V3: Package versions and dependencies
Introduction
RWsearch stands for « Search in R packages, task views, CRAN and in the Web ».
This vignette introduces the following features cited in the README file:
.7. Plot the graph of the package dependencies in an html page in your browser.
.8. List the parent and children dependencies of one or several packages, either the first level or the full list (termed recursive).
.9. Use this information to reinstall or install packages and their dependencies in the right order, from the lowest number to the largest number of dependencies, and avoid mic-mac (very useful).
Package dependencies
Let’s assume we wish to update or install the packages RWsearch, tfestimators and their respective dependencies depicted in the image below. In which order should we install them ?
A simple strategy is to apply install.packages(c(“RWsearch”, “tfestimators”)) and let the function manage the dependency order. A better strategy is to list all packages from the ones with no parent to the ones with many parents, then install these packages according to the list order. (Parallel) Installation is much faster and more secure as there is no recursive call and no branching. The list and the package order is provided with no pain by the instruction p_vers_deps().
As from RWsearch_4.9.4 published on 2021-06-08, the table binarydb provides additional information about the binary version and the source version available on CRAN. This table is optional but useful for Windows and macOS users.
crandb_load(system.file("data", "zcrandb.rda", package = "RWsearch"))
binarydb_load(system.file("data", "zbinarydb.rda", package = "RWsearch"))
# or
# crandb_down()
# binarydb_down()p_vers_deps() lists the package dependencies in a proper order
This list is the one in 2019, without the information provided by binarydb.
dfr <- p_vers_deps(tfestimators, RWsearch) ; dfr # 2019
# nsloaded version crandb compare tdeps ndeps
# forge FALSE <NA> <NA> 3 0 0
# tfruns FALSE <NA> <NA> 3 0 0
# reticulate FALSE 1.20 <NA> 2 0 0
# config FALSE 0.3.1 <NA> 2 0 0
# yaml FALSE 2.2.1 <NA> 2 0 0
# pkgconfig TRUE 2.0.3 <NA> 2 0 0
# digest TRUE 0.6.25 <NA> 2 0 0
# base64enc FALSE 0.1-3 <NA> 2 0 0
# progress FALSE 1.2.2 <NA> 2 0 0
# tidyselect FALSE 1.1.1 <NA> 2 0 0
# purrr FALSE 0.3.4 <NA> 2 0 0
# tibble FALSE 3.1.2 <NA> 2 0 0
# tidyr FALSE 1.1.3 <NA> 2 0 0
# processx FALSE 3.4.2 <NA> 2 0 0
# rstudioapi FALSE 0.13 <NA> 2 0 0
# rlang TRUE 0.4.10 0.4.11 -1 0 0
# brew TRUE 1.0-6 1.0-6 0 0 0
# latexpdf TRUE 0.1.6 0.1.6 0 0 0
# sig TRUE 0.0-5 0.0-5 0 0 0
# magrittr TRUE 2.0.1 2.0.1 0 0 0
# jsonlite FALSE 1.7.1 1.7.2 -1 1 0
# XML TRUE 3.99-0.3 3.99-0.6 -1 2 0
# sos TRUE 2.0-2 2.0-2 0 1 1
# igraph TRUE 1.2.5 1.2.6 -1 10 2
# htmltools TRUE 0.4.0 0.5.1.1 -1 5 3
# htmlwidgets TRUE 1.5.1 1.5.3 -1 9 6
# tensorflow FALSE <NA> 2.4.0 -2 9 7
# networkD3 TRUE 0.4 0.4 0 18 10
# tfestimators FALSE <NA> 1.9.1 -2 18 16
# RWsearch TRUE 4.9.1 4.8.0 1 24 16 This list is the one in 2024, with the information added by binarydb. The much higher number of dependencies comes from the rmarkdown and htmlwidgets package which now use packages from the tidyverse. See the graphs below.
dfr <- p_vers_deps(tfestimators, RWsearch) ; dfr # 2024
# nsloaded version binary crandb difvb difbc compare gcc tdeps ndeps
# brew TRUE 1.0-10 1.0-10 1.0-10 0 0 0 no 0 0
# latexpdf TRUE 0.1.8 0.1.8 0.1.8 0 0 0 no 0 0
# sig TRUE 0.0-6 0.0-6 0.0-6 0 0 0 no 0 0
# magrittr TRUE 2.0.3 2.0.3 2.0.3 0 0 0 yes 0 0
# yaml TRUE 2.3.10 2.3.10 2.3.10 0 0 0 yes 0 0
# cpp11 FALSE 0.5.0 0.5.0 0.5.0 0 0 0 no 0 0
# base64enc FALSE 0.1-3 0.1-3 0.1-3 0 0 0 yes 0 0
# fastmap TRUE 1.2.0 1.2.0 1.2.0 0 0 0 yes 0 0
# R6 FALSE 2.5.1 2.5.1 2.5.1 0 0 0 no 0 0
# rappdirs FALSE 0.3.3 0.3.3 0.3.3 0 0 0 yes 0 0
# evaluate FALSE 1.0.0 1.0.0 1.0.1 0 -1 -1 no 0 0
# jsonlite TRUE 1.8.9 1.8.9 1.8.9 0 0 0 yes 1 0
# cli TRUE 3.6.3 3.6.3 3.6.3 0 0 0 yes 1 0
# pkgconfig TRUE 2.0.3 2.0.3 2.0.3 0 0 0 no 1 0
# rlang TRUE 1.1.4 1.1.4 1.1.4 0 0 0 yes 1 0
# glue FALSE 1.8.0 1.8.0 1.8.0 0 0 0 yes 1 0
# mime FALSE 0.12 0.12 0.12 0 0 0 yes 1 0
# fs FALSE 1.6.4 1.6.4 1.6.4 0 0 0 yes 1 0
# digest TRUE 0.6.36 0.6.37 0.6.37 -1 0 -1 yes 1 0
# XML TRUE 3.99-0.17 3.99-0.17 3.99-0.17 0 0 0 yes 2 0
# xfun FALSE 0.48 0.48 0.48 0 0 0 yes 3 0
# sos TRUE 2.1-8 2.1-8 2.1-8 0 0 0 no 1 1
# highr FALSE 0.11 0.11 0.11 0 0 0 no 4 1
# tinytex FALSE 0.53 0.53 0.53 0 0 0 no 4 1
# cachem FALSE 1.1.0 1.1.0 1.1.0 0 0 0 yes 3 2
# memoise FALSE 2.0.1 2.0.1 2.0.1 0 0 0 no 4 3
# lifecycle TRUE 1.0.4 1.0.4 1.0.4 0 0 0 no 5 3
# htmltools TRUE 0.5.8.1 0.5.8.1 0.5.8.1 0 0 0 yes 6 4
# vctrs FALSE 0.6.5 0.6.5 0.6.5 0 0 0 yes 6 4
# knitr FALSE 1.48 1.48 1.48 0 0 0 no 8 4
# fontawesome FALSE 0.5.2 0.5.2 0.5.2 0 0 0 no 7 5
# jquerylib FALSE 0.1.4 0.1.4 0.1.4 0 0 0 no 7 5
# sass FALSE 0.4.9 0.4.9 0.4.9 0 0 0 yes 11 8
# igraph TRUE 2.0.3 2.0.3 2.0.3 0 0 0 yes 16 8
# bslib FALSE 0.8.0 0.8.0 0.8.0 0 0 0 no 21 17
# rmarkdown FALSE 2.28 2.28 2.28 0 0 0 no 30 25
# htmlwidgets TRUE 1.6.4 1.6.4 1.6.4 0 0 0 no 31 26
# networkD3 TRUE 0.4 0.4 0.4 0 0 0 no 41 32
# RWsearch TRUE 5.1.9 5.1.4 5.1.4 1 0 1 no 47 38The colnames are:
- nsloaded: is the package loaded in the namespace or not? Here, RWsearch and its dependencies are looded and locked. A (ns)loaded package cannot be updated on the spot (especially if it has a dll/shared library, or S4 classes) but requires a fresh version of R with no external package loaded. Use preferably install.packages() or in some extreme cases R CMD INSTALL.
- version is the current version installed in your computer or NA if the package is not installed.
- binarydb is the binary version recorded in the binarydb/CRAN data.frame, either for macOS or Windows.
- crandb is the tar.gz version recorded in the crandb/CRAN data.frame.
- difvb compares the version intalled on the computer with the binary version in binarydb/CRAN.
- difbc compares the binary version in binarydb/CRAN with the source version in crandb/CRAN.
- compare compares the two version numbers:
- -2 is for a package not installed on the computer but available on CRAN.
- -1 is for a package installed on the computer but with an older version than the one on CRAN.
- 0 are for installed packages up to date.
- +1 is for a fresh package that is more recent than the one on CRAN, typically the packages under development or downloaded from another repository.
- +2 is for a package installed on the computer but not available on CRAN. It can be either a package that comes from another repository (github, gitlab, Bioconductor, R-forge, etc) or a pckage that has been removed from CRAN and archived.
- +3 is for a package that is not installed on the computer and not available. It can ba for a instance a package available on another repository (github, gitlab, Bioconductor, R-forge, etc).
- gcc informs that a compilation is needed to build the binary and install the package.
- tdeps is the total number of recursive parent dependencies, including the generic packages (stats, utils, tools, parallel, etc).
- ndeps is the total number of recursive parent dependencies excluding the generic packages, since they are provided with every R version and are not stored on CRAN.
Packages with comparison number -2 and -1 can be installed according to the list order, which is from low values to high values of ndeps. On Windows, I tend to wait for a few days and install the binary packages.
p_graphF() generates the graph of the package dependencies
p_graphF() generates a graph of the package dependencies and send it to the browser. The first image is the one in 2019 and the second image (only for RWsearch) is for September 2024. The much higher number of dependencies comes from the rmarkdown and htmlwidgets package which now use packages from the tidyverse.
Subset the data.frame of the package dependencies
From the p_vers() dfr table above, the packages to be updated are:
subset(dfr, compare < 0)
nsloaded version binary crandb difvb difbc compare gcc tdeps ndeps
evaluate FALSE 1.0.0 1.0.0 1.0.1 0 -1 -1 no 0 0
digest TRUE 0.6.36 0.6.37 0.6.37 -1 0 -1 yes 1 0but digest is loaded in the namespace and cannot be installed immediatly with RWsearch::p_inst. RWsearch must be closed and digest must be installed with the classical install.packages() function.