gagolews/datastructures

General data structures and algorithms for use with R

Prof. Marek Gagolewski

My research interests include machine learning, data aggregation and clustering, computational and applied statistics, and mathematical modelling (the science of science, sport, economics, social sciences, psychometrics, bibliometrics, etc.).

In my spare time, I write books for my students and develop open-source data analysis software.

Open-access textbooks

• Deep R Programming (HTML) (PDF) (paper copy) (GitHub)

• Minimalist Data Wrangling in Python (HTML) (PDF) (paper copy) (GitHub)

Software

Python packages

• genieclust – Fast and robust hierarchical clustering with noise point detection (GitHub) (PyPI) (paper)
• clustering-benchmarks – A framework for benchmarking clustering algorithms (GitHub) (PyPI) (paper)

R packages

• stringi – Fast and portable character string processing in R (one of the most often downloaded packages for R) (GitHub) (CRAN) (paper)
• genieclust – Fast and robust hierarchical clustering with noise point detection (GitHub) (CRAN) (paper)
• stringx – Drop-in replacements for base R string functions powered by stringi (GitHub) (CRAN)
• realtest – Where expectations meet reality: Realistic unit testing in R (GitHub) (CRAN)
• TurtleGraphics – Learn computer programming in R while having a jolly time! (GitHub) (CRAN)

Data

• Clustering benchmarks (framework, datasets, results)
• Datasets for teaching

datastructures's Issues

Nice interface by classes

I think it would be nice to have something more convenient than bunch of methods like vector_push_back() or queue_pop(), where external pointer is needed as an argument. More readable would be something like vector.push_back(6).

And my proposition is like that:

dsVector <- setRefClass("Vector",
                    fields=c("vec"
                               ),
                    methods = list(
                       initialize=function(n=0)
                       {
                          vec <<- vector_create(n)

                       },

                       push_back=function(obj)
                       {
                          vector_push_back(vec, obj)
                       },

                       back=function(obj)
                       {
                          vector_back(vec)
                       },

                       front=function(obj)
                       {
                          vector_front(vec)
                       },

                       size=function()
                       {
                           vector_size(vec)
                       }

))

and usage;

vec2 <- dsVector$new(5)
vec2$size()
vec2$push_back(1)
vec2$front()
vec2$back()

My only concern is how to make indexer, like "vec2[3] = 5", but I think it is possible in one way or another.

OK, maybe another concern: how to do something like as.character(vector) or print(vector).

stack, queue: add as.list

It'd be lovely if we were able to convert a stack or a queue to an R list object

so we need

as.list.stack and as.list.queue

moreover, queue/list constructor should mark class of retval to queue/stack
print.stack, print.queue and maybe format.stack and format.queue should be defined

Vector class

We should implement a vector class, using std:vector from C++, which would allow adding new items to a vector with amortized constant complexity (it would solve a "dynamically extending vector" problem).

The main question is: should it be a vector of any R objects (it would be like an R list), or maybe one type of objects (more like an R vector)?

And is there any way to implement operator [], like vector[5]? Or the only way is to make a function get_element(vector, 5) and set_element(vector, 5)?