The only real change I see is how reference tracking is implemented and that’s good. They probably got inspired by http://www.pqr-project.org. I wish they would take more of those features on or all of them. Now that would qualify as a new major release.
To me that’s kind of upsetting that this is only happening now… That seems like a feature of an actual programming language, that should be included from version 1.0 and on. The fact it was deemed nonessential for so long speaks volumes on to what was prioritized. No offense to R or whatever, my entire dissertation was performed using it, it’s just not what maybe some people think that it is.
2 Likes
@anon92994695 yes, I have described RCall in Data Analysis with Julia using the example of a PCA and factor analysis. This works quite well, only if the return of R is too complex, it becomes “cumbersome” to use the desired return parameters in Julia. R is “old” and you can notice that from time to time (I started with R about 2006). And that …
Julia learned from the mistakes of both of these languages and corrected them from day 1.
… I expect from Julia, too. 
(But often I still feel like a beginner! 
)
1 Like
Oh I am a beginner. I think the beauty of it is, pretty much everyone is right now. Don’t get me wrong the people contributing to the core language aren’t. But, I feel like even they still scratch their head sometimes and think “I guess you can do that?”. I think that’s whats so great about it in the first place. It’s generic and flexible enough that people can actually “discover” things rather then “force” them into existance or memorize a million doc’s pages. Not always, but it’s the best mileage I’ve gotten from minimal investment with any language. That’s why I’m obsessed and posting on here all the time about often trivial/contrived things. I’m a resident cheerleader of sorts :P.
5 Likes
exactly. R seems very slow in terms of incorporating performance-enhancing features.
I talked to some statistics professors who exclusively use R for their teaching and research. They told me that R is the only language that is written by statisticians and for statisticians. In addition, they are not worried about the performance of R since Rcpp is very neat to use.
2 Likes
It is a pity that the maintainer of pqR, Radford Neal, is not a member in the R foundation.
2 Likes
I have heard this before, but I am not sure I understand what it means.
People sometimes say this to imply that R can be used by people with limited programming experience. Which is true for just coding scripts. But when it comes to writing performant R packages (which many statisticians do to demo new results), then it’s up C++ creek without a paddle.
5 Likes
To be fair, there is a paddle. It’s Rcpp, but then you realize there are some holes in the paddle. So you look at Python for a bit. There’s like three ways to do the same thing as Rcpp in Python which opens some nice possibilities. Then you need to do something more in line with classic statistics than machine learning and it would require tons of statistical packages in in C++ and connecting them to Python. After that you find Julia.
6 Likes
Perfect explanaition… Just perfect… Should be framed in my office somewhere…
edit - one day I hope the wiki reads something more like this
Julia language is an offshoot of the popular language called "Necessity". Although necessity is not programmatic, Julia is. This makes the language appealing to people who need to get things done. For other non-need based uses Julia is unpopular. Languages such as heaxagony, brainfuck, python and R have empirically been shown to meet these necessary minimal conditions[1,2,3,4,5,...100000]. Although, people struggling with getting things to work and work reliably, felt it was a good idea to look at the short comings of their means to meet their needs and create Julia.
Now I’m being arrogant… Time to get off here before I say more unemployable things - I still use python and R from time to time as last resorts. Usually because someone else is too scared to leave their comfortzone and needs me to look at something… From time to time I even troubleshoot handcoded VB6.0 routines for matrix multiplication for people. Usually while having existential crisis’, contemplating joining the priesthood, going on a spirit journey, investing in a ponzi scheme, anything to find meaning/feel alive, but by all means I do what’s asked…
5 Likes
Love the Australianism. Rcpp is highly regarded though. I used it to do a simple function. But not as nice as Julia. Julia just needs to advance more on the TTFP porblem and feel less lethargic to use.
2 Likes
Xiaodai that’s why we have a wizard like you here :).
2 Likes
Rcpp is great, but while it makes interfacing with C++ code much easier, it does not make C++ programming any nicer per se. Many statisticians code a C++ version of their algorithm, and then wrap with Rcpp to demo in R.
1 Like