On Thursday, 5 March 2015 17:49:24 UTC, Stefan Karpinski wrote:
>
> On Thu, Mar 5, 2015 at 7:43 AM, Christoph Ortner <[email protected] 
> <javascript:>> wrote:
>
>> For this reason, while I am happy to talk about how nice Julia is, I
>>> will not try to convince people to switch to it. IMO the people who are
>>> potential switchers at this stage have already looked at Julia, and
>>> evangelizing more aggressively could be counterproductive at this stage.
>>
>>
>> I think this is really important. Personally, I am thrilled with Julia, 
>> because I write code that does not need any packages other than plotting 
>> and File I/O. But I really need the combination of rapid development 
>> (scripting, dynamic) and then being able to optimise certain passages, 
>> without ever having to switch to C.  But I would never recommend Julia 
>> (at this stage) to a "production user", only to people who might like to 
>> "play with it". 
>>
>
> I think this depends on how much pain the alternatives are causing. I 
> would never try to sell someone on Julia if they're perfectly happy with 
> Matlab or Python or R – if those are working for them, great! It's the 
> people who are desperately unhappy with what they currently use that might 
> really benefit – and those people do exist. There are people now using 
> Julia in production for whom the alternatives were to prototype in 
> Matlab/R/Python and then rewrite everything in C++ or Java for performance; 
> but the amount of development time and effort entailed in that process just 
> wasn't feasible. Even with the relative immaturity of the Julia ecosystem, 
> it is still more productive in some circumstances to be able to prototype 
> *and* deploy in the same language. It's a matter of picking your poison: 
> you can go with the established languages and have a lot of pain around the 
> performance vs. productivity tradeoff; or you can go with Julia and that 
> won't be an issue, but you'll have to deal with sometimes implementing 
> things that other language already have packages for and with packages 
> sometimes breaking when you upgrade them (the secret is don't upgrade 
> often). As long as that tradeoff is clear, I think it's ok to recommend 
> Julia, but one does have to set expectations honestly and not oversell it.
>

Maybe my statement was a bit too strong and also should have been  clear 
about "production user": people who need mature packages (and in fact here 
it depends very much on what they need as e.g. Julia has much of Numpy and 
SciPy in Base already) and a mature development environment. 
I like Juno and I think ESS is ok, but I would call neither "mature", e.g., 
missing debugger, profiling is not built-in, auto-complete is far from 
perfect and getting help text in the editor does not work consistently 
either. And finally breaking backward compatibility every few months? As I 
said above I am hugely enjoying Julia, but these are the things I warn 
people about.

(Btw, some of my friends and colleagues have even started mocking me about 
my enthusiasm for Julia - so it is not as if I am doing a bad 
advertising-job ;).)

Reply via email to