Migration Guide

Coming from pandas (Python), Polars, or SQL? Here is your cheat sheet.

From pandas (Python)

Operation	pandas (Python)	Molniya (TypeScript)
Read CSV	pd.read_csv("data.csv")	readCsv("data.csv")
Inspect	df.head()	df.head().print()
Select Col	df["price"]	df.col("price")
Filter	df[df["age"] > 18]	df.filter(r => r.age > 18)
Sort	df.sort_values("date")	df.sort("date")
New Column	df["total"] = df["a"] + df["b"]	df = df.assign("total", r => r.a + r.b)
GroupBy	df.groupby("dept").sum()	df.groupby("dept").sum("val")
Map	df["a"].map(fn)	df.col("a").map(fn)
Unique	df["a"].unique()	df.col("a").unique()

NOTE

Key Difference: pandas modifies dataframes in-place for many operations (or has inplace=True). Molniya is always immutable. You must assign the result: df = df.sort(...)

From SQL

Concept	SQL	Molniya
SELECT	SELECT name, age	.select('name', 'age')
WHERE	WHERE age > 18	.filter(r => r.age > 18)
ORDER BY	ORDER BY date DESC	.sort('date', false)
LIMIT	LIMIT 5	.head(5)
GROUP BY	GROUP BY dept	.groupby('dept')
HAVING	HAVING count > 10	.groupby(...).count().filter(...)
JOIN	LEFT JOIN other ON ...	.merge(other, { how: 'left', ... })

From Plain JavaScript (Arrays)

Operation	Array of Objects	Molniya DataFrame
Memory	Heavy (objects per row)	Efficient (typed arrays)
Filter	arr.filter(fn)	df.filter(fn)
Map	arr.map(fn)	df.assign(name, fn)
Sort	arr.sort((a,b) => a - b)	df.sort('col')
Stats	Manual (reduce loops)	.mean(), .sum(), .std()

Type Safety

One major upgrade from pandas is strict typing.

pandas: Columns can contain mixed types ("10", 10, None) silently.

Molniya: Columns are strictly typed. If a column is int32, you cannot put a string in it. This prevents an entire class of "cleaning" bugs.