Fix nested broadcast of AbstractBlockTuple

[ogauthe]

This PR allows for nested broadcast on a BlockedTuple (e.g blocklengths.(dual.(axes))).

[codecov]

Codecov Report

Attention: Patch coverage is 28.57143% with 5 lines in your changes missing coverage. Please review.

Project coverage is 91.18%. Comparing base (b876749) to head (92edeaa).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/blockedtuple.jl	28.57%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #14      +/-   ##
==========================================
- Coverage   92.54%   91.18%   -1.37%     
==========================================
  Files          16       16              
  Lines         322      329       +7     
==========================================
+ Hits          298      300       +2     
- Misses         24       29       +5     

Flag	Coverage Δ
docs	0.00% <0.00%> (ø)

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.

[mtfishman]

Looks like there is a missing getindex overload in Julia 1.10.

Interestingly, Tuple broadcasting works without having ndims(::Tuple) defined. If this is the easiest route to getting it working, I'm ok with defining ndims for now, though we should treat it as an "internal" definition only used by broadcasting.

[ogauthe]

I am puzzled by the test failure on Julia 1.10. As for ndims, I followed the compiler advice, it is not clear to me why it is needed. I am not planning to define it for the object itself.

[mtfishman]

I am puzzled by the test failure on Julia 1.10. As for ndims, I followed the compiler advice, it is not clear to me why it is needed. I am not planning to define it for the object itself.

I think what's happening is that in Julia 1.10 broadcasting made use of indexing with CartesianIndex. For AbstractArray subtypes, CartesianIndex indexing gets forwarded to indexing with integers, but that isn't the case right now for AbstractBlockTuple. You could try defining:

Base.getindex(bt::AbstractBlockTuple, i::CartesinaIndex{1}) = bt[Tuple(i)...]

and see if that fixes things.