driedfigs

Preserve your matplotlib figures with their data for reproducible research.

driedfigs makes it simple to save the data behind your matplotlib plots alongside your figures, improving reproducibility in academic publications.

Key Features

• Zero effort reproducibility: Automatically capture plot data with minimal code changes
• Non-intrusive: Works with existing matplotlib code - just add name parameters
• Flexible: Supports most matplotlib plotting functions (plot, scatter, contour, etc.)
• Simple API: Just two main functions - start_figure() and save_figuredata()

Installation

pip install driedfigs

For xarray support (automatic conversion of DataArrays to numpy):

pip install driedfigs[xarray]

Quick Start

import driedfigs as df
import matplotlib.pyplot as plt
import numpy as np

# 1. Start capturing data for a figure
fig_data = df.start_figure("example_plot")

# 2. Create plots with the 'name' parameter to identify each artist
x = np.linspace(0, 10, 100)
plt.plot(x, np.sin(x), name="sin_wave")
plt.scatter([1, 2, 3], [0.5, 0.8, 1.2], name="data_points")

# 3. Save the data
df.save_figuredata("./data/example_plot.pkl")

# 4. Save the figure (as you would normally)
plt.savefig("./figures/example_plot.pdf")

This creates:

• ./figures/example_plot.pdf (the figure, saved normally)
• ./data/example_plot.pkl (the captured data)

The way driedfigs works is simple:

• When driedfigs is imported, it overwrites commonly used matplotlib plotting functions and methods with "hooked" versions of the function/method that first store the data before calling the function/method normally
• When you call start_figure, a FigureData object for storing that figure's data is created
• When you call any of the "hooked" functions/methods with a name keyword argument, the data is stored in the current FigureData object
• When you call save_figuredata, all of the data in the FigureData object is saved to disk using python's pickle module

Detailed Usage

Basic Workflow

import driedfigs as df
import matplotlib.pyplot as plt

# Start a new figure
fig_data = df.start_figure("my_analysis", verbose=True)

# Make plots - use 'name' to capture data
plt.plot(x_data, y_data, name="raw_data")
plt.plot(x_model, y_model, '--', name="model_fit")

# Save the data
df.save_figuredata("./data/my_analysis.pkl")

# Save the figure (standard matplotlib)
plt.savefig("./figures/my_analysis.pdf")
plt.savefig("./figures/my_analysis.png", dpi=300)

Loading Saved Data

import pickle

# Load the saved data
with open("data/my_analysis.pkl", "rb") as f:
    data = pickle.load(f)

# Access specific plot data
raw_data = data["raw_data"]  # Returns tuple of args passed to plt.plot
x, y = raw_data[0], raw_data[1]

Supported Matplotlib Functions

pyplot functions:

plot, scatter, bar, hist, imshow, contour, contourf, pcolormesh, fill_between, errorbar, boxplot, violin, pie, polar, loglog, semilogx, semilogy

Axes methods:

plot, scatter, bar, hist, imshow, contour, contourf, pcolormesh, fill_between, errorbar, boxplot, violin

Managing Multiple Figures

# Create separate figure data objects
fig1 = df.start_figure("figure_1")
plt.plot(x1, y1, name="data1")
fig1.save("./data/figure1.pkl")  # Use the .save() method

fig2 = df.start_figure("figure_2")
plt.plot(x2, y2, name="data2")
fig2.save("./data/figure2.pkl")

Real-World Example

import driedfigs as df
import matplotlib.pyplot as plt
import numpy as np

# Simulate some research data
time = np.linspace(0, 10, 100)
measurement = np.sin(time) + np.random.normal(0, 0.1, 100)
model = np.sin(time)

# Start capturing
df.start_figure("experiment_results")

# Create publication-quality plot
plt.figure(figsize=(10, 6))
plt.plot(time, measurement, 'o', alpha=0.5, label="Measurements", name="raw_measurements")
plt.plot(time, model, '-', linewidth=2, label="Model", name="theoretical_model")
plt.xlabel("Time (s)")
plt.ylabel("Signal (V)")
plt.legend()
plt.title("Experimental Results vs. Theoretical Model")

# Save the data for reproducibility
df.save_figuredata("./manuscript/data/experiment_results.pkl")

# Save the figure
plt.savefig("./manuscript/figures/experiment_results.pdf")
plt.savefig("./manuscript/figures/experiment_results.png", dpi=300)

# Now you can upload ./manuscript/ with your paper submission!

API Reference

Main Functions

• start_figure(name, verbose=False): Begin capturing data for a new figure

  • Returns a FigureData instance
  • Automatically installs hooks on first use

• save_figuredata(output_path, figuredata=None): Save captured data to a pickle file

  • If figuredata is None, uses the currently active FigureData

FigureData Class

• save(output_path): Save this figure's data to file
• figure_name: Name of the figure
• figure_data: Dictionary mapping artist names to their data
• is_active: Whether this FigureData is currently active