Marek Kaluba

@BioTurboNick You can already use code from this branch directly (well, the only thing that doesn't work is censored samples) for an implementation based on Royston approximation of S-W test;...

@andreasnoack or anyone else in this thread ;) I'd need your advice. I have the SW-coeffs computed up to `eps(Float64)` precision for `n=3:189` (I could go further, but I wasn't...

ok, it seems that I was fitting Normal distribution wrong;) Box-Cox transform `W → (1-W)^λ/λ` for `n in 7:189` seems superior to shifted logarithmic, when optimizing log likelihood (currently using...

also https://github.com/JuliaCollections/LRUCache.jl/issues/18 the packages don't seem to be actively maintained? @kmsquire ?

@johnfgibson, @Stargateur: for comparison with `n`-factorial: we don't simply count the permutations, but sum the first entry of every permutation (otherwise future compilers may skip the `swap` part entirely); `p[n]...

@Stargateur, sure the exact value does not matter, it was just to explain the differences in the results. If I am not mistaken, You could start from `a[0] = 0`,...

a naive python version ```python def heaps(n): elts = [i+1 for i in range(n)] c = [0 for i in range(n)] n = 0 countfirst = elts[0] while n <...

defining `heaps_c(n) = Int(ccall(("heaps", Clib), Cuintmax_t , (Csize_t,), n))` I get: On julia-0.6.4: ```julia julia> @btime allperms(12) 2.862 s (3 allocations: 384 bytes) 3113510400 julia> @btime heaps_c(12) 1.470 s (0...

ok, so I figured it out: it's a matter of `redirect_std|out|err`: Just before `redirect_stdout(out)` You need to call `Base.Libc.flush_cstdio()` to flush all C-buffers;

Yeah, I found also this: https://github.com/JuliaLang/julia/issues/8765 I am not proficient enough to try to solve it...