Matplotlib Dynamic Downsampler

High Performance Time-Series Visualization

A python package for universal data downsampling and interactive visualization using Matplotlib and tsdownsampler.

matplotlib_resampler improves scalability of Matplotlib for visualizing massive time-series datasets. The library dynamically resamples time-series data respective to the current graph view, ensuring fast, responsive updates during panning or zooming.

This aggregation functionality is achieved by utilizing highly optimized data point selection algorithms from tsdownsample. Our default data aggregation method is MinMaxLTTB (defaulting to 5,000 points per view), which guarantees maximum visual fidelity and preserves narrow data spikes.

This works with more than 1 plot and can visualize over 1 Billion data points.

Features

• Convenient and "Plug & Play": Just pass your existing plt object to our class constructor. Preserves all original Matplotlib styling, colors, labels and figure configurations.
• No Backend Servers Required: Requires zero web servers, Dash apps, or Jupyter widget overhead. Runs entirely via Matplotlib's native event backend.
• Efficient visualization of large datasets: Dynamic zooming in and out while keeping high fidelity based on user's point preference.

Installation

Clone the repository: git clone https://github.com/JGPOG/matplotlib_resampler.git

Install the package: pip install matplotlib_resampler\dist\matplotlib_resampler-0.1.0-py3-none-any.whl

Usage

Add dynamic aggregation to your figure with minimal overhead.

import matplotlib.pyplot as plt
import numpy as np
from matplotlib_resampler import DynamicDownsampler

plt.figure(figsize=(10, 5))
plt.plot(t, signal, label="Noisy 1 Hz Sine Wave")

plt.title("1 Hz Noisy Sine Wave with 10 Random Spikes")
plt.xlabel("Time (s)")
plt.ylabel("Amplitude")
plt.grid(True, linestyle="--", alpha=0.6)
plt.legend(loc="upper right")
plt.tight_layout()
# Simply call the class and place the object
DynamicDownsampler(plt)
plt.show()

In example demo, 200,000,000 data points across 4 subplots (50M each) are visualized!

Matplotlib Standalone vs Matplotlib + Resampler

DynamicDownsampler() acts as an invisible wrapper around standard Matplotlib figures. It adds visualization scalability to line charts by resampling the data based on GUI zoom or pan events.

• Standard Matplotlib attempts to render more points than the computer may be capable of handling.
• matplotlib_enhancer intercepts these points before rendering and only plots a mathematically representative downsampled version. When you zoom in, it automatically recalculates and reveals the high resolution details for that window.

Important Considerations & Tips

• Sorting Requirement: The underlying tsdownsample Rust engine requires your X-axis data to be monotonic/sorted and free of NaN values.
• Aliasing and Signal Spikes: MinMaxLTTB ensures extreme peaks and troughs in raw data are never accidentally skipped when rendering a zoomed out view.
• None Networked/Portable Installations: If you have installed using a non networked or portable folder, there is a chance you receive an error about the figure not being interactive and it won't plot. In this case you need to install PyQt6.