It’s good that he recognizes and admits what I quoted below. This is more of an experiment than a serious attempt to convince people not to use R. Also, “Replacing all languages with Lisp” is not a new goal, so I would be a lot more argumentative without the disclaimer.
The title of this pamphlet may suggest that it is written against the R
programming language, but it is actually written against the majority of
programming languages in use today. The ultimate goal of this text is to
violently eliminate them all.
Recognizing that the aforementioned goal is rather quixotic, I will be
satisfied instead if anyone finds any value in this pamphlet, whether it be
education, entertainment or personal hygiene
I would recommend learning R because it passes the Perlis test:
A language that doesn’t affect the way you think about programming, is not worth knowing.
In fact you could treat R as a shortcut for a programmer to learn data manipulation and statistics. It’s probably going to sink in better than doing problems on paper in a statistics class.
Also, don’t underestimate that data manipulation is 90% of the problem. R implicitly “knows” this because it is used for real data analysis problems all the time, unlike Guile Scheme.
One thing to know about R is that it has actually been reinvented with Lisp-like macros in the form of the tidyverse, by Hadley Wickham:
That is, someone outside R core redesigned most of APIs and some of the language, and it’s actually better. It is based on the relational model, which computer scientists already know about:
I haven’t looked closely at R for quite while… but I imagine the core flaws and strengths are still there.
It’s core strength was that it was a free version of S that was good enough and close enough to attract a lot of academic statistician who encoded their knowledge as libraries for R.
Which is a sincere pity, since if they had made a slightly larger step life would be a lot better, but they didn’t.
Now a huge bundle of our civilizations knowledge of this subject is encoded in a language that would make a programming language academic weep.
(Actually it’s not Bad as languages go…. but then I have met some truly terrible ones.)
Like Latex and Perl, R was written with great attention to end user ergonomics, with a particular niche of end user in mind, but with not enough attention to having a good kernel underneath, oen that would shed cruft rather than accumulating it.
Since R came from S, the next language should be Q, and it should indeed be something Lisp-ish that provides a good front end for statisticians.
It’s good that he recognizes and admits what I quoted below. This is more of an experiment than a serious attempt to convince people not to use R. Also, “Replacing all languages with Lisp” is not a new goal, so I would be a lot more argumentative without the disclaimer.
I would recommend learning R because it passes the Perlis test:
A language that doesn’t affect the way you think about programming, is not worth knowing.
In fact you could treat R as a shortcut for a programmer to learn data manipulation and statistics. It’s probably going to sink in better than doing problems on paper in a statistics class.
Also, don’t underestimate that data manipulation is 90% of the problem. R implicitly “knows” this because it is used for real data analysis problems all the time, unlike Guile Scheme.
One thing to know about R is that it has actually been reinvented with Lisp-like macros in the form of the tidyverse, by Hadley Wickham:
https://www.tidyverse.org/
That is, someone outside R core redesigned most of APIs and some of the language, and it’s actually better. It is based on the relational model, which computer scientists already know about:
http://vita.had.co.nz/papers/tidy-data.html
In 2017, it makes sense to skip the messy, evolved base R APIs and use the tidyverse APIs instead.
I haven’t looked closely at R for quite while… but I imagine the core flaws and strengths are still there.
It’s core strength was that it was a free version of S that was good enough and close enough to attract a lot of academic statistician who encoded their knowledge as libraries for R.
Which is a sincere pity, since if they had made a slightly larger step life would be a lot better, but they didn’t.
Now a huge bundle of our civilizations knowledge of this subject is encoded in a language that would make a programming language academic weep.
(Actually it’s not Bad as languages go…. but then I have met some truly terrible ones.)
Like Latex and Perl, R was written with great attention to end user ergonomics, with a particular niche of end user in mind, but with not enough attention to having a good kernel underneath, oen that would shed cruft rather than accumulating it.
Since R came from S, the next language should be Q, and it should indeed be something Lisp-ish that provides a good front end for statisticians.