First steps
Type system reference
- Built-in types
- Type inference and type annotations
- Kinds of types
- Class basics
- Annotation issues at runtime
- Protocols and structural subtyping
- Dynamically typed code
- Type narrowing
- Duck type compatibility
- Stub files
- Generics
- More types
- Literal types and Enums
- TypedDict
- Final names, methods and classes
- Metaclasses
Configuring and running mypy
- Running mypy and managing imports
- The mypy command line
- The mypy configuration file
- Inline configuration
- Mypy daemon (mypy server)
- Using installed packages
- Extending and integrating mypy
- Automatic stub generation (stubgen)
- Automatic stub testing (stubtest)
Miscellaneous
- Common issues and solutions
- Supported Python features
- Error codes
- Error codes enabled by default
- Error codes for optional checks
- Additional features
- Frequently Asked Questions
- Mypy Release Notes
Project Links
Metaclasses¶
Ametaclass is a class that describesthe construction and behavior of other classes, similarly to how classesdescribe the construction and behavior of objects.The default metaclass istype, but it’s possible to use other metaclasses.Metaclasses allows one to create “a different kind of class”, such asEnums,NamedTuples and singletons.
Mypy has some special understanding ofABCMeta andEnumMeta.
Defining a metaclass¶
classM(type):passclassA(metaclass=M):pass
Metaclass usage example¶
Mypy supports the lookup of attributes in the metaclass:
fromtypingimportClassVar,TypeVarS=TypeVar("S")classM(type):count:ClassVar[int]=0defmake(cls:type[S])->S:M.count+=1returncls()classA(metaclass=M):passa:A=A.make()# make() is looked up at M; the result is an object of type Aprint(A.count)classB(A):passb:B=B.make()# metaclasses are inheritedprint(B.count+" objects were created")# Error: Unsupported operand types for + ("int" and "str")
Gotchas and limitations of metaclass support¶
Note that metaclasses pose some requirements on the inheritance structure,so it’s better not to combine metaclasses and class hierarchies:
classM1(type):passclassM2(type):passclassA1(metaclass=M1):passclassA2(metaclass=M2):passclassB1(A1,metaclass=M2):pass# Mypy Error: metaclass conflict# At runtime the above definition raises an exception# TypeError: metaclass conflict: the metaclass of a derived class must be a (non-strict) subclass of the metaclasses of all its basesclassB12(A1,A2):pass# Mypy Error: metaclass conflict# This can be solved via a common metaclass subtype:classCorrectMeta(M1,M2):passclassB2(A1,A2,metaclass=CorrectMeta):pass# OK, runtime is also OK
Mypy does not understand dynamically-computed metaclasses,such as
classA(metaclass=f()):...Mypy does not and cannot understand arbitrary metaclass code.
Mypy only recognizes subclasses of
typeas potential metaclasses.Selfis not allowed as annotation in metaclasses as perPEP 673.
For some builtin types, mypy may think their metaclass isabc.ABCMetaeven if it istype at runtime. In those cases, you can either:
use
abc.ABCMetainstead oftypeas thesuperclass of your metaclass if that works in your use-casemute the error with
#type:ignore[metaclass]
importabcasserttype(tuple)istype# metaclass of tuple is type at runtime# The problem:classM0(type):passclassA0(tuple,metaclass=M0):pass# Mypy Error: metaclass conflict# Option 1: use ABCMeta instead of typeclassM1(abc.ABCMeta):passclassA1(tuple,metaclass=M1):pass# Option 2: mute the errorclassM2(type):passclassA2(tuple,metaclass=M2):pass# type: ignore[metaclass]