Matplotlib figures: render a labeled report without a display server

Operation contract

The dispatch report uses the Agg backend before importing pyplot, then saves a labeled bar chart to an owned byte buffer. It closes the figure after saving. The PNG signature check confirms the returned container format; it does not evaluate whether the chart communicates its data honestly. Minor-unit labels prevent the displayed numbers from being mistaken for rupee amounts.

Failure and ownership boundary

This example uses fixture totals, no remote images and no browser screenshot. Rendering a chart does not repair incorrect aggregation. Repeated unclosed figures can retain memory even after an output buffer is discarded. Pandas groupby: retain missing keys and define all-null totals and Python ExitStack: unwind partially acquired resources provide the preceding and following boundaries.

Tested environment

Dependency check: this program was executed on CPython 3.14.6 with matplotlib==3.11.2. Install these versions in a separate virtual environment. The download includes the recorded environment snapshot; no third-party package is part of the website runtime.

Working program

from io import BytesIO
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt

figure, axes = plt.subplots(figsize=(6, 3.5), constrained_layout=True)
try:
    axes.bar(["DEL", "BOM", "BLR"], [375, 250, 175], color="#286c83", width=0.6)
    axes.set(title="Accepted receipts by region", ylabel="Amount (minor units)")
    axes.set_ylim(bottom=0)
    axes.spines[["top", "right"]].set_visible(False)
    with BytesIO() as output:
        figure.savefig(output, format="png", dpi=120)
        image = output.getvalue()
    print("PNG container:", image.startswith(b"\x89PNG\r\n\x1a\n"))
finally:
    plt.close(figure)
print("open figures:", len(plt.get_fignums()))

Output

PNG container: True
open figures: 0

Costs and limits

Rendering work depends on artists, pixel dimensions and backend. Raster output retains pixel-related storage; increasing each dimension multiplies the pixel count. This fixture does not print compressed byte size because rendering details can change across package versions.

Common Mistakes

• Select a noninteractive backend before importing pyplot.
• A saved image does not automatically release its figure.

Connected lessons

Pandas groupby: retain missing keys and define all-null totals, Pandas chunked CSV aggregation: validate every chunk before returning totals, Python ExitStack: unwind partially acquired resources.