How to check inferred numba type with nopython true

Numba Support: How do I do ...?
1

i have the question same as :
How to check inferred numba type with nopython true
assume that we have two njit functions, one function will call another. and then use the first’s return result to create a numba type

@nb.njit()
def fn_1():
    a = np.random.randint(1, 100, size=24).reshape(4, 6)
    b = np.empty((4, 6), 'f8')
    c = np.empty((2, 8), 'i8')
    return a, b, c
    pass


@nb.njit()
def fn_2():
    rs = fn_1()
    numba_type=get_numba_type_from(rs)??????????????????? how could i do this?
    # my_list = nb.typed.typedlist.List.empty_list(numba_type)
    # my_list = nb.typed.typedlist.List([rs])
    # my_list.append(rs)
    pass
2

hi @guzuomuse , the link seems wrong. The text says “how to check …” but the link is to a thread on testing Numba 0.53.

3

@luk-f-a thanks for inform that. modified. do you know how to get the inferred type with onpython mode as the question described? thank you

4

From you code example above, it seems that you don’t need typeof. Would this do what you want?

@njit()
def fn_1():
    a = np.random.randint(1, 100, size=24).reshape(4, 6)
    b = np.empty((4, 6), 'f8')
    c = np.empty((2, 8), 'i8')
    return a, b, c
    

@njit()
def fn_2():
    rs = fn_1()
    my_list = []
    my_list.append(rs)
    return my_list
5

@luk-f-a the code works. but tha’t not what i want! the only thing i want here is to get the type on-the-fly programmatically .

6

creating a numba type cannot be done inside a jitted function. Types are a compile-time object, and they can be created in non-jitted code. They cannot be created at runtime, inside the jitted function.

I’m assuming that getting the type on-the-fly is only a means to an end. If you explain what you are trying to achieve, I might be able to suggest something that numba can do. Have you looked at the documentation of generated_jit?

Luk

7

just for learning purpose.
let it be it ,thanks for your help.