pysuricata 0.0.10

A lightweight EDA tool inspired by the curious nature of suricates. Built just for fun 🔬.

pysuricata

A lightweight Python library to generate self-contained HTML reports for exploratory data analysis (EDA).

📖 Read the documentation

Installation

Install pysuricata directly from PyPI:

pip install pysuricata

Why use pysuricata?

• Instant reports: Generate clean, self-contained HTML reports directly from pandas DataFrames.
• Out-of-core option (v2): Consume in-memory DataFrame chunks and profile datasets larger than RAM.
• No heavy deps: Minimal runtime dependencies (pandas/pyarrow optional depending on source).
• Rich insights: Summaries for numeric, categorical, datetime columns, missing values, duplicates, correlations, and sample rows.
• Portable: Reports are standalone HTML (with inline CSS/JS/images) that can be easily shared.
• Customizable: Title, sample display, and output path can be tailored to your needs.

Quick Example (classic, in-memory DataFrame)

The following example demonstrates how to generate an EDA report using the Iris dataset with Pandas:

import pandas as pd
from pysuricata import profile

# Load the Iris dataset directly using Pandas
iris_url = "https://raw.githubusercontent.com/mwaskom/seaborn-data/master/iris.csv"
iris_df = pd.read_csv(iris_url)

# Build the report and save to a file
rep = profile(iris_df)
rep.save_html("iris_report.html")

Streaming report (low memory)

For large datasets, stream in-memory DataFrame chunks you control.

from pysuricata import profile, ReportConfig
import pandas as pd

def chunk_iter():
    for i in range(10):
        yield pd.read_csv(f"part-{i}.csv")  # You manage chunking externally

rep = profile((ch for ch in chunk_iter()), config=ReportConfig())
rep.save_html("report.html")

# Optional: stats-only
from pysuricata import summarize
stats = summarize(iris_df)

Highlights:

• Streams data in chunks, low peak memory.
• Shows processed bytes (≈) and precise generation time (e.g., 0.02s).
• Approximate distinct (KMV), heavy hitters (Misra–Gries), quantiles/histograms via reservoir sampling.
• Numeric extras: 95% CI for mean, coefficient of variation, heaping %, granularity hints, bimodality.
• Categorical extras: case/trim variants, empty strings, length stats.
• Datetime details: per-hour, day-of-week, and month breakdown tables + timeline chart.
• Correlation chips (streaming) for numeric columns.
• Hardened HTML escaping for column names and labels.