Sheehan Olver
Sheehan Olver
Put another way: if package B depends on A, then it’s natural that an extension for B depends on an extension on A
I don't understand your example. What's the alternative solution? It might be possible I can move functions/types into LazyArrays.jl in this case. But it seems like a major flaw in...
I still don’t understand the issue being solved. A directed tree doesn’t form cycles….
I think I need a mathematical description to be convinced it’s not an implementational issue. Packages with (or without) extensions are simply a tree. How does a cycle come up?...
Could a simple solution be that if an extensionA depends on extensionB it needs to list it in the Project.toml? I.e. https://github.com/JuliaArrays/LazyArrays.jl/blob/d3e51078551d98ba951525931c77b2f259bfea29/Project.toml#L25 Would become ``` LazyArraysBlockBandedMatricesExt = ["LazyArraysBandedMatricesExt", "BlockBandedMatrices"] ```
"my_" is a very bad naming convention. I have no idea what these are for!
I think even just beginning with`_` would be fine, eg., `LazyArrays._issparse`.
But we would need a comment explaining the design
If you bump the minor version (1.x) I can merge and tag
Note that `mul!` uses native BLAS, where `default_blasmul!` and `tiled_blasmul!` are Julia code, that were copied from LinearAlgebra. Here gives an indication of what's going on: ```julia julia> @time mul!(C,...