Use DuckDB to query Parquet files, inspect metadata, join tables, and convert results to pandas for analysis; commonly precedes ydata-eda-profiling for EDA on extracted tables.
SKILL.md
DuckDB Parquet Lab Workflow
Purpose
Standardize the pattern of loading Parquet files into DuckDB, inspecting schema, running SQL joins, and converting results to pandas DataFrames.
Usage
"load Parquet with DuckDB and join tables"
"describe DuckDB table schema"
"convert DuckDB query to pandas"
Instructions
Read Parquet data with duckdb.query or duckdb.sql using SQL strings.
Inspect schema using DESCRIBE SELECT * FROM <table> and display with .show().
Use explicit joins with clear LEFT or RIGHT semantics to preserve row counts.
Convert results to pandas with .to_df() for downstream modeling.
Use ./templates/duckdb_snippets.md for the standard SQL patterns.
