Sacha Verweij
                                            Sacha Verweij
                                        
                                    > good generic constructs to specify the output container type Ref. #16740, work in that direction. Best!
> Which would Vector(a) do? For general iterable `a`, `Vector(a)` should treat `a` as a list starting with the first item? Best!
👍 for @abraunst's comments; optimal methods for hypersparse arrays often differ from those for sparse arrays. As an alternative to the keyword idea, we could extend `sum!` to accept a...
Hello Augustin! Out of curiosity, what is your application? The extreme aspect ratio matrix with which you benchmark (`A = sprand(10^8, 10^3, 10.0^(-5))`) is a pathological case for the existing...
> As a matter of principle, sparse operations should scale with nnz and never have to allocate the full vector / matrix. `` A brief technical/historical digression :). Many operations...
Good catch! Looks like either: (1) `sgwi` should be initialized to zero and the condition in the `while` changed to `sgwi < sgwimax`; or (2) `sgwi += 1` should move...
Also a good catch! There are a number of ways you can fix that issue without adding another sweep (`countnz`). For one, you could count the number of nonzero values...
I wholeheartedly sympathize. Could we postpone this discussion till after the release though? I have been meaning to write up some thoughts on the matter, but cannot spare the time...
> Yes, once we have getfield overloading, any changes here should be non-breaking. That upside of getfield overloading had not occurred to me. Fantastic! 😄
Linking https://github.com/JuliaLang/julia/issues/24910#issuecomment-349699740. Best!