conference-plot

name	conference-plot
description	Create publication-quality matplotlib figures for two-column conference papers (ACM, IEEE, USENIX). Provides style presets, colorblind-safe Wong palette, and dual-output (PDF+PNG) workflow. Use when the user wants to create, style, or fix plots/figures/charts for academic paper submission, or asks for conference-quality matplotlib figures.

Conference Paper Plotting

Quick Start

Always import plot_utils.py from this skill's scripts/ directory and use the paper_style() context manager to set up correct dimensions and rcParams.

Bar chart (ACM single-column)

import sys
sys.path.insert(0, "/home/xly/.claude/skills/conference-plot/scripts")
from plot_utils import paper_style, WONG_PALETTE, HATCHES, save_dual_output
import matplotlib.pyplot as plt
import numpy as np
from pathlib import Path

labels = ["A", "B", "C", "D"]
values = [4.2, 3.8, 5.1, 2.9]

with paper_style(width=3.3, height=2.5):
    fig, ax = plt.subplots()
    bars = ax.bar(labels, values, color=WONG_PALETTE[1:5],
                  edgecolor="black", linewidth=0.4,
                  hatch=[HATCHES[i] for i in range(len(labels))])
    ax.set_ylabel("Throughput (Gbps)")
    ax.set_xlabel("Configuration")
    save_dual_output(fig, Path("throughput.pdf"), None)
    plt.close(fig)

Line plot (USENIX single-column)

from plot_utils import paper_style, WONG_PALETTE, save_dual_output
import matplotlib.pyplot as plt
import numpy as np
from pathlib import Path

markers = ["o", "s", "^", "D", "v", "P", "X", "*"]
x = np.arange(1, 6)
with paper_style(width=3.33, height=2.5):
    fig, ax = plt.subplots()
    for i, label in enumerate(["System A", "System B", "System C"]):
        ax.plot(x, np.random.rand(5) * 10, color=WONG_PALETTE[i + 1],
                marker=markers[i], markersize=4, label=label)
    ax.set_xlabel("Thread Count")
    ax.set_ylabel("Latency (ms)")
    ax.legend()
    save_dual_output(fig, Path("latency.pdf"), None)
    plt.close(fig)

Heatmap (IEEE double-column)

from plot_utils import paper_style, save_dual_output
import matplotlib.pyplot as plt
import numpy as np
from pathlib import Path

data = np.random.rand(5, 8)
with paper_style(width=7.16, height=2.5):
    fig, ax = plt.subplots()
    im = ax.imshow(data, cmap="RdYlGn_r", aspect="auto")
    fig.colorbar(im, ax=ax, shrink=0.8)
    ax.set_xlabel("Benchmark")
    ax.set_ylabel("Configuration")
    save_dual_output(fig, Path("heatmap.pdf"), None)
    plt.close(fig)

Usage

1. Import from scripts/plot_utils.py (add the skill's scripts/ dir to sys.path).
2. Wrap all plotting code in paper_style().
3. Save with save_dual_output(fig, pdf_path, png_path). By default it saves both PDF and PNG (auto-deriving the missing path). Pass save_both=False to save only the explicitly provided path(s).

Style Reference

Color Palette (WONG_PALETTE)

Index	Hex	Color	Typical Use
0	#000000	Black	Baselines, reference
1	#EEBA0C	Orange	Primary comparison
2	#56B4E9	Sky Blue	Secondary comparison
3	#009E73	Bluish Green	Third system
4	#F0E442	Yellow	Fourth (use sparingly)
5	#0072B2	Blue	Fifth system
6	#D55E00	Vermillion	Sixth / alert
7	#CC79A7	Reddish Purple	Seventh system
8	#0000FF	Blue Saturated	Eighth system
9	#FF0000	Red Saturated	Ninth system

This is a 10-color palette based on the Wong colorblind-safe palette (Nature Methods, 2011), extended with two saturated colors. Safe for deuteranopia, protanopia, and tritanopia. Start from index 1 for data series (index 0 is black, best for baselines or outlines).

Markers

Use inline marker strings directly: "o", "s", "^", "D", "v", "P", "X", "*" — circle, square, triangle-up, diamond, triangle-down, plus-filled, X-filled, star.

Hatches (HATCHES)

['', '//', '\\\\', 'xx', '..', '++', 'OO', '**'] — none, diagonal, back-diagonal, cross-hatch, dots, plus, circles, stars.

Venue Dimensions

Venue	Single-Column	Double-Column	Recommended Call
ACM (SIGCOMM, MOBICOM, etc.)	3.3"	7.0"	paper_style(width=3.3)
IEEE (INFOCOM, MICRO, etc.)	3.5"	7.16"	paper_style(width=3.5)
USENIX (OSDI, NSDI, ATC, etc.)	3.33"	7.0"	paper_style(width=3.33)

Pass explicit width and height to paper_style(). Default is width=3.3, height=2.5.

Common Patterns

Grouped bars with hatches

x = np.arange(len(groups))
w = 0.25
for i, (series, vals) in enumerate(data.items()):
    ax.bar(x + i * w, vals, w, label=series,
           color=WONG_PALETTE[i + 1], edgecolor="black", linewidth=0.4,
           hatch=HATCHES[i + 1])
ax.set_xticks(x + w * (len(data) - 1) / 2)
ax.set_xticklabels(groups)

Dual-axis (twinx) with color-coded labels

ax1 = fig.add_subplot(111)
ax2 = ax1.twinx()
ax1.bar(x, throughput, color=WONG_PALETTE[1], label="Throughput")
ax2.plot(x, latency, color=WONG_PALETTE[5], marker="o", label="Latency")
ax1.set_ylabel("Throughput (Gbps)", color=WONG_PALETTE[1])
ax2.set_ylabel("Latency (ms)", color=WONG_PALETTE[5])
ax1.tick_params(axis="y", colors=WONG_PALETTE[1])
ax2.tick_params(axis="y", colors=WONG_PALETTE[5])

Error bars

ax.bar(x, means, yerr=stds, capsize=2, color=WONG_PALETTE[1:],
       edgecolor="black", linewidth=0.4, error_kw={"linewidth": 0.6})

Heatmap with annotation

im = ax.imshow(data, cmap="RdYlGn_r", aspect="auto")
for i in range(data.shape[0]):
    for j in range(data.shape[1]):
        ax.text(j, i, f"{data[i, j]:.1f}", ha="center", va="center", fontsize=6)
fig.colorbar(im, ax=ax, shrink=0.8)

Guidelines

• Always use paper_style() — it sets pdf.fonttype=42 and ps.fonttype=42 to avoid Type 3 font rejection by venues.
• Keep fonts >= 6pt for readability. The default 8pt base is safe.
• Use WONG_PALETTE (10 colors) for colorblind safety. Start from index 1 for data series; index 0 (black) is best for baselines or outlines.
• Export PDF for paper, PNG for review. Use save_dual_output(fig, pdf_path, png_path).
• Always plt.close(fig) after saving to avoid memory leaks in batch scripts.
• Pair hatches with colors for bar charts so the plot remains distinguishable in grayscale printouts.

Detailed Venue Specs

See references/venue-specs.md for column widths, font requirements, LaTeX \includegraphics patterns, and common gotchas (Type 3 fonts, minimum font size, vector vs raster) for ACM, IEEE, and USENIX.