Add example

[erwanp]

Could you add an example of use in the Readme or in the Docs https://numba-scipy.readthedocs.io/en/latest/user/index.html ,
the user-manual is currently empty

[esc]

@erwanp thank you for submitting this. I have labeled it as a feature request. Pull-requests always welcome! 😇

[clarkedavida]

Please even a very simple example would be helpful! Can you maybe give a hint how to use this module? Does it come automatically with import numba? The following code doesn't work, and I can't tell why.

from numba import jit
import scipy.special as sp


@jit(nopython=True)
def integral(a,b):
    sp.beta(a,b)

[esc]

Please even a very simple example would be helpful! Can you maybe give a hint how to use this module? Does it come automatically with import numba? The following code doesn't work, and I can't tell why.

from numba import jit
import scipy.special as sp


@jit(nopython=True)
def integral(a,b):
    sp.beta(a,b)

What error message do you get?

[clarkedavida]

>>> from numba import jit
>>> import scipy.special as sp
>>> @jit(nopython=True)
... def integral(a,b):
...     sp.beta(a,b)
... 
>>> integral(1,2)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/david/.local/lib/python3.11/site-packages/numba/core/dispatcher.py", line 468, in _compile_for_args
    error_rewrite(e, 'typing')
  File "/home/david/.local/lib/python3.11/site-packages/numba/core/dispatcher.py", line 409, in error_rewrite
    raise e.with_traceback(None)
numba.core.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Unknown attribute 'beta' of type Module(<module 'scipy.special' from '/usr/lib/python3.11/site-packages/scipy/special/__init__.py'>)

File "<stdin>", line 3:
<source missing, REPL/exec in use?>

During: typing of get attribute at <stdin> (3)

File "<stdin>", line 3:
<source missing, REPL/exec in use?>

[esc]

So, I tried to reproduce this and found two things. The error you are seeing is because numba-scipy is either not installed or not installed correctly. After pip install numba-scipy I get:

Traceback (most recent call last):
  File "/Users/esc/git/numba/issue_numba_scipy_74.py", line 9, in <module>
    integral(1, 2)
  File "/Users/esc/git/numba/numba/core/dispatcher.py", line 468, in _compile_for_args
    error_rewrite(e, 'typing')
  File "/Users/esc/git/numba/numba/core/dispatcher.py", line 409, in error_rewrite
    raise e.with_traceback(None)
numba.core.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend)
No implementation of function Function(<ufunc 'beta'>) found for signature:

 >>> beta(int64, int64)

There are 2 candidate implementations:
  - Of which 2 did not match due to:
  Overload of function 'beta': File: ../../miniconda3-arm64/envs/numba_3.10/lib/python3.10/site-packages/numba_scipy/special/overloads.py: Line 9.
    With argument(s): '(int64, int64)':
   No match.

During: resolving callee type: Function(<ufunc 'beta'>)
During: typing of call at /Users/esc/git/numba/issue_numba_scipy_74.py (7)


File "issue_numba_scipy_74.py", line 7:
def integral(a,b):
    sp.beta(a,b)

Which may be a bug in numba-scipy.

[clarkedavida]

OK you're correct, I somehow lost my numba-scipy installation... That's something I'll have to fix on my own.

Anyway do you think you could maybe give 1-2 short examples where numba-scipy works? I was originally interested in this to try to speed up double integrals of the form

scipy.integrate.quad( scipy.integrate.quad ( func, ... ) )

My strategy so far has been to Numba as much of func as I can, but for my use-case there are a couple instances where it would be convenient to just numba-scipy a scipy.integrate.

Thanks for taking some time to look into this, I appreciate it.

EDIT: Wait I'm just now noticing... Does numba-scipy actually work with the integrate module?

[esc]

OK you're correct, I somehow lost my numba-scipy installation... That's something I'll have to fix on my own.

Anyway do you think you could maybe give 1-2 short examples where numba-scipy works? I was originally interested in this to try to speed up double integrals of the form

scipy.integrate.quad( scipy.integrate.quad ( func, ... ) )

My strategy so far has been to Numba as much of func as I can, but for my use-case there are a couple instances where it would be convenient to just numba-scipy a scipy.integrate.

Thanks for taking some time to look into this, I appreciate it.

EDIT: Wait I'm just now noticing... Does numba-scipy actually work with the integrate module?

No, I don't think so. The currently supported functions are listed on: https://numba-scipy.readthedocs.io/en/latest/reference/index.html -- this is the scope of the scipy functions that can be used within a Numba @njit decorated context.