Conflicting types - Unable to transform a ndarray into a PyArray

[Matti88]

Hello team. While working with the library I wanted to pass out a PyArray from a function but I am unable to convert a ndarray into a PyArray the following is the error code:

mismatched types

expected struct `ndarray::ArrayBase`, found a different struct `ndarray::ArrayBase`

note: expected struct `ndarray::ArrayBase<ndarray::data_repr::OwnedRepr<_>, _>`
         found struct `ndarray::ArrayBase<ndarray::OwnedRepr<i32>, ndarray::Dim<[usize; 2]>>`
note: perhaps two different versions of crate `ndarray` are being used?rustc(E0308)
lib.rs(76, 58): expected struct `ndarray::ArrayBase`, found a different struct `ndarray::ArrayBase`

yielded from the simple code

let mut return_arr  = Array2::<i32>::zeros((10, 10));
let return_Second_ = PyArray::from_owned_array(_py , return_arr);

I was not able to find in ndarray documentation the type:
ndarray::ArrayBase<ndarray::data_repr::OwnedRepr<_>, _>

[kngwyu]

I guess this is because of version mismatch.
What is in your Cargo.toml?

[Matti88]

Hello @kngwyu ,

here is what is in my cargo:

[lib]
name = "algo_match"
crate-type = ["cdylib"]

[dependencies.pyo3]
version = "0.12.4"
features = ["extension-module"]


[dependencies]
polars = "0.9.0"
numpy = "0.12"
ndarray = { version = "0.13", features = ["rayon"] }
num = "0.3.1"

[Matti88]

I guess this is because of version mismatch.
What is in your Cargo.toml?

I guess you would like to see the code for replication?

[relf]

I guess I have the same kind of problem.

error[E0308]: mismatched types
  --> src\lib.rs:18:32
   |
18 |             lhs: Some(LHS::new(&xlimits.as_array().to_owned()).seed(seed)),
   |                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected struct `ndarray::ArrayBase`, found a different struct `ndarray::ArrayBase`
   |
   = note: expected reference `&ndarray::ArrayBase<_, ndarray::dimension::dim::Dim<[usize; 2]>>`
              found reference `&ndarray::ArrayBase, ndarray::dimension::dim::Dim<[usize; 2]>>`
   = note: perhaps two different versions of crate `ndarray` are being used?

error[E0599]: no method named `into_pyarray` found for struct `ndarray::ArrayBase, ndarray::dimension::dim::Dim<[usize; 2]>>` in the current scope
  --> src\lib.rs:28:42
   |
28 |             .map(|m| m.sample(n_samples).into_pyarray(gil.python()).to_owned());
   |                                          ^^^^^^^^^^^^ method not found in `ndarray::ArrayBase, ndarray::dimension::dim::Dim<[usize; 2]>>`

Indeed it seems to be related to ndarray 0.13 vs 0.14, as ndarray-linalg is still with ndarray=0.13 even if I specify 0.13 in my Cargo.toml, it pulls 0.13 and 0.14. I do not understand why 0.14 is installed.

@kngwyu, I've seen the linalg example which works with ndarray=0.14 on a special ndarray-linalg branch of yours.
So I tried to build it with ndarray="0.13" and ndarray-linalg = "0.12" and I get:

(base) D:\rlafage\workspace\rust-numpy\examples\linalg>cargo build
   Compiling autocfg v1.0.1
   Compiling proc-macro2 v1.0.24
   Compiling unicode-xid v0.2.1
   Compiling winapi-x86_64-pc-windows-gnu v0.4.0
   Compiling syn v1.0.54
   Compiling winapi v0.3.9
   Compiling libc v0.2.81
   Compiling proc-macro-hack v0.5.19
   Compiling serde v1.0.118
   Compiling rand_core v0.4.2
   Compiling rawpointer v0.2.1
   Compiling cfg-if v1.0.0
   Compiling inventory v0.1.10
   Compiling smallvec v1.5.1
   Compiling scopeguard v1.1.0
   Compiling unindent v0.1.7
   Compiling pyo3 v0.12.4
   Compiling ndarray v0.13.1
   Compiling ndarray v0.14.0
   Compiling blas-src v0.2.1
   Compiling lapack-src v0.6.0
   Compiling blas-src v0.6.1
   Compiling cfg-if v0.1.10
   Compiling num-traits v0.2.14
   Compiling num-complex v0.2.4
   Compiling num-integer v0.1.44
   Compiling matrixmultiply v0.2.4
   Compiling instant v0.1.9
   Compiling rand_core v0.3.1
   Compiling lock_api v0.4.2
   Compiling cblas-sys v0.1.4
   Compiling lapacke-sys v0.1.4
   Compiling quote v1.0.7
   Compiling num-complex v0.3.1
   Compiling approx v0.3.2
   Compiling paste-impl v0.1.18
   Compiling rand v0.5.6
   Compiling parking_lot_core v0.8.1
   Compiling parking_lot v0.11.1
   Compiling paste v0.1.18
   Compiling lapacke v0.2.0
   Compiling cauchy v0.2.2
   Compiling pyo3-derive-backend v0.12.4
   Compiling ndarray-linalg v0.12.1
   Compiling inventory-impl v0.1.10
   Compiling indoc-impl v0.3.6
   Compiling ctor v0.1.16
   Compiling ghost v0.1.2
   Compiling pyo3cls v0.12.4
   Compiling indoc v0.3.6
   Compiling numpy v0.12.1 (D:\rlafage\workspace\rust-numpy)
   Compiling numpy-linalg-example v0.1.0 (D:\rlafage\workspace\rust-numpy\examples\linalg)
error[E0599]: no method named `inv` found for struct `ndarray::ArrayBase, ndarray::dimension::dim::Dim<[usize; 2]>>` in the current scope
  --> examples\linalg\src\lib.rs:12:14
   |
12 |             .inv()
   |              ^^^ method not found in `ndarray::ArrayBase, ndarray::dimension::dim::Dim<[usize; 2]>>`

warning: unused import: `ndarray_linalg::solve::Inverse`
 --> examples\linalg\src\lib.rs:1:5
  |
1 | use ndarray_linalg::solve::Inverse;
  |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  |
  = note: `#[warn(unused_imports)]` on by default

warning: unused import: `IntoPyArray`
 --> examples\linalg\src\lib.rs:2:13
  |
2 | use numpy::{IntoPyArray, PyArray2, PyReadonlyArray2};
  |             ^^^^^^^^^^^

error: aborting due to previous error; 2 warnings emitted

For more information about this error, try `rustc --explain E0599`.
error: could not compile `numpy-linalg-example`

To learn more, run the command again with --verbose.

(base) D:\rlafage\workspace\rust-numpy\examples\linalg>

It reproduces the ndarray 0.13 and 0.14 duplicate installation (and fails because of the mixture I suppose)
The modified Cargo.toml is:

[package]
name = "numpy-linalg-example"
version = "0.1.0"
authors = ["Yuji Kanagawa "]
edition = "2018"

[lib]
name = "rust_linalg"
crate-type = ["cdylib"]

[dependencies]
numpy = { path = "../.." }
ndarray = "0.13.1"
# ndarray-linalg = { git = "https://github.com/kngwyu/ndarray-linalg", branch = "ndarray-014", features = ["openblas-static"] }
ndarray-linalg = "0.12"

[dependencies.pyo3]
version = "0.12"
features = ["extension-module"]

So is there a way to make it work with ndarray-linalg=0.12 or should we wait for your pull request to be merged (I tried to use your ndarray-linalg branch but it ends up with several ndarray-linalg versions installation)?
In any case thank you for this project and your help.

[morenol]

I also had this problem. In my case, I also fixed this by upgrading ndarray to 0.14

smartcorelib/smartcore#39

[Matti88]

I also had this problem. In my case, I also fixed this by upgrading ndarray to 0.14

smartcorelib/smartcore#39

I have performed the same change and it worked just fine. I would consider it solved (somehow)

[relf]

Yes, if you can bump ndarray=0.14 it is all fine but with a ndarray-linalg dependency depending on ndarray=0.13 it is not possible. This topic was also discussed in #165 and from my standpoint, it is still an open issue. At the moment, I consider the example linalg is not really working as it relies on an unofficial ndarray-linalg version. Hopefully ndarray-linalg will catch up with rust-ndarray/ndarray-linalg/pull/258 🙏

[kngwyu]

Thank you @relf @morenol for your helps. I'm now closing this.

is there a way to make it work with ndarray-linalg=0.12

If you have a nightly toolchain, cargo +nightly update -Z minimal-versions would work.
But actually specifying ndarray >= 0.13 was my mistake. I should have used the more restricted version. Sorry about that.