Update #4
Update #4
Conversation
|
I have no vote here but I'd say don't mention Graal. It is outside the scope of the document. It is not even mentioned in this fairly current and extensive list on the SO info tag which already has a number of somewhat semi-random extra links. I'd also recommend against quoting the exact number of CRAN packages as this changes by the hour. As such I recommend against this PR #3 and recommend, ahem, mine in #2. |
|
Just quote that there are more than 14000 packages, for example. It is also important to make it clear that there are several packages that are not in CRAN. The python community cites a lot of contributions on GitHub. The R community is also very active on GitHub. |
|
You didn't look at https://github.com/matloff/R-vs.-Python-for-Data-Science/pull/2/files did you? |
|
And I think there should be talk about new technologies being created and with which R will benefit. Authors with Hadley Wickham already cited this; http://adv-r.had.co.nz/Performance.html. The project's https://github.com/oracle/fastr could be oused that others might be interested. Numba in Python started like this. |
|
As I said above not in an overview piece of a few hundred words. Everybody has their favourite but remaining concise and succinct is an art. @matloff is pretty good at it and can do without our editorialising. But maybe you want to start a blog somewhere stressing what you find important, and are excited about? |
|
Do you have an aversion to anything that tries to accelerate R but other than Rcpp? I do not know why you do not think it's feasible. Calm. Rcpp will not die. Think. |
|
It's amazing that any comment, be it in GitHub or in stack overflow you position yourself in a rude way. This behavior is deplorable. You do not have to feel threatened when some big company tries to contribute to improving the performance of the R language. |
|
I did not mention Rcpp. I do not think it even belonged in the piece. I remain convinced, however, that
|
|
Dirk, their arguments are loose. For you the problem of R's performance should only be solved with Rcpp. You released the Rcpp project when it was in the beginning. You announced. Phrases such as "I am wasting my time with you. Good bye." are quite childish. ;) Be a little more humble like Hadley who is not afraid to quote jobs other than his. The Community thanks you! |
|
Rcpp is stable, well-integrated into R, and widely used. In contrast, alternate R implementations might be worth mentioning in a "horizon scanning" document, but not in a practical document on R and Python in data science. Yes, GraalVM and fastR and pqR and all that might just be ready for production in a few years, but nobody doing data science now needs to know that. We all know that faster things are coming in the future. If you [@prdm0] want to write a "what's on the far horizon for R and Python" document then that might be interesting reading. But its not this. |
|
I still prefer to follow Hadley's vision of sharing information and making projects public, etc. The example of the Rccp that when it was not stable was disclosed. Well, the owner of the repository is free to decide. However, it is very strange that Dirk always opposes anything that says to improve the performance of R. Hadley always seemed to me to be a person who communicates data science well. He's been talking about FastR for years. Besides, the reader is smart. Just let us know that this is a project that could yield great things in the future. |
@prdm0 you seem to be making many unfounded assumptions here and then projecting. Its kind of like "I think you thought something I consider thought crime, so I'm going to insult you for how you think about it!". That's not fair and turns you into the very monster your seek to admonish others from becoming. It leaves the other party with no real avenue to respectfully respond with dignity, except for exiting the conversation. As an impartial spectator, I'd suggest that you work on separating what seems to be extreme negative emotions toward those you disagree with, from the merits of the subject at hand. It will get you better results. In that spirit, I want to give you a heads up by giving you a Trigger Warning: In the following paragraph, I am also going to disagree with your pull request. While the GraalVM you mention is interesting, its my understanding Python can be used on GraalVM as well, which you don't mention in your pull request patch to add a GraalVM section. Therefore, categorizing this as an advantageous feature for R vs Python maybe premature and somewhat or entirely negated. One could compare the merits of both R and Python implementations on GraalVM and Python Numba, with benchmarks and more, along with all other competing implementations. But that would best be handled in a separate note which I'm sure would be so very interesting for everyone to read. I know I issued a formal Trigger Warning, but I'm new at this so I'm not sure if said warning completely absolves me from being responsible for your emotions. In this spirit, I want you to know that I understand how difficult it is when what seems like a good idea, is logically dismantled by others in public. It can be tough 😢 So I just want to say Thank You so much for participating in this discussion. If there were ever a book containing examples of bad pull requests for others to learn form, this would be an excellent submission, and so I want you to know that this PR and ensuing dialogue, has earned you the virtual participation trophy of the day 🏆 👏 👏 👏 🙇♂️ |
|
@JustinMShea, how much nonsense you write in a single text. I'm not guessing anything. Dirk always tries to refute and take the merit of everything that comes to "improving the performance of R". This is notorious. Comparison X vesus Y allows the relation of equality. That way, what you said does not make sense. I believe VS in the file does not mean stick with silly comparisons of R vs. Python (This would be very amateur). crime? : Do not try to be funny. You're not good at it! |
|
Oh Dear... It appears you don't understand even the basic subject and format of the OPs note, which explains so many things.It could be forgiven, if not for your rather unsavory disposition! Sadly, It seems a history of disagreements with one of the posters has triggered such deep emotional trauma, that you are unable to engage in rational conversation at this time without insult. I hope this state of affairs improves for you. In the mean time, I hope this daily affirmation from Stuart Smalley helps you feel less threatened and assuages your damaged ego. |
Tbh, @eddelbuettel just stays factual. If anyone is "positioning themselves in a rude way", it would be the person who seems to have a habit of lashing out at others posting genuine concerns. It's also pretty annoying to discover that the personality clash is used as fake arguments on a Github repo for crying out loud. This is not a Hadley vs. TheBadWorldOfBaddies issue. Keep those things for twitter please. That said, GraalVM is legit and interesting. And in no way related to "how R and Python compare" as it is about interoperability between languages. I also like to point out that GraalVM states on their own website :
https://www.graalvm.org/docs/why-graal/#for-ruby-r-or-python So I'd agree with @eddelbuettel that mentioning GraalVM in this post doesn't give any added value. |
It is important to mention that several R packages are kept outside of CRAN, in GitHub and Git Lab, for example.
It would be interesting to discuss about GraalVM, an important technology that has been created by Oracle and will resemble Python Numba. This can greatly increase the computing power of the R language, just as it is possible with Rcpp today.