TopolearnEverything in Polars starts with a DataFrame of typed, named columns, so you need to read its shape and schema fluently.
A Polars DataFrame is a table of named columns, and each column is a Series: a typed, one-dimensional array stored in the Apache Arrow format. Every column has exactly one data type, such as Int64, Float64, String, Boolean or Date. The mapping of names to types is the schema.
The quickest way to build one is from a dict of lists: pl.DataFrame({"name": [...], "age": [...]}). Polars infers the types, or you can pass schema={"age": pl.Int32} to choose them.
Things to know straight away:
df.shape,df.height,df.widthgive the size;df.columnsthe names;df.schemathe names with types.df.head(n),df.tail(n)anddf.sample(n)show rows;df.glimpse()shows one line per column, handy for wide tables.df.describe()gives count, null count, mean, min, max and quantiles.df["age"]returns a Series;df.to_dicts()anddf.rows()give plain Python data.
Unlike pandas there is no index. Rows are identified by position or by the values in your columns, which makes joins and group-bys explicit and removes a whole class of alignment surprises. Polars also never mutates a frame in place: methods return a new DataFrame, and the underlying column buffers are shared rather than copied, so this is cheap.
Import it as import polars as pl, the universal convention.
Example
import polars as pl
df = pl.DataFrame(
{
"name": ["Ana", "Ben", "Cy"],
"age": [34, 28, None],
"city": ["Oslo", "Lima", "Oslo"],
},
schema_overrides={"age": pl.Int32},
)
print(df.shape) # (3, 3)
print(df.schema)
# Schema({'name': String, 'age': Int32, 'city': String})
print(df["age"].null_count()) # 1
print(df.head(2).rows()) # [('Ana', 34, 'Oslo'), ('Ben', 28, 'Lima')]
print(df.to_dicts()[2]) # {'name': 'Cy', 'age': None, 'city': 'Oslo'}Remember
- A DataFrame is a set of named, typed Series sharing one length
- Check
df.schemaearly: types drive everything else - There is no index; rows are found by their values
Watch out for
- Expecting
df.loc/ index alignment from pandas: Polars has neither.