Segfault when converting COO to Dense

[mtsokol]

Hi @willow-ahrens,

I'm trying to convert empty COO tensor to a dense one but I get a segfault. Should I write it differently or is it a bug?

using Finch

tns_coo = Tensor(SparseCOO{2}(Element(0), (3,3)))  # empty coo
# with Tensor(SparseCOO{2}(Element(0), (3,3)), zeros(3,3)) it works

Tensor(Dense(Dense(Element(0))), tns_coo)  # segfault instead of dense 3x3

[mtsokol]

It looks like ptr field in uninitialized one is:

1-element Vector{Int64}:
 1

and for initialized it's:

2-element Vector{Int64}:
 1
 1

[willow-ahrens]

That syntax isn't supposed to initialize a tensor. Instead, pass the dimensions to the Tensor function. It looks like the relevant docs in https://github.com/willow-ahrens/Finch.jl/blob/c9588e252e5fb50b0b217fa73b0fae522d7817f0/src/tensors/fibers.jl#L24-L31 somehow didn't make it into the website, so I made a PR to do that.

[mtsokol]

Ah, then I used incorrect semantics, thank you!