145 add support for constructing from data without copying (ktensor)

[dmdunla]

Addresses the ktensor part of:

• Add support for constructing tensor/ktensor from data without copying #145

---

📚 Documentation preview 📚: https://pyttb--182.org.readthedocs.build/en/182/

[dmdunla]

The following methods currently return a copy make a copy of the ktensor:

• permute
• symmetrize

The following methods alter the underlying data members and do not make a copy:

• arrange
• fixsigns
• normalize

Is this what we want? Which should be in-place (no copy) and which should make copies?

[dmdunla]

@ntjohnson1 Please review changes to ktensor at this point. Once this has been reviewed, I'll move to tensor.

[ntjohnson1]

The following methods currently return a copy make a copy of the ktensor:

• permute
• symmetrize

The following methods alter the underlying data members and do not make a copy:

• arrange
• fixsigns
• normalize

Is this what we want? Which should be in-place (no copy) and which should make copies?

We originally decided to match MATLAB's in-place vs copy choice. We are consistent with the methods you listed.

Returns copy

• permute
• symmetrize

In-place

• arrange
• fixsigns
• normalize

[ntjohnson1]

Looks good to me. A few NITs you can address at your discretion. Always a fan of smaller PRs, my vote would be to merge this then open a second PR for tensor since the changes should be aligned in spirit but I doubt the implementation overlaps.

[dmdunla]

Will finalize this PR to address ktensor only in #145.

@ntjohnson1 Would you suggest breaking up #145 or just submitting several PRs for the individual classes?

[ntjohnson1]

Will finalize this PR to address ktensor only in #145.

@ntjohnson1 Would you suggest breaking up #145 or just submitting several PRs for the individual classes?

Whatever you think is most obvious. No strong preference. I mostly just wanted this to land today since I branched off it to start in on ktensor linting and typing here (finished linting and halfway through typing).