Fix isstored with trailing singletons, vector show

[mtfishman]

This is a step towards addressing ITensor/BlockSparseArrays.jl#77.

To-do:

• Fix linear indexing in isstored.
• Add tests.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 75.59%. Comparing base (248dd87) to head (27294cc).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #45      +/-   ##
==========================================
+ Coverage   74.88%   75.59%   +0.70%     
==========================================
  Files           7        7              
  Lines         450      463      +13     
==========================================
+ Hits          337      350      +13     
  Misses        113      113              

Flag	Coverage Δ
docs	34.92% <25.00%> (-0.80%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[mtfishman]

@lkdvos the root of ITensor/BlockSparseArrays.jl#77 is that isstored wasn't handling trailing singleton indices properly, i.e.:

a = SparseArrayDOK{Float64}(undef, 2)
isstored(a, 1, 1)

should have output false but was outputting true, which is fixed in this PR. With that issue fixed, fixing ITensor/BlockSparseArrays.jl#77 is just a matter of generalizing Base.replace_in_print_matrix to accept matrices or vectors, which is demonstrated in this PR for sparse matrices/vectors and in ITensor/BlockSparseArrays.jl#79 for block sparse matrices/vectors.

You PR #39 refactors isstored, does that PR handle trailing singleton dimensions? I think that PR would be the correct place to address that issue since it involves a broader refactor of the isstored logic (though I do have some questions about which part of the logic around canonicalizing the index inputs should be handled at the level of isstored vs. @interface isstored, but we can discuss that in #39).

[lkdvos]

It should handle that correctly, since there are the calls to to_indices in there somewhere, but it's good that you brought this up so I can double check

[mtfishman]

It should handle that correctly, since there are the calls to to_indices in there somewhere, but it's good that you brought this up so I can double check

This issue led me to wonder how Base handles that logic, apparently it isn't in Base.to_indices:

julia> Base.to_indices(randn(2), (1, 1))
(1, 1)

But anyway, however it is done, let's handle that logic as part of #39 and this PR will be limited to changing the definition of Base.replace_in_print_matrix (or we can do that in #39 as well since it is a small change anyway).

[lkdvos]

Just double-checked it, in case you were interested, it's the combination of Base.to_indices and Base._to_linear_index or Base._to_subscript_indices that should resolve all combinations of indices. (which indeed is fixed in that PR)

Definitely agree on the rest though, if this works this is definitely a good intermediate patch.

[mtfishman]

Alight, as you say I guess this is a reasonable intermediate solution, so I'll go ahead and merge this with the assumption that the isstored logic will be rewritten in #39 anyway, but at the very least it is useful for ITensor/BlockSparseArrays.jl#77 and ITensor/BlockSparseArrays.jl#79.