Ctx(dict):
+        def __missing__(self, key):
+            return ""
+
+    # {N} 等价于 {cN}
+    return re.sub(r"\{(\d+)\}", r"{c\1}", fmt).format_map(_Ctx(ctx))
+
+
+def resolve_column(name, columns):
+    """Resolve the --color-by argument to a raw CSV column index; None if unresolvable."""
+    if name in columns:
+        return 3 + columns.index(name)
+    if name in ("worker", "begin", "duration"):
+        return ("worker", "begin", "duration").index(name)
+    m = re.fullmatch(r"c?(\d+)", name)
+    return int(m.group(1)) if m else None
+
+
+def field_at(task, col_idx):
+    if col_idx == 0:
+        return task.worker
+    if col_idx == 1:
+        return fmt_num(task.begin)
+    if col_idx == 2:
+        return fmt_num(task.duration)
+    return task.tags[col_idx - 3] if col_idx - 3 < len(task.tags) else ""
+
+
+# ---------------------------------------------------------------- 绘图
+
+def render(data, args):
+    meta, columns, tasks = data.meta, data.columns, data.tasks
+    # 行排序方式：natural / lex / appearance
+    if args.sort == "lex":
+        workers = sorted({t.worker for t in tasks})
+    elif args.sort == "appearance":
+        workers = list(dict.fromkeys(t.worker for t in tasks))
+    else:
+        workers = sorted({t.worker for t in tasks}, key=natural_key)
+    row_of = {w: i for i, w in enumerate(workers)}
+
+    # --color-by：按列值从彩虹 colorbar 均匀采样分类色
+    # 默认按 worker 着色，同一 worker 内相邻任务交替深浅，避免连在一起的段看不清
+    has_legend = False
+    legend_items = []
+    if args.color_by:
+        col_idx = resolve_column(args.color_by, columns)
+        if col_idx is None:
+            print(f"error: cannot resolve --color-by column '{args.color_by}', "
+                  f"available: {', '.join(columns) or '(no tag columns)'}", file=sys.stderr)
+            sys.exit(1)
+        values = list(dict.fromkeys(field_at(t, col_idx) for t in tasks))
+        value_color = dict(zip(values, sample_colors(len(values))))
+        task_color = {id(t): value_color[field_at(t, col_idx)] for t in tasks}
+        legend_items = list(value_color.items())
+        has_legend = len(legend_items) >= 2
+    else:
+        worker_color = dict(zip(workers, sample_colors(len(workers))))
+        seq = {}
+        task_color = {}
+        for t in tasks:
+            i = seq.get(t.worker, 0)
+            task_color[id(t)] = (worker_color[t.worker] if i % 2 == 0
+                                 else lighten(worker_color[t.worker], ALT_LIGHTEN))
+            # 宽度接近零的任务不参与同行颜色 toggle
+            if t.duration > 0:
+                seq[t.worker] = i + 1
+
+    # 默认 16:9；行数超过 BASE_ROWS 时保持每行高度不变，按比例增大图高
+    fig_h = FIG_W * FIG_ASPECT * max(1.0, len(workers) / BASE_ROWS)
+    fig, ax = plt.subplots(figsize=(FIG_W, fig_h))
+    labels = []
+    for t in tasks:
+        y = row_of[t.worker]
+        ax.broken_barh([(t.begin, max(t.duration, 1e-9))], (y - 0.5, 1.0),
+                       facecolors=task_color[id(t)], edgecolor="none")
+        if args.label:
+            artist = ax.text(t.begin + t.duration / 2, y,
+                             make_label(args.label, t, columns),
+                             ha="center", va="center", fontsize=8)
+            labels.append((artist, t.begin, t.begin + t.duration))
+
+    # label 文字宽度超过任务 bar 宽度则不显示
+    if labels:
+        fig.canvas.draw()
+        renderer = fig.canvas.get_renderer()
+        for artist, x0, x1 in labels:
+            bar_px = abs(ax.transData.transform((x1, 0))[0]
+                         - ax.transData.transform((x0, 0))[0])
+            if artist.get_window_extent(renderer).width > bar_px:
+                artist.remove()
+
+    ax.set_ylim(len(workers) - 0.5, -0.5)  # 首个 worker 在最上方
+    ax.margins(x=0)  # 横坐标左右两端不留空隙
+    ax.set_yticks(range(len(workers)))
+    ax.set_yticklabels(workers)
+    ax.tick_params(axis="y", length=0)
+    ax.grid(axis="x", linewidth=0.8, alpha=0.5)
+    ax.set_axisbelow(True)
+    ax.set_xlabel("time (s)")
+
+    title = args.title or "Gantt Chart"
+    sub = meta.get("timestamp")
+    if sub:
+        ax.set_title(f"{title}\n{sub}", fontsize=12)
+    else:
+        ax.set_title(title, fontsize=12)
+    if has_legend:
+        ax.legend(handles=[plt.Rectangle((0, 0), 1, 1, color=c, label=v)
+                           for v, c in legend_items],
+                  loc="lower left", bbox_to_anchor=(0, 1.02), ncol=8,
+                  frameon=False, fontsize=9)
+
+    fig.savefig(args.output, bbox_inches="tight")
+
+
+# ---------------------------------------------------------------- CLI
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Draw a Gantt chart from program task timing data "
+                    "(input format see input.txt)")
+    parser.add_argument("input", help="input file path")
+    parser.add_argument("-o", "--output", help="output image path, default <input>.png")
+    parser.add_argument("--label",
+                        help="format string for bar labels, e.g. '{task} ({duration}s)', "
+                             "see module docstring")
+    parser.add_argument("--color-by",
+                        help="color bars by this column (tag name or index like c3), "
+                             "default single color")
+    parser.add_argument("--title", help="custom title")
+    parser.add_argument("--sort", choices=["natural", "lex", "appearance"],
+                        default="natural",
+                        help="worker row order: natural (Windows-style, default), "
+                             "lex (lexicographic), appearance (first-seen order)")
+    args = parser.parse_args()
+
+    path = Path(args.input)
+    if not path.is_file():
+        print(f"error: input path '{args.input}' does not exist or is not a file",
+              file=sys.stderr)
+        sys.exit(1)
+
+    data = parse_input(path)
+
+    if not data.tasks:
+        print("error: no valid tasks parsed (missing 'gantt results:' section?)",
+              file=sys.stderr)
+        sys.exit(1)
+
+    args.output = args.output or str(path.with_suffix(".png"))
+    render(data, args)
+    print(f"gantt saved to {args.output}")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/gantt/input.txt b/gantt/input.txt
new file mode 100644
index 0000000..88c23c3
--- /dev/null
+++ b/gantt/input.txt
@@ -0,0 +1,7 @@
+timestamp: 2026-09-08 19:38:07.431900
+
+gantt results:
+#, begin, duration, <tag1>, <tag2>
+...
+
+此后每行，#打印worker_id，可以是string，表示任务在甘特图的哪个行上；begin表示任务开始时间距离timestamp的时间，浮点型，单位s；duration表示任务长度，浮点型，单位s；后面的tag则是任务的自定义字段，脚本支持输入一个包含tag的格式化字符串，将字符串内容输入到对应任务的甘特图bar上
\ No newline at end of file
diff --git a/gantt/test_input.png b/gantt/test_input.png
new file mode 100644
index 0000000..11bc29f
Binary files /dev/null and b/gantt/test_input.png differ
diff --git a/gantt/test_input.txt b/gantt/test_input.txt
new file mode 100644
index 0000000..897b2fc
--- /dev/null
+++ b/gantt/test_input.txt
@@ -0,0 +1,16 @@
+timestamp: 2026-09-08 19:38:07.431900
+
+gantt results:
+#, begin, duration, task, size
+main, 0.0, 0.8, init, small
+main, 0.8, 12.5, compute, large
+main, 13.3, 0, sync, small
+main, 13.3, 1.2, cleanup, small
+io-worker, 1.0, 5.5, read, medium
+io-worker, 7.0, 4.0, write, medium
+io-worker, 11.5, 0.02, fsync, small
+gpu, 1.5, 3.25, kernel-A, large
+gpu, 5.0, 6.75, kernel-B, large
+net, 2.0, 2.5, recv, medium
+net, 9.0, 3.0, send, medium
+net, 14.5, 0.001, ping, small