itemsize

Itemsize is the number of bytes used to store a single element of a data type in an array or buffer. It reflects the memory requirement of one element and is a key factor in calculating the total memory footprint of a collection of elements, as well as in understanding how data is laid out in memory.

In Python, the array module exposes an itemsize attribute on array objects, reporting the size of one element as stored by the platform’s C type implementation. For example, an array of type 'i' (signed int) typically has an itemsize of 4 on common 32- and 64-bit platforms, though the exact value depends on the platform’s C type sizes.

In NumPy, the size of one element is given by the dtype's itemsize (or by the ndarray.itemsize property). If you create an array with dtype=np.float64, its itemsize is 8, since a 64-bit floating point number uses eight bytes.

To estimate memory usage, multiply the itemsize by the number of elements. Note that this gives the raw element data size; actual memory usage may differ due to object headers, alignment, or additional metadata in higher-level containers.

Common values correspond to standard primitive types: 1 for int8/uint8 or bool, 2 for int16/uint16, 4 for int32/float32, and 8 for int64/float64. The concept is related to, but distinct from, alignment/padding and to the broader notion of size in languages like C (sizeof) or Python’s struct module.

See also: sizeof, dtype.itemsize, ndarray.itemsize.