[WIP] Update dependencies

[StefanUlbrich]

Open issues:

• Some changes in the slice syntax (make.rs, evaluate.rs)
• sprs is less actively maintained. Sprs_ldl has not been updated on crate.io

[StefanUlbrich]

Temporary workaround by installing sprs and sprs_ldl directly from git

[StefanUlbrich]

Resolved compile errors related to ndarray. Remainder related to sprs. A few question related to explicit casting

[StefanUlbrich]

This cast seems necessary

[StefanUlbrich]

@espdev , do you have an idea why this explicit cast is necessary?

[StefanUlbrich]

The last issue blocking compilation: The Mul trait is not implemented for the CsMatBase I am not sure whether it can be implemented outside the sprs crate (I tried at the bottom of trait.rs -- commented out. @espdev , any ideas?

[StefanUlbrich]

Fixed all but one issue connected to sprs

[espdev]

Hi @StefanUlbrich,

Thank you for your work and participation in csaps-rs!

Do I understand correctly that sprs version 0.11 does not work with csaps (the last issue)?

[StefanUlbrich]

1. Until the linalg part is not published, we’ll need to install from GIT directly.
2. We can’t use the scalar product at two places but that can be worked around easily.

is an external package mandatory? we only have band matrices so maybe this can be exploited?

[StefanUlbrich]

Hi @StefanUlbrich,

Thank you for your work and participation in csaps-rs!

Do I understand correctly that sprs version 0.11 does not work with csaps (the last issue)?

This compiles at least

let a1 = qtwq.map(|el| s1 * *el );
let a2 = r.map(|el| *el * smooth);

[mulimoen]

@StefanUlbrich sprs-ldl has been released for the newest sprs

[StefanUlbrich]

@StefanUlbrich sprs-ldl has been released for the newest sprs

Thanks! Yes, there is some movement in sprs. I still have an compiler error that I could not figure out on my own. I doubt that this is resolved. Could we have a look together, @espdev (my rust experience is still limited)?

[espdev]

Sorry guys. Now there is not enough time for the project. I won't be able to participate for a couple of weeks.

[StefanUlbrich]

No worries @espdev . It seems that upgrading to sprs is blocked by sparsemat/sprs#278 ... using generic floats is an unresolved issue. It would not be a problem if we were dealing withf32 or f64 directly

[StefanUlbrich]

@espdev , I managed to find a way around the issue with generic floats without the need to fix anything sprs (see sparsemat/sprs#278 (comment)). The package builds (!) and the unit tests pass except for one (and this seems to be a dimension issue, see below). There is still tons of comments and cleanup left, but it seems the issue is close to be finished.

test test_evaluate_3d ... FAILED

failures:

---- test_evaluate_3d stdout ----
thread 'test_evaluate_3d' panicked at 'assertion failed: `(left == right)`
  left: `[[[1.0, 2.0, 3.0, 4.0],
  [1.0, 3.0, 5.0, 7.0]],

 [[2.0, 4.0, 6.0, 8.0],
  [3.0, 5.0, 7.0, 9.0]]], shape=[2, 2, 4], strides=[1, 2, 4], layout=Ff (0xa), const ndim=3`,
 right: `[[[1.0, 2.0, 3.0, 4.0],
  [2.0, 4.0, 6.0, 8.0]],

 [[1.0, 3.0, 5.0, 7.0],
  [3.0, 5.0, 7.0, 9.0]]], shape=[2, 2, 4], strides=[8, 4, 1], layout=Cc (0x5), const ndim=3`', tests/umv_evaluate.rs:64:5
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace


failures:
    test_evaluate_3d

test result: FAILED. 3 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

[espdev]

but it seems the issue is close to be finished

@StefanUlbrich great news!

[StefanUlbrich]

Any idea on why the one unit test might fail (i.e., did it run on main)?

[espdev]

@StefanUlbrich
Thank you! great work!
I've made some other changes also, but now all test fail. :)

[espdev]

Fixed tests except test_evaluate_3d.

Any idea on why the one unit test might fail (i.e., did it run on main)?

I have no idea why the test has failed. Array memory layout and strides do not match. This test has definitely worked before (with old ndarray/sprs versions).