Skip to main content

Sandbox — pandas & charts

This page is a lab, not a student chapter. It checks three ways to “chart” on Piston:

  1. Pandas tables (already fast)
  2. ASCII bars (testable, no image)
  3. matplotlib PNG (needs longer CPU time + a PISTON_PNG: line)
  4. seaborn (same PNG trick; seaborn sits on matplotlib)

Run with batch mode (interactive off) so the full PNG arrives in one response.

Print a PNG like this:

print("PISTON_PNG:" + base64.b64encode(png_bytes).decode("ascii"))

The runner hides that line and shows the image.

1. Pandas table​

2. ASCII bar chart​

Graded exams can check this output. No matplotlib.

3. matplotlib PNG​

Uses Agg (no window). First import is slow; Piston CPU/time was raised for this.

If this is killed (Time limit exceeded), restart Piston after the compose change:

cd /Users/nenewang/Documents/GIT/piston
docker compose up -d api

Limits in docker-compose.yaml: PISTON_RUN_TIMEOUT=20000, PISTON_RUN_CPU_TIME=15000, PISTON_OUTPUT_MAX_SIZE=250000.

4. seaborn PNG​

Same Agg backend. sns.set_theme() then a barplot.