plotlibs

Fast, beautiful Python plotting for everyone.

Renamed: plotlib → plotlibs. Old code keeps working via a shim (import plotlib), but please migrate to import plotlibs. New home: https://github.com/salim-studio/plotlibs

import plotlibs as pl

pl.plot([1, 2, 3], [1, 4, 9], label="x²")
pl.xlabel("x"); pl.ylabel("y"); pl.legend()
pl.savefig("out.png")  # or pl.show()

Why plotlibs?

Install

pip install plotlibs                    # minimal: numpy + pillow
pip install "plotlibs[all]"             # pandas + sqlalchemy + duckdb + sklearn + matplotlib
pip install -e ".[dev]"                 # contributors

30-second tour

import plotlibs as pl

# 1) EDA in one line
df = pl.load_csv("sales.csv")
print(pl.describe(df))
pl.quick_eda(df)

# 2) Stats plots (all matplotlib-style signatures)
pl.boxplot([[1, 2, 3], [2, 3, 4]], labels=["A", "B"])
pl.violinplot([[1, 2, 3], [2, 3, 4]])
pl.kde(df["sales"])
pl.heatmap([[1, 0.8], [0.8, 1]], annot=True)
pl.corr(df, annot=True)
pl.countplot(["Cairo", "Oran", "Cairo"])
pl.area([1, 2, 3], [1, 1, 1], [2, 2, 2], labels=["a", "b"])
pl.hist2d([1, 2, 2, 3], [1, 1, 2, 2])
pl.stem([1, 2, 3], [2, 1, 3])

# 3) SQL -> plot
conn = pl.db_connect("sales.db")   # or ":memory:", "sqlite:///x.db", "duckdb://shop.ddb"
pl.to_sql(df, "sales", conn)
print(pl.db.list_tables(conn))
df2 = pl.read_sql("SELECT region, SUM(amount) FROM sales GROUP BY region", conn)
pl.db.plot_from_sql("SELECT qty, amount FROM sales LIMIT 500", conn, kind="scatter")

# 4) ML / DL
pl.plot_history({"loss": [0.9, 0.5, 0.3], "val_loss": [1.0, 0.6, 0.4]})
pl.plot_confusion_matrix(y_true, y_pred, normalize=True)
pl.plot_roc(y_true, y_score, label="xgb")
pl.plot_feature_importance(["age", "salary", "city"], [0.5, 0.3, 0.2])
pl.plot_images(mnist_images)

Visual identity

Brand: midnight ink #0F172A + indigo #4F46E5 + cyan #06B6D4, highlights lime #A3E635 / amber #FACC15, background paper #F8FAFC. Logo: rounded-square gradient with a white line chart. See assets/ and BRANDING.md.

Signature theme:

import plotlibs as pl
pl.style.use("plotlibs")        # light signature
pl.style.use("plotlibs-dark")   # dark signature
print(pl.style.available())

Matplotlib compatibility

Why faster?

1. Min-max decimation — lines over 2000 points are bucketed (min+max preserved), O(n) in numpy.
2. Pillow rasterizer — direct C-level drawing, no heavy layout engine.
3. Vectorized pipeline — data→pixel in one pass, numpy ticks/limits.
4. No heavy objects — artists are plain dicts, render is deferred.

Benchmark (n=200k, benchmarks/bench.py):

plotlibs   : ~370 ms/plot
matplotlib : ~1560 ms/plot
speedup    : ~4.2x faster

Project layout

plotlibs/              # repo root (package name == repo name)
  plotlibs/            # real package
    __init__.py  pyplot.py  figure.py  colors.py  fast.py  style.py
    data.py  db.py  eda.py  ml.py
    backends/renderer.py
  plotlib/             # deprecated shim (renamed alias)
  assets/              # logo.svg, banner.svg
  tests/  benchmarks/  examples/

Migration from plotlib

- import plotlib as pl
+ import plotlibs as pl

import plotlib still works in 0.3.x (emits DeprecationWarning) and will be removed in 1.0.

Roadmap

• 0.3.1 — Fix PyPI banner/logo images (absolute URLs), polish docs
• 0.3.0 — Rename to plotlibs, English docs, brand identity
• 0.4.0 — Interactive HTML export (pl.save_html), dashboards, RTL/Arabic labels
• 0.5.0 — Geo + time-series + sklearn/HF pipeline helpers
• Suggest ideas: open an issue with [idea] at https://github.com/salim-studio/plotlibs/issues

Contributing

python -m pytest tests -q
python examples/demo.py
python examples/demo_fame.py

License

MIT — see LICENSE.