plotify-charts 0.1.0

A Python plotting package with a dual Seaborn/Matplotlib + Plotly backend.

Plotify

Plotify is a Python plotting package built around two ideas:

1. Auto-pick — given a dataframe and column names, Plotify infers the right chart from the data shape and types. No need to remember which function to call.
2. Publication-ready by default — every chart ships with a colourblind-safe palette, smart tick formatting (1.2K, 3.4M), minimalist spines, and consistent typography on both backends.

Every plot can render on either Seaborn/Matplotlib (static images) or Plotly (interactive HTML).

The chart catalogue and category split (Numerical, Categoric, Num & Cat, Time series, Network, Maps) are inspired by data-to-viz.com.

Demo notebooks

The fastest way to see what Plotify can do — runnable notebooks under notebooks/ with every chart class rendered on both backends:

• 00_quickstart.ipynb — auto, suggest, themes
• 01_numerical.ipynb — Boxplot, Density, Violin, Scatter, ConnectedScatter
• 02_categorical.ipynb — all 12 categoric charts
• 03_num_cat.ipynb — GroupedBar, StackedBar, GroupedScatter, BoxPlotByGroup
• 04_timeseries.ipynb — Line, Area, StackedArea, StreamGraph
• 05_network.ipynb — Network, Chord, Sankey, Arc, HierarchicalEdgeBundling
• 06_maps.ipynb — Choropleth, Bubble, Hexbin, Cartogram, Connection

Outputs are committed, so the notebooks render fully on GitHub without running them.

Two-line example

import pandas as pd, plotify

df = pd.DataFrame({
    "date": pd.date_range("2024-01-01", periods=12, freq="MS"),
    "sales": [120, 140, 90, 200, 220, 260, 240, 310, 280, 330, 360, 410],
    "region": ["North"] * 6 + ["South"] * 6,
})

plotify.auto(df, x="date", y="sales", color="region", backend="plotly")\
       .save_plot("sales.html")

plotify.auto looks at the dtypes (datetime, numeric, low-cardinality group), picks LineChart, and returns a beautifully themed figure.

What auto actually picks

plotify.suggest(df, x="date", y="sales", color="region")
# → [Suggestion(LineChart,        score=0.96, reason="datetime x + numeric y + low-cardinality group → multi-series line chart"),
#    Suggestion(StackedAreaChart, score=0.72, reason="stacked area variant for compositional time series")]

Suggestions come ranked, each with a one-line reason — useful for debugging or for showing the user why a chart was picked.

You can also pass an intent to override the rules:

plotify.auto(df, x="region", y="sales", intent="distribution")
# → Violinplot

Themes

plotify.theme controls the look of every chart on both backends.

import plotify

plotify.theme.get_current().name      # "publication" — applied automatically
plotify.theme.set("none")             # opt out, use library defaults
plotify.theme.register(plotify.theme.Theme(name="brand", palette=("#FF1493", "#00CED1")))
plotify.theme.set("brand")

Package structure

plotify/
├── base.py              # BasePlot — backend dispatch + save
├── numerical/           # Boxplot, DensityPlot, Violinplot, ScatterPlot, ConnectedScatterPlot
├── categorical/         # BarPlot, PieChart, Treemap, Sunburst, Venn, Dendrogram, etc. (12)
├── num_cat/             # GroupedBarPlot, StackedBarPlot, GroupedScatter, BoxPlotByGroup
├── timeseries/          # LineChart, AreaChart, StackedAreaChart, StreamGraph
├── network/             # NetworkDiagram, ChordDiagram, SankeyDiagram, ArcDiagram, HierarchicalEdgeBundling
└── maps/                # ChoroplethMap, BubbleMap, HexbinMap, Cartogram, ConnectionMap

Installation

pip install plotify-charts          # PyPI distribution name
pip install plotify-charts[maps]    # adds geopandas for ChoroplethMap's seaborn path
pip install plotify-charts[full]    # all optional extras

The Python import name is plotify regardless of which install variant you pick:

import plotify

If you've cloned the repo for development, install with Poetry instead:

poetry install --with dev --extras full

Usage

import pandas as pd
from plotify.numerical import Boxplot
from plotify.categorical import BarPlot, PieChart
from plotify.timeseries import LineChart
from plotify.network import SankeyDiagram

df = pd.DataFrame({"cat": list("AAABBB"), "val": [1, 2, 3, 4, 5, 6]})

# Seaborn/Matplotlib backend → static image.
Boxplot(df, x="cat", y="val", backend="seaborn").save_plot("box.png")

# Plotly backend → interactive HTML (or static image via kaleido).
BarPlot(df, x="cat", y="val", backend="plotly").save_plot("bar.html")
PieChart(df, names="cat", values="val", backend="plotly").save_plot("pie.html")

# Sankey flow with explicit source/target/value lists.
SankeyDiagram(
    source=[0, 1, 0],
    target=[2, 2, 3],
    value=[8, 4, 2],
    labels=["A", "B", "C", "D"],
    backend="plotly",
).save_plot("sankey.html")

Every plot class accepts backend="seaborn" (default) or backend="plotly". A few classes only support one backend — WordCloudPlot, VennDiagram, CircularPacking, StreamGraph, ChordDiagram, ArcDiagram, HierarchicalEdgeBundling, and Cartogram are Seaborn-only because Plotly has no native equivalent. Requesting an unsupported backend raises a ValueError at construction time.

Chart catalogue (all implemented)

Category	Classes
Numerical	Boxplot, DensityPlot, Violinplot, ScatterPlot, ConnectedScatterPlot
Categoric	BarPlot, LollipopChart, RadarChart, ParallelCoordinates, WordCloudPlot, PieChart, DoughnutChart, Treemap, CircularPacking, SunburstDiagram, VennDiagram, Dendrogram
Num & Cat	GroupedBarPlot, StackedBarPlot, GroupedScatter, BoxPlotByGroup
Time series	LineChart, AreaChart, StackedAreaChart, StreamGraph
Network	NetworkDiagram, ChordDiagram, SankeyDiagram, ArcDiagram, HierarchicalEdgeBundling
Maps	ChoroplethMap, BubbleMap, HexbinMap, Cartogram, ConnectionMap

Development

Run the test suite:

poetry run pytest

The notebooks under notebooks/ are generated from scripts/build_notebooks.py — please don't hand-edit them. After changing an API or a chart class, rebuild and re-execute in a single step:

python scripts/build_notebooks.py --execute

The --execute flag runs each notebook via jupyter nbconvert so the rendered outputs are embedded in the committed .ipynb files. The script sets PYTHONPATH to the repo root automatically, so it works without pip-installing the package.