Does Numba support in-built python function e.g. `setitem`

Question

TypingError: Failed in nopython mode pipeline (step: nopython frontend)
No implementation of function Function(<built-in function setitem>) found for signature:

This error is encountered when tried to set all the elements less than a given threshold in matrix the SG_Z to zero by using the following command SG_Z[SG_Z < Threshold] = 0

This command is being used in a function that is parallelized using Numba @jit. Due to the existence of this line, the function is not running.

Answer 1

You don't say much about SG_Z but I suspect it is 2d (or higher). numba has limited multidimensional indexing abilities (compared to numpy)

In [133]: arr = np.random.rand(3,4)
In [134]: arr
Out[134]: 
array([[0.8466427 , 0.37340328, 0.07712635, 0.34466743],
       [0.86591184, 0.32048868, 0.1260246 , 0.9811717 ],
       [0.28948191, 0.32099879, 0.54819722, 0.78863841]])
In [135]: arr<.5
Out[135]: 
array([[False,  True,  True,  True],
       [False,  True,  True, False],
       [ True,  True, False, False]])
In [136]: arr[arr<.5]
Out[136]: 
array([0.37340328, 0.07712635, 0.34466743, 0.32048868, 0.1260246 ,
       0.28948191, 0.32099879])

The numba:

In [137]: import numba
In [138]: @numba.njit
     ...: def foo(arr, thresh):
     ...:     arr[arr<.5]=0
     ...:     return arr
     ...: 
In [139]: foo(arr,.5)
Traceback (most recent call last):
  File "<ipython-input-139-33ea2fda1ea2>", line 1, in <module>
    foo(arr,.5)
  File "/usr/local/lib/python3.8/dist-packages/numba/core/dispatcher.py", line 420, in _compile_for_args
    error_rewrite(e, 'typing')
  File "/usr/local/lib/python3.8/dist-packages/numba/core/dispatcher.py", line 361, in error_rewrite
    raise e.with_traceback(None)
TypingError: No implementation of function Function(<built-in function setitem>) found for signature:
 
 >>> setitem(array(float64, 2d, C), array(bool, 2d, C), Literal[int](0))
 
There are 16 candidate implementations:
  - Of which 14 did not match due to:
  Overload of function 'setitem': File: <numerous>: Line N/A.
    With argument(s): '(array(float64, 2d, C), array(bool, 2d, C), int64)':
   No match.
  - Of which 2 did not match due to:
  Overload in function 'SetItemBuffer.generic': File: numba/core/typing/arraydecl.py: Line 171.
    With argument(s): '(array(float64, 2d, C), array(bool, 2d, C), int64)':
   Rejected as the implementation raised a specific error:
     TypeError: unsupported array index type array(bool, 2d, C) in [array(bool, 2d, C)]
  raised from /usr/local/lib/python3.8/dist-packages/numba/core/typing/arraydecl.py:68

During: typing of setitem at <ipython-input-138-6861f217f595> (3)

It's not setitem in general that missing; numba does this all the time. It's setitem for this particular combination of arguments.

It does work if I first ravel the array.

In [140]: foo(arr.ravel(),.5)
Out[140]: 
array([0.8466427 , 0.        , 0.        , 0.        , 0.86591184,
       0.        , 0.        , 0.9811717 , 0.        , 0.        ,
       0.54819722, 0.78863841])

But in numba we don't need to be afraid of iteration, so for a 2d input we could iterate on rows:

In [148]: @numba.njit
     ...: def foo(arr, thresh):
     ...:     for i in arr:
     ...:         i[i<thresh] = 0
     ...:     return arr
     ...: 
In [149]: foo(arr,.5)
Out[149]: 
array([[0.8466427 , 0.        , 0.        , 0.        ],
       [0.86591184, 0.        , 0.        , 0.9811717 ],
       [0.        , 0.        , 0.54819722, 0.78863841]])

There may be more general ways of writing this, and providing signatures, but this should give some idea(s) of how to address this issue.