Why Are We Teaching Pandas Instead of SQL? | HackerNoon

Abstraction: Argument that SQL is superior to pandas for analytics education

Key points:

• SQL's declarative syntax forces thinking in column/table operations and prevents common beginner antipatterns (row-by-row loops, mixed types, CSV misuse) that pandas readily allows
• Pandas index design causes subtle bugs: the index can contain duplicates, nulls, and mixed types; loc vs iloc confusion is widespread among beginners
• Pandas object dtype conflates strings, mixed types, and Python objects—no static schema enforcement—while SQL columns have enforced types with constraints
• CSVs (the de facto pandas persistence format) lose schema on export; recommended alternatives are Parquet or a SQL database; pickling is also discouraged
• Boolean indexing in pandas requires repeating the dataframe variable and depends on index alignment with the filter Series, making code verbose and error-prone
• For analytics at scale, the author recommends SQL + data warehouses (e.g., BigQuery) or PySpark over pandas optimization hacks

Connections: Pandas · Sql · Data Analysis · Declarative Programming

Source: https://hackernoon.com/why-are-we-teaching-pandas-instead-of-sql