Read Shape, Dtype, and Axis Meaning

An array stores values in a numerical layout. Before computing with it, inspect both the layout and what each direction means in the problem.

Read Shape, Dimensions, and Size

Use the same measurement table from the previous lessons:

shape: (3, 2)
dimensions: 2
elements: 6

These three attributes answer different questions:

The shape is a tuple because an array can have more than one axis. Multiplying the axis lengths gives the size: $3 \times 2 = 6$.

Do not use size as a substitute for shape. Arrays with shapes (3, 2) and (2, 3) both contain six values, but their layouts differ.

Give Every Axis a Meaning

Axis numbers describe storage directions. The program or dataset description must supply their meaning.

For this table, we choose:

The shape can therefore be read as

(observations, sensors) = (3, 2)

Axis 0 is the first position in the shape tuple, not “the horizontal axis” in every program. Axis 1 is the second position. Whether rows mean observations, people, images, or days depends on the data contract. Here each row is one observation and its two entries are the two sensor readings.

Writing the semantic names beside the shape prevents a common error: code can be valid Python and still operate along the wrong real-world direction. Later lessons will select and summarize values along these axes, so the names should be settled first.

Inspect and Convert the Dtype Deliberately

An array's dtype describes the representation used for all its elements:

print(readings.dtype)

For these decimal readings, NumPy normally creates a floating-point array. A specific environment may display a precision in the name, such as float64.

NumPy must choose one dtype for the complete array. If numerical input mixes integers and decimal values, it infers a type that can represent both:

[18.  20.5 22. ]
float64

The integer-looking values are represented in the same floating-point dtype as 20.5. The exact dtype name can depend on the input and environment, so inspect it rather than assuming.

Sometimes validated numerical values arrive as text:

This is a text array, not a numerical one. If the program has already established that every item is valid numerical text, astype(float) can make the required conversion explicit:

[18.5 20.  21.5]
float64

astype returns an array with the requested dtype. It is not a repair for unknown or malformed data: converting "missing" to float raises an error. Validate the boundary first, then convert for a stated numerical reason.

Later numerical work will rely on the dtype when it interprets a computation. Here the important habit is smaller: inspect the dtype, explain why it exists, and convert only when the program needs a different representation.

numeric_readings = text_readings.astype(float)