Hi Tim,

Indeed, ChainedVectors is an excellent start. It implements the desired 
functionality but purely for vectors. What I am looking for could be called 
ChainedMatrices. 

Thanks for the tip.

Best Regards,
Jan

Dňa utorok, 17. marca 2015 11:49:55 UTC+1 Tim Holy napísal(-a):
>
> Not currently, but you can create your own new AbstractArray types. In 
> this 
> case, https://github.com/tanmaykm/ChainedVectors.jl 
> might be a good model for how to proceed. 
>
> Best, 
> --Tim 
>
> On Tuesday, March 17, 2015 01:35:03 AM Ján Dolinský wrote: 
> > Hi, 
> > 
> > Is it possible to create a view which refers to e.g. two different 
> matrices 
> > ? I am on Julia v0.36. 
> > 
> > X = rand(1000,1000) 
> > Y = rand(1000,100) 
> > 
> > @time a = view(X, :, :) 
> > elapsed time: 6.576e-6 seconds (184 bytes allocated) 
> > 
> > @time b = view(Y, :, :) 
> > elapsed time: 5.83e-6 seconds (168 bytes allocated) 
> > 
> > @time c = [a b] 
> > elapsed time: 0.080207052 seconds (8802560 bytes allocated) 
> > 
> > a and b are apparently "views" and little memory is allocated. Is it 
> > possible to create a composite view which refers to X and Y at the same 
> > time ? Expression "[a b]" seems to create a copy of X and Y. 
> > 
> > My motivation is that I am concatenating two matrices in a loop where 
> the 
> > first one is usually fixed and the other one may have different number 
> of 
> > columns at each iteration. 
> > 
> > Thanks, 
> > Jan 
>
>

Reply via email to