Skip to main content

Series Data Structure

A Series is a one-dimensional array with an Index. Values have a dtype; labels identify and align them. That alignment behavior is the important difference from a plain Python list or NumPy array.

import pandas as pd

scores = pd.Series(
    [91, 84, pd.NA],
    index=["Ada", "Lin", "Sam"],
    name="score",
    dtype="Int64",
)

Mental model​

  • scores.array holds the data using a pandas extension array when applicable.
  • scores.index holds labels; labels need not be consecutive integers.
  • scores.name becomes a column label in many table operations.
  • scores.dtype controls representation and missing-value behavior.

Do not memorize inferred dtypes. They can depend on input and pandas version. Specify a dtype at boundaries when strings, nullable integers, booleans, or dates must have a particular representation.

Construction and alignment​

Construct from a sequence when order is primary, or from a mapping when labels are primary:

left = pd.Series({"a": 10, "b": 20})
right = pd.Series({"b": 1, "c": 2})

left + right
# a -> missing, b -> 21, c -> missing

left.add(right, fill_value=0)
# a -> 10, b -> 21, c -> 2

Arithmetic aligns by label, not by physical position. This is powerful but can silently introduce missing values when indexes differ. Compare indexes or use explicit alignment when the relationship is important.

Missing values​

Use isna() and notna() rather than equality comparisons. Depending on the dtype, missingness may be represented by pd.NA, NaN, or NaT; code should normally rely on the shared missing-data API rather than the scalar sentinel.

Boundary​

Use a NumPy array when positional homogeneous computation is the whole job. Use a Series when labels, nullable dtypes, alignment, or pandas table operations carry meaning.

Source​