Software Design and Development for Statistical Data Analysis Packages in R, Lecture notes of Data Structures and Algorithms

Download Software Design and Development for Statistical Data Analysis Packages in R and more Lecture notes Data Structures and Algorithms in PDF only on Docsity! Software Design and Development for Statistical Data Analysis Packages in R Kyle D. Bemis 4 March 2015 Purdue University Outline • Why plan your package? • Object-oriented programming in R • S3 • S4 • Using C or C++ in an R function • Creating a package • Example: Cardinal • Debugging and development tips Object-oriented programming in R • R is primarily a functional language • It has some object-oriented features • S3 • S4 S3 is simple • Less formal style of object-oriented programming • class(object) <- “Animal” • Objects are usually lists • Method dispatch based on naming scheme • foo.Animal <- function(…) { … } • Better for simpler data structures, such as results and output from analysis S4 is more complex • Formal class declaration • setClass(“Animal”, …) • Objects can be very complex data structures • Method dispatch implemented in “methods” package • setMethod(“foo”, “Animal”, function(…) {…}) • Better for more complex data structures, such as experiment workflows (genomics, proteomics, etc.) Build and install a package • From R: • install.packages(“MyPackage”, repos=NULL, type=“source”) • From command line: • R CMD build MyPackage • R CMD INSTALL MyPackage_0.1.tar.gz Example: Mass spectrometry imaging 200 400 600 800 1000 0 20 40 60 80 10 0 12 0 DESI Spectrum (x,y) = (33,80) MZ In te ns ity ● 20 40 60 80 100 120 10 20 30 40 50 60 70 M/Z = 887.67 x y thousands of mass spectra hundreds of ions MS image data structure x = 1, y = 1 x = 1, y = 2 x = 2, y = 1 x = 2, y = 2 m/z = 100 0.1 0.6 2.1 1.3 m/z = 101 3.4 0.8 2.3 1.7 m/z = 102 1.5 1.9 1.6 0.2 Debugging and development tips • Debugging in R: • browser() • options(error=recover) • Source control (Github, SVN, etc.) Questions? Thank you!