Extending the Numba Frontend¶
The Numba APIs described in this document are not currently guaranteed to be stable. External packages that rely on these APIs may break with new Numba releases.
The frontend of Numba analyzes the control flow of a function and performs type inference in order to (attempt to) deduce the types of all intermediate values in the function and identify points where types must be coerced.
A Numba type is really a category label for values that is used by the back- end to match appropriate code generators with the values they operate on. All Numba types are instances of classes that inherit from numba.types.Type. Numba types can be parameterized (for example, arrays and records), in which case their Type classes will take constructor arguments defining the parameters. Different instances of a parameterized type usually denote distinct types and can trigger different, specialized code generation in the backend.
In the rest of this document, when we refer to a “type”, we mean the Numba type unless we explicitly write “Python type”.
Mapping Python Types to Numba Types¶
Although the @jit decorator allows explicit declarations of the Numba types in a function signature, sometimes Numba needs to infer the Numba type associated with a particular Python type. If automatic JIT compilation is being used, then Numba will determine the types of function arguments from the Python values passed as function arguments. Additionally, if a function accesses global variables, Numba types will also be inferred from the Python values of those globals.
Once the types of all the externally defined values (function arguments and globals) have been specified, the type inference engine needs to propagate these types through all of the expressions in the function.
Numba needs type signatures for:
- Object attributes: This can include the attributes of instances of Python classes, or modules.
- Global values: Objects (such as functions) accessed from the global namespace.
- Operators and other “implicit” functions: Certain Python syntax (like a + b, or iter(o)) triggers special function calls. To overload these operations, a type signature for the appropriate function must be registered.
- Other entities not described in this document, such as builtin functions.
All of the tasks below will work with an example class and function:
class Interval(object): '''A half-open interval on the real number line.''' def __init__(self, lo, hi): self.lo = lo self.hi = hi def __repr__(self): return 'Interval(%f, %f)' % (self.lo, self.hi) # global function def valid_interval(interval): '''Return True if interval.lo <= interval.hi''' pass # This is a stub. We will implement the function in LLVM
Organizing Type Signatures with a Registry¶
If you have a lot of type signatures in a module, it can be cumbersome to make type information easily portable between targets. The numba.typing.templates.Registry class simplifies this process by collecting lists of attribute, global and operator type signatures that can be installed into a typing context all at once.
A common pattern in the Numba code is to collect all the type information for a particular package into a module that begins with:
from numba.typing.templates import (AttributeTemplate, ConcreteTemplate, signature, Registry) registry = Registry() # A new registry for our new set of types builtin = registry.register builtin_attr = registry.register_attr builtin_global = registry.register_global
Then those three functions are used to record different type signatures in the registry (see examples below). When the registry is fully populated, it is installed in the typing context:
from numba.targets.registry import target_registry # Assuming the CPU target target = target_registry['cpu'] target.targetdescr.typing_context.install(registry)
Creating a New Numba Type¶
To create a new Numba type, subclass numba.types.Type and make a single instance of it:
class IntervalType(numba.types.Type): def __init__(self): super(IntervalType, self).__init__(name='Interval') interval_type = IntervalType()
interval_type can now be used to declare argument and return types in @jit decorations:
@jit(numba.types.bool_(interval_type, numba.types.float32)) def inside(interval, x): return interval.lo <= x < interval.hi
The string form of the JIT signature @jit("bool_(interval_type, float32)") cannot be used in the above example unless interval_type has been added to the numba.types module. This shortcoming will be fixed in a future Numba version.
Adding an Attribute Value Type Signature¶
We can add type signatures for attributes of instances of Interval, so that lo and hi are recognized as returning float32 types. This requires creating a subclass of numba.typing.templates.AttributeTemplate:
from numba.types import float32 from numba.typing.templates import AttributeTemplate @builtin_attr class IntervalAttributes(AttributeTemplate): key = interval_type # We will store the interval bounds as 32-bit floats _attributes = dict(lo=float32, hi=float32) def generic_resolve(self, value, attr): return self._attributes[attr]
The key attribute of the template contains the Numba type that needs to be matched to use this template. It can either be an instance of a Type subclass, or the subclass itself, for parametric types.
The AttributeTemplate will first look for a method of the form resolve_<attribute name> to get the type of a specific attribute, otherwise it will delegate to the generic_resolve() method. This call takes both the Numba type instance (useful for parametric types) of the value being accessed, and the name of the attribute. The return value from generic_resolve() is the type of the value returned by the attribute access.
Adding a Function Type Signature¶
In order for the Numba type inference engine to recognize the valid_interval global function, we need to provide a type signature for it. This is done using a numba.typing.templates.ConcreteTemplate:
from numba.types import bool_, Function from numba.targets.registry import target_registry from numba.typing.templates import ConcreteTemplate, signature # Assuming the CPU target target = target_registry['cpu'] typing_context = target.targetdescr.typing_context class ValidIntervalSignature(ConcreteTemplate): key = valid_interval cases = [ signature(bool_, interval_type) ] builtin_global(valid_interval, Function(ValidIntervalSignature))
The key for looking up the function type is the Python function itself, valid_interval in this example. The cases attribute lists all of the supported function signature combinations. The first argument to signature is the return type, and the remaining arguments are the types of the function arguments. Only positional arguments are supported for function types (i.e. no keyword arguments).
Overloading Elementary Operations¶
Suppose we want to add support for a + operation between two intervals. We need to make a ConcreteTemplate where the key is the string "+":
from numba.targets.registry import target_registry from numba.typing.templates import ConcreteTemplate, signature # Assuming the CPU target target = target_registry['cpu'] typing_context = target.targetdescr.typing_context @builtin class AdditionSignature(ConcreteTemplate): key = '+' cases = [ signature(interval_type, interval_type, interval_type) ]
Several templates with the same key can be inserted, and each will be checked for a matching function signatures in the order of insertion. This is what allows the same key to be overloaded with different numbers of arguments and different argument types.
The list of special function keys includes:
|+||Addition (2 args) and unary positive (1 arg)|
|-||Subtraction (2 args) and unary negative (1 arg)|
|/?||Divide (only Python 2)|
|getiter||Get an iterator (equivalent to __iter__())|
|iternext||Return the next element from an iterator (equivalent to __next__())|
|getitem||Get an item (equivalent to __getitem__())|
These keys come directly from operations in the Numba IR (see Stage 2: Generate the Numba IR).
In-place operations (like a += b) are assumed to have the same signature as the right-hand side of the expanded form (a = a + b).