Apart from the Language part below, which applies to both object mode and nopython mode, this page only lists the features supported in nopython mode.
Numba strives to support as much of the Python language as possible, but some language features are not available inside Numba-compiled functions. The following Python language features are not currently supported:
try .. except
, try .. finally
)with
statement)yield from
)The raise
statement is supported in several forms:
raise
(to re-raise the current exception)raise SomeException
raise SomeException(<arguments>)
: in nopython mode, constructor
arguments must be compile-time constantsSimilarly, the assert
statement is supported with or without an error
message.
Numba supports function calls using positional and named arguments, as well
as arguments with default values and *args
(note the argument for
*args
can only be a tuple, not a list). Explicit **kwargs
are
not supported.
Function calls to locally defined inner functions are supported as long as they can be fully inlined.
Functions can be passed as argument into another function. But, they cannot be returned. For example:
from numba import jit
@jit
def add1(x):
return x + 1
@jit
def bar(fn, x):
return fn(x)
@jit
def foo(x):
return bar(add1, x)
# Passing add1 within numba compiled code.
print(foo(1))
# Passing add1 into bar from interpreted code
print(bar(add1, 1))
Note
Numba does not handle function objects as real objects. Once a function is assigned to a variable, the variable cannot be re-assigned to a different function.
Numba now supports inner functions as long as they are non-recursive and only called locally, but not passed as argument or returned as result. The use of closure variables (variables defined in outer scopes) within an inner function is also supported.
Most recursive call patterns are supported. The only restriction is that the recursive callee must have a control-flow path that returns without recursing. Numba is able to type-infer recursive functions without specifying the function type signature (which is required in numba 0.28 and earlier). Recursive calls can even call into a different overload of the function.
Numba supports generator functions and is able to compile them in object mode and nopython mode. The returned generator can be used both from Numba-compiled code and from regular Python code.
Coroutine features of generators are not supported (i.e. the
generator.send()
, generator.throw()
, generator.close()
methods).
Arithmetic operations as well as truth values are supported.
The following attributes and methods are supported:
.conjugate()
.real
.imag
Arithmetic operations as well as truth values are supported.
The following attributes and methods are supported:
.conjugate()
.real
.imag
The following operations are supported:
Creating and returning lists from JIT-compiled functions is supported,
as well as all methods and operations. Lists must be strictly homogenous:
Numba will reject any list containing objects of different types, even if
the types are compatible (for example, [1, 2.5]
is rejected as it
contains a int
and a float
).
Note
When passing a list into a JIT-compiled function, any modifications made to the list will not be visible to the Python interpreter until the function returns.
Warning
List sorting currently uses a quicksort algorithm, which has different performance characterics than the algorithm used by Python.
Numba supports list comprehension, but not the creation of nested list.
Numba also supports “array comprehension” that is a list comprehension
followed immediately by a call to numpy.array()
. The following
is an example that produces a 2D Numpy array:
from numba import jit
import numpy as np
@jit(nopython=True)
def f(n):
return np.array([ [ x * y for x in range(n) ] for y in range(n) ])
In this case, Numba is able to optimize the program to allocate and initialize the result array directly without allocating intermediate list objects. Therefore, the nesting of list comprehension here is not a problem since a multi-dimensional array is being created here instead of a nested list.
Additionally, Numba supports parallel array comphension when combined with the parallel option on CPUs.
All methods and operations on sets are supported in JIT-compiled functions.
Sets must be strictly homogenous: Numba will reject any set containing
objects of different types, even if the types are compatible (for example,
{1, 2.5}
is rejected as it contains a int
and a float
).
Note
When passing a set into a JIT-compiled function, any modifications made to the set will not be visible to the Python interpreter until the function returns.
The bytearray
type and, on Python 3, the bytes
type
support indexing, iteration and retrieving the len().
The memoryview
type supports indexing, slicing, iteration,
retrieving the len(), and also the following attributes:
The following built-in functions are supported:
abs()
bool
complex
divmod()
enumerate()
float
int
: only the one-argument formiter()
: only the one-argument formlen()
min()
max()
next()
: only the one-argument formprint()
: only numbers and strings; no file
or sep
argumentrange
: semantics are similar to those of Python 3 even in Python 2:
a range object is returned instead of an array of values.round()
sorted()
: the key
argument is not supportedtype()
: only the one-argument form, and only on some types
(e.g. numbers and named tuples)zip()
array
¶Limited support for the array.array
type is provided through
the buffer protocol. Indexing, iteration and taking the len() is supported.
All type codes are supported except for "u"
.
collections
¶Named tuple classes, as returned by collections.namedtuple()
, are
supported in the same way regular tuples are supported. Attribute access
and named parameters in the constructor are also supported.
Creating a named tuple class inside Numba code is not supported; the class must be created at the global level.
ctypes
¶Numba is able to call ctypes-declared functions with the following argument and return types:
enum
¶Both enum.Enum
and enum.IntEnum
subclasses are supported.
math
¶The following functions from the math
module are supported:
math.acos()
math.acosh()
math.asin()
math.asinh()
math.atan()
math.atan2()
math.atanh()
math.ceil()
math.copysign()
math.cos()
math.cosh()
math.degrees()
math.erf()
math.erfc()
math.exp()
math.expm1()
math.fabs()
math.floor()
math.frexp()
math.gamma()
math.hypot()
math.isfinite()
math.isinf()
math.isnan()
math.ldexp()
math.lgamma()
math.log()
math.log10()
math.log1p()
math.pow()
math.radians()
math.sin()
math.sinh()
math.sqrt()
math.tan()
math.tanh()
math.trunc()
operator
¶The following functions from the operator
module are supported:
operator.add()
operator.and_()
operator.div()
(Python 2 only)operator.eq()
operator.floordiv()
operator.ge()
operator.gt()
operator.iadd()
operator.iand()
operator.idiv()
(Python 2 only)operator.ifloordiv()
operator.ilshift()
operator.imatmul()
(Python 3.5 and above)operator.imod()
operator.imul()
operator.invert()
operator.ior()
operator.ipow()
operator.irshift()
operator.isub()
operator.itruediv()
operator.ixor()
operator.le()
operator.lshift()
operator.lt()
operator.matmul()
(Python 3.5 and above)operator.mod()
operator.mul()
operator.ne()
operator.neg()
operator.not_()
operator.or_()
operator.pos()
operator.pow()
operator.rshift()
operator.sub()
operator.truediv()
operator.xor()
functools
¶The functools.reduce()
function is supported but the initializer
argument is required.
random
¶Numba supports top-level functions from the random
module, but does
not allow you to create individual Random instances. A Mersenne-Twister
generator is used, with a dedicated internal state. It is initialized at
startup with entropy drawn from the operating system.
random.betavariate()
random.expovariate()
random.gammavariate()
random.gauss()
random.getrandbits()
: number of bits must not be greater than 64random.lognormvariate()
random.normalvariate()
random.paretovariate()
random.randint()
random.random()
random.randrange()
random.seed()
: with an integer argument onlyrandom.shuffle()
: the sequence argument must be a one-dimension
Numpy array or buffer-providing object (such as a bytearray
or array.array
); the second (optional) argument is not supportedrandom.uniform()
random.triangular()
random.vonmisesvariate()
random.weibullvariate()
Note
Calling random.seed()
from non-Numba code (or from object mode
code) will seed the Python random generator, not the Numba random generator.
Note
Since version 0.28.0, the generator is thread-safe and fork-safe. Each thread and each process will produce independent streams of random numbers.
See also
Numba also supports most additional distributions from the Numpy random module.
cffi
¶Similarly to ctypes, Numba is able to call into cffi-declared external functions, using the following C types and any derived pointer types:
char
short
int
long
long long
unsigned char
unsigned short
unsigned int
unsigned long
unsigned long long
int8_t
uint8_t
int16_t
uint16_t
int32_t
uint32_t
int64_t
uint64_t
float
double
ssize_t
size_t
void
The from_buffer()
method of cffi.FFI
and CompiledFFI
objects is
supported for passing Numpy arrays and other buffer-like objects. Only
contiguous arguments are accepted. The argument to from_buffer()
is converted to a raw pointer of the appropriate C type (for example a
double *
for a float64
array).
Additional type mappings for the conversion from a buffer to the appropriate C type may be registered with Numba. This may include struct types, though it is only permitted to call functions that accept pointers to structs - passing a struct by value is unsupported. For registering a mapping, use:
numba.cffi_support.
register_type
(cffi_type, numba_type)¶Out-of-line cffi modules must be registered with Numba prior to the use of any of their functions from within Numba-compiled functions:
numba.cffi_support.
register_module
(mod)¶Register the cffi out-of-line module mod
with Numba.
Inline cffi modules require no registration.