Skip to content

python: Add explicit index and bytes support #4388

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
zcbenz merged 2 commits into from
Aug 29, 2026
Merged

python: Add explicit index and bytes support #4388

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
34aaf8b
feat: Added explicit index and bytes
aaishwarymishra Aug 24, 2026
7df74b5
add explicit dim and type check in index
aaishwarymishra Aug 28, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 16 additions & 0 deletions python/src/array.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1034,6 +1034,22 @@ void init_array(nb::module_& m) {
[](mx::array& a) {
return nb::cast<std::complex<double>>(to_scalar(a));
})
.def(
"__index__",
[](mx::array& a) {
if (!mx::issubdtype(a.dtype(), mx::integer) || a.ndim() != 0) {
throw nb::type_error(
"Only 0-dimensional integer arrays can be converted to an index.");
Comment thread
zcbenz marked this conversation as resolved.
Resolved
}
return nb::int_(to_scalar(a));
})
.def(
"__bytes__",
[](mx::array& a) {
a.eval();
return nb::bytes(

@simeetnayan81 simeetnayan81 Sep 1, 2026 •
edited
Loading

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I was going through the code when I realized this:
We probably need to call a.contiguous here as well.
eval() only makes the buffer readable

If a refers to transpose/slice/view of some array b. byte(a) might be wrong silently.
I am still not 100% sure, but that's what I think.

Example:

x = mx.arange(10, int32)     # 0 1 2 3 4 5 6 7 8 9
y = x[::2]                # 0 2 4 6 8
bytes(y) #This would generate a false dump since y is not packed still. It'd take 0 1 2 3 4

I'll file a new issue with more detail

reinterpret_cast<const char*>(a.data<void>()), a.nbytes());
})
.def(
"__format__",
[](mx::array& a, nb::object format_spec) {
Expand Down
4 changes: 4 additions & 0 deletions python/src/indexing.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@

#include <nanobind/ndarray.h>

#include "mlx/dtype.h"
#include "mlx/ops.h"
#include "python/src/convert.h"
#include "python/src/indexing.h"
Expand All @@ -20,6 +21,9 @@ bool is_index_scalar(const nb::object& obj) {
if (nb::isinstance<nb::bool_>(obj)) {
return false;
}
if (nb::isinstance<mx::array>(obj)) {
return false;
}
if (!PyIndex_Check(obj.ptr())) {
return false;
}
Expand Down
24 changes: 24 additions & 0 deletions python/tests/test_array.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -140,6 +140,30 @@ def test_scalar_conversion(self):
self.assertListEqual(list(z.shape), list(x.shape))
self.assertListEqual(list(z.shape), list(y.shape))

def test_index_conversion(self):
for dtype in [
mx.uint8,
mx.uint16,
mx.uint32,
mx.uint64,
mx.int8,
mx.int16,
mx.int32,
mx.int64,
]:
with self.subTest(dtype=dtype):
self.assertEqual(operator.index(mx.array(2, dtype)), 2)
self.assertEqual(list(range(mx.array(3, dtype))), [0, 1, 2])

def test_index_conversion_invalid(self):
for dtype in [mx.float16, mx.float32, mx.bfloat16, mx.complex64, mx.bool_]:
with self.subTest(dtype=dtype):
with self.assertRaises(TypeError):
operator.index(mx.array(2, dtype))

with self.assertRaises(TypeError):
list(range(mx.array(3, dtype)))

def test_finfo(self):
with self.assertRaises(ValueError):
mx.finfo(mx.int32)
Expand Down
Loading