Julia's still pretty new and under very heavy development. We all agree 
with your sentiment that distributing a set of commonly-useful packages 
along with the binary installation makes a lot of sense. I think there's an 
issue for that, it's on the roadmap as it were - it might be entirely 
doable within the next version or two of Julia, stay tuned.

Hopefully the bug you ran into that was causing trouble for Julia's package 
manager on Windows 8 is now resolved. Once you can use Pkg.add(), I think 
you'll come around to some of the things Julia has going for it over 
Matlab, despite the decades of head-start Matlab development had, and the 
enormous resources of thousands of Mathworks employees and millions in 
revenue behind it, and the difference in price :). Package manager speed 
and better discoverability and distributing common packages along with 
Julia binaries are acknowledged deficiencies that Julia needs work on. 
Solutions will come.


On Saturday, May 10, 2014 5:14:15 AM UTC-7, kdb wrote:
>
> @Isaiah Norton
>
> Sorry, when rewriting my post I forgot to include that I know the 
> bugreport. By now a patch has been committed, but I don't know if I'll have 
> the time to set up a build environment for testing it before the next 
> binary pre-release.
>
> @Tony Kelman
>
> Summing it up: For each of the points a solution exists but none of them 
> are treated as core part of the system and a new user would have to 
> *find*them and install them separately first -- main issue being the "find" 
> part, 
> especially once there are multiple tools that seem to fulfill the same 
> purpose.
>
> That's not something someone, who just needs to get data analysis done, 
> wants to bother with, so they'll likely stick with MATLAB. Having a default 
> setup of such tools prepackaged and prominently promoted through the Julia 
> home-page (e.g. as a separate download variant) would likely help getting 
> such users into the community. Even for me as a technically inclined user, 
> worrying about extra packages and tools from the get go is a big deterrent 
> to trying a new tool. When the typical end-user compares solutions, he will 
> likely compare the freshly installed state without any extra stuff and 
> there MATLAB would have a HUGE advantage.
>
>
> 2014-05-10 5:41 GMT+02:00 Isaiah Norton <[email protected] <javascript:>
> >:
>
>> The specific issue mentioned #4 is known: 
>> https://github.com/JuliaLang/julia/issues/5574
>> Unfortunately I can't fully reproduce it locally (no Windows 8.1) and the 
>> hint of a lead that I found hasn't really panned out (yet, but I think it's 
>> the right direction). If anyone (a) has this problem and (b) is willing to 
>> do a debugging session over IRC and/or TeamViewer, that would be really 
>> great.
>>  
>>
>> On Fri, May 9, 2014 at 11:32 PM, Tony Kelman <[email protected]<javascript:>
>> > wrote:
>>
>>> 1. There's Julia Studio by Forio, Julia-specific plugins for IPython, 
>>> Sublime Text, and Light Table, as well as a few native-Julia IDE's (using 
>>> Gtk or OpenGL or other toolkits) being worked on by various people.
>>>
>>> 2. https://github.com/malmaud/Autoreload.jl might be one solution for 
>>> this?
>>>
>>> 3. Don't think so, but I could be wrong. Discoverability and good 
>>> documentation across the package ecosystem is a tough problem, there are 
>>> some ideas and experiments floating around but no universal great solution 
>>> yet AFAIK.
>>>
>>> 4. I feel your pain regarding the brokenness of package management in 
>>> other languages on Windows. Yes, work is being done. A number of packages 
>>> are nicely set up to download any binary dependencies automatically on 
>>> Windows, I hope we can cover everything over time. It's a bit challenging 
>>> since most package developers don't use Windows very often if at all, but 
>>> there's usually no technical impediment to setting things up in Julia so 
>>> they Just Work (tm). Make noise and flag issues when things don't work on 
>>> Windows.
>>>
>>>
>>>
>>> On Friday, May 9, 2014 8:21:32 AM UTC-7, kdb wrote:
>>>>
>>>> Reading about julia, I came to quickly like the language. Everyday use 
>>>> however would be more about quick-and-dirty data analysis, so some 
>>>> questions surfaced regarding usability for non-programming usecases, 
>>>> specifically in MS Windows environments. Some of these expectations have 
>>>> been set by MATLAB, though I used it only shortly during a voluntary 
>>>> lecture roughly 7 years ago. 
>>>>
>>>> *1. A GUI for interactive use?*
>>>>
>>>> While the REPL that comes with Julia 0.3.0 works perfectly fine on 
>>>> Windows, MATLABs GUI makes it easier to just dive into the work. 
>>>> User-defined variables are shown, and can be manipulated with the mouse 
>>>> and 
>>>> it comes with a builtin editor for quick-and-dirty scripts, when the REPL 
>>>> alone doesn't cut it anymore.
>>>>
>>>> While the *help* and *apropos* functions are great, a GUI interface 
>>>> would help, mostly because it smoothes the learning curve for beginners.
>>>>
>>>> *2. Quick-and-dirty scripts?*
>>>>
>>>> In MATLAB, you can just create *.m* files in the working directory, 
>>>> which then are treated as function definitions that can be easily modified 
>>>> at any time and, as far as I've read, function definitions are updated 
>>>> automatically when the file is changed. Does Julia provide such facilities 
>>>> for interactive developement/data analysis? Note that already having to 
>>>> type something like *reload mymodule.jl* would already be a 
>>>> disadvantage. 
>>>>
>>>> *3. Auto-Import and global search?*
>>>>
>>>> Is it possible to just write *somemodule.somefunction(x)* rather than 
>>>> explicitly importing the module first? And does e.g. *apropos(string)* 
>>>> search 
>>>> globally all installed modules, maybe including user files? Both would 
>>>> help 
>>>> a new user in exploring the possibilities of the environment. 
>>>>
>>>> *4. Windows-compatible package management?*
>>>>
>>>> As a prime example, currently *Pkg.update(), Pkg.add(...)* are broken 
>>>> on Windows 8. For some reason, *git* is not able to send DNS when 
>>>> called from within Julia; The bug is being worked on, but it illustrates 
>>>> an 
>>>> issue I have found across the module management systems I tried, be it 
>>>> Perl 
>>>> (CPAN), or Python (easy_install, pip): They tend to not work reliably on 
>>>> Windows. Is work being done to make the prospects in that regard better 
>>>> for 
>>>> Julia?
>>>>
>>>>
>>>> I'm asking these questions not only out of personal interest, but also 
>>>> because I think that they might be important for convincing less-technical 
>>>> users to give Julia an honest try and thus to spread basic knowledge of 
>>>> the 
>>>> language. 
>>>>
>>>
>>
>

Reply via email to