Python Review · Shallow and Deep Copy

Jan 9, 2016 | Tech Software

Mutable and Immutable Types

Data	Copy Behavior
Immutable	Data cannot be modified after creation, modifications create new objects Shallow and deep copy all refer to the original object; no copy needed, modifications won't affect original data
Mutable	Data can be modified directly after creation; modifications do not create new objects The difference between shallow and deep copy becomes apparent with nested structures of mutable types; not much with single-layer mutable types

Nested Mutable Data

Feature	Shallow Copy	Deep Copy
Scope	Only copies the outer container, does not copy nested elements.	Recursively copies the outer container and all nested elements, including multi-level nested mutable types.
Memory	Nested mutable elements share the same memory address.	No memory sharing at all, the new object is completely independent from the original.
Impact	Modifying nested mutable elements will affect the original object.	Modifying either the new or the original object will not affect the other.
Implementation	`copy.copy()`, `[:]`, or `foo.copy()` if supported	`copy.deepcopy()`, the only built-in valid method.
Performance	Lower overhead.	Higher overhead due to recursive copying.

Custom Classes

in general, use copy.copy() and copy.deepcopy() directly
the difference only becomes apparent when dealing with nested mutable structures in custom classes, similar to built-in types
if necessary, use __copy__(self) and __deepcopy__(self, memo) for customized behavior

Examples

Immutable Types

Mutable Types

Nested Shallow Copy

Nested Deep Copy

Class by Module

Class by Dunders