I guess I'm not quite clear on whether ArrayViews, or ArrayViewsAPL would support synthetic broadcasting. IIUC, what's needed is to introduce a new dimension, with stride set to 0.
Tim Holy wrote: > That's kind of an "inverse slice." Neat. > > This should be easy to do with https://github.com/timholy/ArrayViewsAPL.jl, > and much easier to use than having to think about strides explicitly. We just > need to finish stagedfunctions. > > --Tim > > On Tuesday, August 26, 2014 08:45:38 AM Neal Becker wrote: >> I was thinking particularly of the case of a function that does not >> broadcast it's arguments. stride tricks could be used to cause >> broadcasting (maybe not that efficiently). >> >> http://stackoverflow.com/questions/25486506/julia-broadcasting-equivalent-of >> -numpy-newaxis?noredirect=1#comment39778672_25486506 >> Tobias Knopp wrote: >> > The ArrayView package will give similar though not equivalent >> > possibilities. See also https://github.com/JuliaLang/julia/pull/5556 >> > and https://github.com/JuliaLang/julia/issues/5932 >> > I think there is also a transpose type that would allow to reverse >> > strides. >> > >> > Cheers, >> > >> > Tobi >> > >> > Am Dienstag, 26. August 2014 14:26:46 UTC+2 schrieb Neal Becker: >> >> In numpy, array contents can be re-interpreted without copying. The >> >> indexing of >> >> an array is defined by it's strides, and by altering the strides we can >> >> get >> >> different views. >> >> >> >> >> >> http://nbviewer.ipython.org/github/ipython-books/cookbook-code/blob/maste >> >> r/notebooks/chapter04_optimization/06_stride_tricks.ipynb >> >> >> >> Does julia have a similar facility? >> >> >> >> -- >> >> -- Those who don't understand recursion are doomed to repeat it -- -- Those who don't understand recursion are doomed to repeat it
