sophiamyang / PythonRcomparison

Python and R package download comparison

Data source

R package download data

• CRAN package download logs
  • http://cran-logs.rstudio.com/
  • e.g., https://www.r-bloggers.com/finally-tracking-cran-packages-downloads/
• Download stats API in R:
  • https://cranlogs.r-pkg.org/
  • JSON API: https://github.com/r-hub/cranlogs.app
  • e.g., https://cranlogs.r-pkg.org/downloads/total/2014-01-01:2014-12-31/ggplot2
  • access through R package canlogs: https://github.com/r-hub/cranlogs
• Python package cranlogs
  • pip install cranlogs
  • https://cranlogs.readthedocs.io/en/latest/cran_downloads.html

Python package download data

• Pypi data:
  • https://packaging.python.org/guides/analyzing-pypi-package-downloads/
  • https://github.com/ofek/pypinfo#installation
• Conda data:
  • https://github.com/ContinuumIO/anaconda-package-data

Python & R package comparison

Purpose	Python package	Python function	R package	R function
Data import and export	pandas	read_csv, to_csv	readr	read.csv, write.csv
	pandas	read_excel	readxl	.xls and .xlsx
	pandas	read_json	jsonlite	json
	json	json.loads	jsonlite	json
	xml	xml.etree.ElementTree.parse	XML	xmlparse
Data processing	pandas	apply, map	purrr	map
	pandas	apply, map	plyr	apply functions, e.g., aaply
	pandas	loc, sort_values, aggregate, groupby, merge	dplyr, data.table	filter, arrange, select, mutate, summarise, groupby, join
	pandas	pivot, melt, stack, unstack, groupby	tidyr	gather, spead
	pandas	.	magrittr	%>%
	pandas, matplotlib		tidyverse	a set of packages that include ggplot2, dplyr, tidyr, readr, purrr, tibble, stringr, and forcats
	datetime		lubridate, hms
Models
Linear models	sklearn.linear_model, statsmodels		stats	lm()
Generalised linear models	statsmodels	GLM	stats	glm()
Generalised additive models	pyGAM, stasmodels.gam		mgcv	gam()
Penalised linear models	sklearn.linear_model	Lasso, ElasticNet	glmnet	glmnet()
Penalised linear models	glmnet		glmnet	glmnet()
Robust linear models	statsmodels	RLM	MASS	rlm()
Linear Mixed-Effects Models	stasmodels	mixedlm	lme4, lmerTest
Structural eqation modeling	semopy		lavaan
k-means	sklearn.cluster	KMeans	stats	kmeans
PCA	sklearn.decomposition	PCA	stats	prcomp
t-SNE	sklearn.manifold	TSNE	Rtsne	Rtsne
KNN	sklearn.neighbors	KNeighborsClassifier	class,FNN	KNN
Decision Trees	sklearn	tree	rpart	rpart()
Random forest	sklearn.ensemble	RandomForestClassifier	randomForest	randomForest()
Gradient boosting	xgboost		xgboost
other	scipy, numpy, sklearn		e1071	collection of models (Functions for latent class analysis, short time Fourier transform, fuzzy clustering, support vector machines, shortest path computation, bagged clustering, naive Bayes classifier,...)
other	sklearn		caret	caret is a set of functions that attempt to streamline the process for creating predictive models
network analysis	networkx		igraph
Deep learning	keras, tensorflow		keras, tensorflow
Visualization	matplotlib, bokeh		ggplot2, graphics
Dashboard	panel, voila		shiny

Data

I retrieved package download data from Conda, PyPI, and Cran from Janurary 2018 to July 2019. Then I combined Conda and PyPI downloads to represent total downloads of Python packages.

Results

Pandas was downloaded a lot more than dplyr, tidyverse and data.table. And the growth of pandas is substantial over time.

scikit-learn is downloaded a lot more than statsmodels in Python, caret in R, and e1071 in R.

Most people use Keras and Tensorflow in Python, few use them in R.

People use matplotlib a lot. ggplot2 is as popular as Bokeh.

For creating dashboard, most people use Shiny. Few use Panel and Voila.

Conclusion

Except for creating dashboard, people seem to download a lot more Python packages than R packages for data manipulation, visualization, machine learning, and deep learning.

Note!

Python package download numbers can be inflated for the following reasons:

• CI systems: A lot of the downloads could be coming from CI systems. And R might have a lot less downloads from CI systems.
• Environments: Python users might be more likely to manage multiple environment and install packages in various environment. However, R users might just use one environment.
• Packaging updates: R users might not update packages as often as Python users.

References

https://r4ds.had.co.nz/
https://lgatto.github.io/IntroMachineLearningWithR/
https://topepo.github.io/caret/index.html
https://daviddalpiaz.github.io/r4sl/
https://www.tidyverse.org/