[breaking] scalar row indexing produces column vector

[ericphanson]

fixes #509

There could be other issues like this out there, but at least this fixes that particular one, and I can't quite think of others off the top of my head.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.82%. Comparing base (bb17c4c) to head (2cb537e).
Report is 2 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #624   +/-   ##
=======================================
  Coverage   97.82%   97.82%           
=======================================
  Files          88       88           
  Lines        5140     5141    +1     
=======================================
+ Hits         5028     5029    +1     
  Misses        112      112           

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

[odow]

The doc failure seems suspicious.

[odow]

This isn't quite correct. Here's a reproducer from the docs:

julia> using Convex

julia> import MathOptInterface as MOI

julia> x = Variable(2, 2)
Variable
size: (2, 2)
sign: real
vexity: affine
id: 885…241

julia> context = Convex.Context{Float64}(MOI.Utilities.Model{Float64});

julia> Y = [1 2; 3 4]
2×2 Matrix{Int64}:
 1  2
 3  4

julia> Convex.to_vaf(Convex.conic_form!(context, x[1, :] .* Y[1, :]'))
┌                              ┐
│0.0 + 1.0 MOI.VariableIndex(1)│
│0.0 + 1.0 MOI.VariableIndex(3)│
│0.0 + 2.0 MOI.VariableIndex(1)│
│0.0 + 2.0 MOI.VariableIndex(3)│
└                              ┘

julia> Convex.to_vaf(Convex.conic_form!(context, x[1:1, :] .* Y[1, :]'))
┌                              ┐
│0.0 + 1.0 MOI.VariableIndex(1)│
│0.0 + 2.0 MOI.VariableIndex(3)│
└                              ┘

These two calls should be the same.

[odow]

Okay, perhaps the docs are wrong, because:

julia> a = [1, 2]
2-element Vector{Int64}:
 1
 2

julia> a .* a'
2×2 Matrix{Int64}:
 1  2
 2  4

[odow]

Ooooofff. This is pretty subtle...

[ericphanson]

Good catch! Btw maybe this could be dot? It used to be bugged in the complex case so maybe the original author didn’t use it

[odow]

Updated to use dot. Yeah, I assume the previous code was just trial and error to find what worked.

[odow]

I think this is worth it. The next version might break existing code, but it's arguably a bug fix because the current code departs from regular Julia. There's a strong chance that any breakage is likely a bug to begin with.