Create intermediate Works with: Any AI

SQL to Pandas Translation Guide

Few shot Has example

Move queries between SQL and pandas correctly.

The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.

Translate between SQL and pandas for my queries. Table schemas: [TABLE_SCHEMAS]. Queries or pandas code to translate: [QUERIES_OR_CODE]. Direction: [SQL_TO_PANDAS_OR_REVERSE].

For each query give:
1. The translation with comments aligning each clause (SELECT, WHERE, GROUP BY, HAVING, ORDER BY, JOIN) to its pandas counterpart.
2. A tiny worked example with four invented rows and the result of both versions side by side.
3. Differences that change results: null handling in comparisons and joins, integer division, case sensitivity of strings, sort stability, ties in ranking, how DISTINCT and drop_duplicates differ, join row multiplication.
4. Performance advice: do filtering and aggregation in the database when data is large, and pull only needed columns.
5. A verification step: run both on a sample and compare with assert_frame_equal after sorting.
Finish with a cheat-sheet table of the 12 most common patterns in my own table and column names.

Note the database engine and version, since null ordering, collation, and integer division differ between engines and change results.

Fill in before use

Optional: type your details and the prompt above updates. Nothing is saved or sent.

Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.

SQL
SELECT region, SUM(sales) AS total FROM orders WHERE year = 2025 GROUP BY region HAVING SUM(sales) > 1000 ORDER BY total DESC;
Pandas
(orders[orders["year"] == 2025]
 .groupby("region", as_index=False)["sales"].sum()
 .rename(columns={"sales": "total"})
 .query("total > 1000")
 .sort_values("total", ascending=False))
Notes: WHERE becomes boolean indexing; HAVING filters after grouping; NULL handling differs (sum skips NaN).
Check: compare row counts and totals with the SQL result.
Join: SQL LEFT JOIN = merge(how="left") with validate.
Gotcha: SQL COUNT(column) ignores NULLs while len(df) counts every row; use df['col'].count() for the SQL behaviour.

• sql-pandas