Could the error message for missing imports be made clearer?

[rajuptvs]

When running the example below, Mojo raises an error that can be confusing at first glance. The root cause was a missing import, but the error message did not make that clear.

from testing import assert_equal
from gpu.host import DeviceContext
# ANCHOR: dot_product_layout_tensor
from gpu import thread_idx, block_idx, block_dim, barrier
from layout import Layout, LayoutTensor

alias TPB = 8
alias SIZE = 8
alias BLOCKS_PER_GRID = (1, 1)
alias THREADS_PER_BLOCK = (TPB, 1)
alias dtype = DType.float32
alias layout = Layout.row_major(SIZE)
alias out_layout = Layout.row_major(1)

fn dot_product[
    in_layout: Layout, out_layout: Layout
](
    output: LayoutTensor[mut=True, dtype, out_layout],
    a: LayoutTensor[mut=True, dtype, in_layout],
    b: LayoutTensor[mut=True, dtype, in_layout],
    size: UInt,
):
    # FILL ME IN (roughly 13 lines)
    shared = LayoutTensor[
        dtype,
        Layout.row_major(TPB),
        MutableAnyOrigin,
        address_space = AddressSpace.SHARED,
    ].stack_allocation()

/home/raju/personal/mojo-gpu-puzzles/problems/p12/p12_layout_tensor.mojo:31:37: error: 'AddressSpace' value has no attribute 'SHARED'
        address_space = AddressSpace.SHARED,
                        ~~~~~~~~~~~~^

Initially, I assumed it was an issue with the AddressSpace type or its attributes.
However, after checking other puzzles, I realized it was simply a missing import.
In Python, a similar situation would produce something like:

NameError: name 'AddressSpace' is not defined

This would immediately hint that the symbol wasn’t imported or defined.

Was wondering if this kind of error message could be improved in the future — or maybe this isn’t really a big deal and is just part of Mojo’s current design philosophy?

[ehsanmok]

Thanks! added to the problem.