None of those are particularly good reasons. It boils down to "Julia doesn't do Java-style OO well", but believe it or not Java-style OO isn't the only way you can build a large program.
To take a couple of specific points as examples, you can use composition <https://github.com/one-more-minute/LNR.jl/blob/22625e0f25ce299b7bab9b1fb20d0a8dd2266ecf/src/LNR.jl#L20-L23> as a replacement for inheritance just fine, and you can get great autocomplete <http://junolab.org/> for Julia as well. On 26 December 2014 at 21:41, Páll Haraldsson <pall.haralds...@gmail.com> wrote: > > I mentioned Julia as a good alternative to Ceylon language in their forum > (for linear algebra). > > First of all, I assume you could call Ceylon (in theory at least, or vice > versa, I don't know to much about Ceylon). And second, I'm not sure I agree > with the responses I got. It seems they (or I?) misunderstand Julia. Am I > wrong to think that Julia has all the important properties of "static > languages" and should really be considered as such? I'm still not sure > about (e.g.) this part "Julia does not have inheritance, so I could not > build complex hierarchies of classes". It seems to me Julia has type > hierarchies and I've not looked enough into multiple dispatch vs. > "conventional OO" to see if it has any big downsides. On its own or > interfacing with other OO languages, say C++. I do not care about every > conceivable OO feature such as C++'s multiple inheritance, just that you > have good abstraction capabilities (and can call other OO code in other > languages). > > https://groups.google.com/forum/#!topic/ceylon-users/Qs9m1SgdThI > > It seems to be that the module system is excellent (no worse than other > languages I know) and the exception handling. The data hiding part I'm > conflicted about. Are there any major trade-offs the designers regret > (related to large-scale) or would like to change (breaking code). > > Best regards, > Palli. > >