Automated data profiling that treats df.describe() as a warmup

What it does fg-data-profiling generates extended exploratory data analysis reports from pandas or Spark DataFrames with minimal code. It infers column types, flags data quality issues like missing values and skewness, and packages everything into shareable HTML or machine-readable JSON. The output covers univariate and multivariate statistics, time-series autocorrelation, text character distributions, and even file metadata like EXIF tags.

The interesting bit The project is essentially an opinionated wrapper around the parts of EDA everyone rewrites: instead of chaining df.info(), df.describe(), and a half-dozen matplotlib calls, you get a single ProfileReport object that renders itself as a Jupyter widget, an iframe, or a flat file. The recent Spark support suggests it is trying to escape the single-machine sandbox, though the README admits that effort still needs contributors.

Key highlights

• Automatic type inference for categorical, numerical, date, and other column types.
• Built-in alerts for data quality issues including high correlation, skewness, uniformity, and constant values.
• Time-series analysis with autocorrelation, seasonality, ACF, and PACF plots.
• Text profiling that breaks down scripts such as Latin and Cyrillic along with separator usage.
• Dataset comparison mode for diffing two DataFrames in one shot.

Caveats

• The project was recently renamed from ydata-profiling to fg-data-profiling, so existing codebases need import updates.
• Spark support is marked as released but still actively seeking contributors, suggesting it may not be as polished as the pandas path.
• The README nudges users toward the commercial YData Fabric Data Catalog for database integration, so deep enterprise connectivity lives behind a paywall.

Verdict Data scientists and analysts who routinely inspect messy DataFrames will save time here; if you already have a bespoke EDA template you love, this will feel like overkill.