Scale, Product Array Combinator

[nullplay]

Do not merge

[codecov]

Codecov Report

Attention: 138 lines in your changes are missing coverage. Please review.

Comparison is base (df553b8) 76.00% compared to head (e2b69b7) 75.77%.

Files	Patch %	Lines
src/tensors/combinators/product.jl	0.00%	79 Missing ⚠️
src/tensors/combinators/scale.jl	42.68%	47 Missing ⚠️
src/transforms/wrapperize.jl	63.15%	7 Missing ⚠️
src/dimensions.jl	77.77%	4 Missing ⚠️
src/util/limits.jl	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #349      +/-   ##
==========================================
- Coverage   76.00%   75.77%   -0.23%     
==========================================
  Files          78       80       +2     
  Lines        6600     6796     +196     
==========================================
+ Hits         5016     5150     +134     
- Misses       1584     1646      +62     

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

[willow-ahrens]

what's the difference between scale and product? Also, does this work for discrete extent when the factor is a positive integer?

[nullplay]

1. Scale is dealing with constant coefficient and Product deals variable coefficient. Scale is similar to Offset combinator and Product is similar to Toeplitz.
2. No. We need some mechanism that restricts scaledim starts and ends at integer endpoint. Current scaledim just divide with a scale. (E.g., scaledim(Extent(3,6), 2) = Extent(1.5, 3.0), but what we want for discrete Extent is Extent(2, 3))

[willow-ahrens]

Can we reverse the direction of scaledim so that it multiplies the extent, rather than dividing? My intuition is that we should be able to write B[i] = A[i*2 + 1], right?