现在用julia
即便统计函数,julia足够了
python有的流行package
就是julia写的
版主: hci
hci 写了: 2024年 5月 5日 11:14 搞不定。
R在统计学中的地位,还不能被撼動,主要是统计学教授只用R,相关的学科,也是如此。比如心理系的研究生们,大都会R。
会R,python就也会了,反之不然。
个人而言,更喜欢R:主要是喜欢 bigtable 胜过 python 的 pandas。经常吐槽 pandas,太难用了。
说实在的,大公司就应该用 SAS,丑是丑一点,但非常适合不是统计专业的人使用,稳定而且提供的选项多。
pseudo 写了: 2024年 5月 5日 14:10 个人而言,更喜欢R:主要是喜欢 bigtable 胜过 python 的 pandas。经常吐槽 pandas,太难用了。
希望科学计算这块将来会统一到 Julia。对了,Julia 的内核本质也是 Lisp。
sqldf很慢,R里最快的是data.table,现在polars也支持R了。fieldman 写了: 2024年 5月 5日 14:57 有那么点意思。我个人也不喜欢学很多小众语言的syntax, 所以在R里用一些sqlite, sqldf之类的接口,无奈太慢了。duckdb速度如何?看来需要研究研究。
wildthing 写了: 2024年 5月 5日 11:19 R semantics is quite strange. Maybe due to historical reasons it is still popular. Its interface is quite pleasant and support for data operation is strong. On the other hand, performance is quite poor compared to Python. It may never go away. Python is stupid in its own right.
wildthing 写了: 2024年 5月 5日 11:19 R semantics is quite strange. Maybe due to historical reasons it is still popular. Its interface is quite pleasant and support for data operation is strong. On the other hand, performance is quite poor compared to Python. It may never go away. Python is stupid in its own right.
Just compare the dataframe of R and dataframe of Pandas. The performance difference is quite obvious.wokao 写了: 2024年 5月 5日 19:37 “performance is quite poor compared to Python”
胡说八道,python缺省是link 到openblas,miniconda连到mkl,R连到自带的简陋线性代数库,但是可以很方便就连到openblas和mkl,然后R和Python就没有性能差异了,毕竟Python已经很慢,R即便比蜗牛慢也慢不了多少。实际上R的popular库大多是用C/C++写的
wildthing 写了: 2024年 5月 5日 20:19 Just compare the dataframe of R and dataframe of Pandas. The performance difference is quite obvious.
I don't know why you have to use blas. If you have to use blas, you might as well use cublas, which is much better.
We are talking about performance of native code, not C libraries. The whole argument of R has good performance because the libraries are in C is complete bullshit. Python has numpy, which has excellent performance in most cases.
wokao 写了: 2024年 5月 6日 06:26 dataframe
https://h2oai.github.io/db-benchmark/
R的data.table比pandas快得多, dplyr和pandas速度相当,但是语法要舒服很多
你大概是文科生,上了个统计课,用了一把R,和里边教学用的dataframe,然后就开始胡说八道了
至于什么用blas,更是胡说八道。你用numpy实际是要调用blas的,只不过Python安装的时候偷偷往你机器里安装了OpenBlas/MKL。跟R一样,Python里边有速度要求的numpy和scipy都是C/C++写的
R和Python都是解释性语言,都慢。想快,用Julia。
你应该跑英文bbs去练英文,跑这里真滑稽。
wildthing 写了: 2024年 5月 6日 08:06 Python is compiled to a virtual machine language. It's not purely interpreted.
The original purpose of pandas is to improve the slow dataframe speed of R. You should read the paper by pandas author.
I study programming language for a living though I don't know much about R except it's pretty horrible.
Insulting my English doesn't make what you said true.