Package deepdep
was created to acquire and visualize
information on dependencies of R packages in a smart and convenient way.
Most of its functionality is contained in two functions:
deepdep
– to get a data.frame
with
dependencies described and plot_dependencies
– visualize
this data.frame
.
Suppose you’re creating an R package and you want to include graph of
its dependencies to your vignette, README.md
file on your
git repository or article on your package. With deepdep
you
simply need to type one line:
But before we describe how this function works in detail, let’s see what are other functionalities of the package.
deepdep
package exports the following functions:
get_available_packages
,get_description
,get_dependencies
,get_downloads
,deepdep
,plot_dependencies
.Those functions rely on each other and are ordered from the lowest to the highest level. We’ll describe what they exactly do and how on examples.
This function lists, as the name indicates, available packages. The default behaviour is listing all CRAN packages.
t <- get_available_packages()
head(t, 20)
#> A3 AATtools ABACUS abbreviate abbyyR abc abc.data
#> "A3" "AATtools" "ABACUS" "abbreviate" "abbyyR" "abc" "abc.data"
#> ABC.RAP abcADM ABCanalysis abclass ABCoptim ABCp2 abcrf
#> "ABC.RAP" "abcADM" "ABCanalysis" "abclass" "ABCoptim" "ABCp2" "abcrf"
#> abcrlda abctools abd abdiv abe abess
#> "abcrlda" "abctools" "abd" "abdiv" "abe" "abess"
However, if you want to check if package is present in a little wider
range – on CRAN or Bioconductor repositories, you simply need to set
argument bioc = TRUE
. In this case function is simply
wrapper around BiocManager::available()
and to use it you
need to have BiocManager
package (available via CRAN)
installed.
t <- get_available_packages(bioc = TRUE)
#> 'getOption("repos")' replaces Bioconductor standard repositories, see '?repositories' for details
#>
#> replacement repositories:
#> CRAN: https://cran.rstudio.com/
head(t, 20)
#> [1] "A3" "a4" "a4Base" "a4Classif" "a4Core" "a4Preproc" "a4Reporting"
#> [8] "AATtools" "ABACUS" "ABarray" "abbreviate" "abbyyR" "abc" "abc.data"
#> [15] "ABC.RAP" "abcADM" "ABCanalysis" "abclass" "ABCoptim" "ABCp2"
Another possibility is checking what packages are installed – you do
it by adding local = TRUE
parameter.
t <- get_available_packages(local = TRUE)
head(t, 20)
#> [1] "abind" "agricolae" "aigoracomponents" "aigoraML" "aigoraOpen"
#> [6] "aigorapack" "AlgDesign" "anytime" "AsioHeaders" "askpass"
#> [11] "assertthat" "attempt" "backports" "base64enc" "base64url"
#> [16] "bayestestR" "BBmisc" "bdsmatrix" "BH" "bigD"
Result of this function is cached (for more details, see Caching section of this vignette).
When you know, that given package is available, you may want to obtain DESCRIPTION of this package, at least the most essential parts of it, especially dependencies. You can do it by calling:
get_description("DALEXtra")
#> DALEXtra: Extension for 'DALEX' Package
#> Maintainer: Szymon Maksymiuk <sz.maksymiuk@gmail.com>
#> Description:
#> Provides wrapper of various machine learning models.
#> In applied machine learning, there
#> is a strong belief that we need to strike a balance
#> between interpretability and accuracy.
#> However, in field of the interpretable machine learning,
#> there are more and more new ideas for explaining black-box models,
#> that are implemented in 'R'.
#> 'DALEXtra' creates 'DALEX' Biecek (2018) <arXiv:1806.08915> explainer for many type of models
#> including those created using 'python' 'scikit-learn' and 'keras' libraries, and 'java' 'h2o' library.
#> Important part of the package is Champion-Challenger analysis and innovative approach
#> to model performance across subsets of test data presented in Funnel Plot.
#> Depends: R DALEX
#> Imports: reticulate ggplot2
#> LinkingTo:
#> Suggests: auditor gbm ggrepel h2o iml ingredients lime localModel mlr mlr3 ranger recipes rmarkdown rpart stacks xgboost testthat tidymodels
#> Enhances:
#> Scrap date: 2022-06-14 08:08:41
Again, you can pass bioc = TRUE
if you want to check for
this package in Bioconductor repository. Notice that if package is not
found there, it will be searched for on CRAN. The reason behind this
type of behaviour is the fact that packages present on Bioconductor are
updated more often than on CRAN and not all of them are present here.
Option local = TRUE
for only installed packages is also
possible. If a package is not available in a given source, the function
will return NULL
value:
get_description("a4")
#> NULL
get_description("a4", bioc = TRUE)
#> a4: Automated Affymetrix Array Analysis Umbrella Package
#> Maintainer: Laure Cougnaud <laure.cougnaud@openanalytics.eu>
#> Description:
#> Umbrella package is available for the entire Automated Affymetrix Array Analysis suite of package.
#> Depends: a4Base a4Preproc a4Classif a4Core a4Reporting
#> Imports:
#> LinkingTo:
#> Suggests: MLP nlcv ALL Cairo Rgraphviz GOstats
#> Enhances:
#> Scrap date:
Result of this function is also cached (for more details, see Caching section of this vignette).
This package allows you obtaining information on how many times specified package was downloaded. However, it works only with CRAN packages.
get_downloads("ggplot2")
#> x
#> last_day 58986
#> last_week 588212
#> last_month 2421187
#> last_quarter 7283366
#> last_half 15698538
#> grand_total 106188432
Results of this function is not cached.
After parsing description file, you can now create a
data.frame
which will describe dependencies between given
package and others. You do it by using this function:
get_dependencies("ggplot2")
#> name version type last_day last_week last_month last_quarter last_half grand_total
#> 1 cli <NA> Imports 37403 540894 2412409 7540831 14636052 65253097
#> 2 glue <NA> Imports 23255 299464 1227913 3316178 7392136 64049386
#> 3 gtable >= 0.1.1 Imports 12722 162527 681744 1919155 4289100 35736269
#> 4 isoband <NA> Imports 13772 178728 755858 2156434 4537600 25733208
#> 5 lifecycle > 1.0.1 Imports 33300 474632 2110034 6295444 12314841 57743668
#> 6 MASS <NA> Imports 3451 54701 240819 548525 1168085 12322995
#> 7 mgcv <NA> Imports 1466 22857 88442 283694 611885 8639283
#> 8 rlang >= 1.0.0 Imports 38603 552818 2426430 7234865 15325972 99564579
#> 9 scales >=\n1.2.0 Imports 19735 262475 1155303 3409107 6550755 46716655
#> 10 tibble <NA> Imports 21206 273333 1140332 3203317 7609105 69099588
#> 11 vctrs >= 0.5.0 Imports 36066 510423 2300925 6420051 13411712 68644002
#> 12 withr >= 2.5.0 Imports 20456 276656 1197213 3456127 6660468 45414397
As with two previously described functions -
get_available_packages
and get_description
,
here you can also use bioc = TRUE
or
local = TRUE
and again, in case the package is not
available, the result will be NULL
. Here you have another
options to set.
The first one is parameter downloads
– should number of
downloads of packages be included? It uses get_downloads
and works only with CRAN packages.
Another, more important parameter is dependency_type
.
You can specify how detailed should be list of dependencies. Default
value is "strong"
, which is a shorthand for
c("Depends", "Imports", "LinkingTo")
, but you can chose any
combination of those and additionally
"Suggests", "Enhances"
.
get_dependencies("ggplot2", downloads = FALSE, dependency_type = c("Imports", "Suggests", "Enhances"))
#> name version type
#> 1 cli <NA> Imports
#> 2 glue <NA> Imports
#> 3 gtable >= 0.1.1 Imports
#> 4 isoband <NA> Imports
#> 5 lifecycle > 1.0.1 Imports
#> 6 MASS <NA> Imports
#> 7 mgcv <NA> Imports
#> 8 rlang >= 1.0.0 Imports
#> 9 scales >=\n1.2.0 Imports
#> 10 tibble <NA> Imports
#> 11 vctrs >= 0.5.0 Imports
#> 12 withr >= 2.5.0 Imports
#> 13 covr <NA> Suggests
#> 14 dplyr <NA> Suggests
#> 15 ggplot2movies <NA> Suggests
#> 16 hexbin <NA> Suggests
#> 17 Hmisc <NA> Suggests
#> 18 knitr <NA> Suggests
#> 19 lattice <NA> Suggests
#> 20 mapproj <NA> Suggests
#> 21 maps <NA> Suggests
#> 22 maptools <NA> Suggests
#> 23 multcomp <NA> Suggests
#> 24 munsell <NA> Suggests
#> 25 nlme <NA> Suggests
#> 26 profvis <NA> Suggests
#> 27 quantreg <NA> Suggests
#> 28 ragg <NA> Suggests
#> 29 RColorBrewer <NA> Suggests
#> 30 rgeos <NA> Suggests
#> 31 rmarkdown <NA> Suggests
#> 32 rpart <NA> Suggests
#> 33 sf >=\n0.7-3 Suggests
#> 34 svglite >= 1.2.0.9001 Suggests
#> 35 testthat >= 3.1.2 Suggests
#> 36 vdiffr >= 1.0.0 Suggests
#> 37 xml2 <NA> Suggests
#> 38 sp <NA> Enhances
Result of this function is not cached (at least yet).
The main function of the package – it is simply wrapper around
get_dependencies
, that allows you getting not only
dependencies, but also dependencies of the dependencies iteratively!
(Now you know, why we called it deepdep).
Parameters are the same as in get_dependencies
, but
additionally you can specify depth
parameter, which
describes how many iterations it function should perform. If
depth
equals 1, it’s simply the same as calling
get_dependencies
.
deepdep("ggplot2", depth = 2)
#> origin name version type origin_level dest_level
#> 1 ggplot2 cli <NA> Imports 0 1
#> 2 ggplot2 glue <NA> Imports 0 1
#> 3 ggplot2 gtable >= 0.1.1 Imports 0 1
#> 4 ggplot2 isoband <NA> Imports 0 1
#> 5 ggplot2 lifecycle > 1.0.1 Imports 0 1
#> 6 ggplot2 MASS <NA> Imports 0 1
#> 7 ggplot2 mgcv <NA> Imports 0 1
#> 8 ggplot2 rlang >= 1.0.0 Imports 0 1
#> 9 ggplot2 scales >=\n1.2.0 Imports 0 1
#> 10 ggplot2 tibble <NA> Imports 0 1
#> 11 ggplot2 vctrs >= 0.5.0 Imports 0 1
#> 12 ggplot2 withr >= 2.5.0 Imports 0 1
#> 13 lifecycle cli >= 3.4.0 Imports 1 1
#> 14 lifecycle glue <NA> Imports 1 1
#> 15 lifecycle rlang >= 1.0.6 Imports 1 1
#> 16 mgcv nlme >= 3.1-64 Depends 1 2
#> 17 mgcv Matrix <NA> Imports 1 2
#> 18 scales farver >= 2.0.3 Imports 1 2
#> 19 scales labeling <NA> Imports 1 2
#> 20 scales lifecycle <NA> Imports 1 1
#> 21 scales munsell >= 0.5 Imports 1 2
#> 22 scales R6 <NA> Imports 1 2
#> 23 scales RColorBrewer <NA> Imports 1 2
#> 24 scales rlang >= 1.0.0 Imports 1 1
#> 25 scales viridisLite <NA> Imports 1 2
#> 26 tibble fansi >= 0.4.0 Imports 1 2
#> 27 tibble lifecycle >= 1.0.0 Imports 1 1
#> 28 tibble magrittr <NA> Imports 1 2
#> 29 tibble pillar >= 1.7.0 Imports 1 2
#> 30 tibble pkgconfig <NA> Imports 1 2
#> 31 tibble rlang >= 1.0.2 Imports 1 1
#> 32 tibble vctrs >= 0.3.8 Imports 1 1
#> 33 vctrs cli >= 3.4.0 Imports 1 1
#> 34 vctrs glue <NA> Imports 1 1
#> 35 vctrs lifecycle >= 1.0.3 Imports 1 1
#> 36 vctrs rlang >= 1.0.6 Imports 1 1
As famous quote says,
A picture is worth more than a thousand words.
That’s why we have plot_dependencies
function. It allows
visualizing easily what are dependencies of specified package.
The function is generic, and currently supports two types of object –
you can pass a deepdep
object, result of the calling the
deepdep
function or just type name of the package. With the
latter option you can also pass arguments to
get_dependencies
as additional parameters.
In each of the plots you can see one package name in the centre and two circles of packages gathered around them. These are dependencies of the first and second level.
Default plot type is circular, as you can see on the examples
presented above. However, you can set plot_type
parameter
to tree
.
Not all dependencies are plotted. To increase readability, dependencies on the same level are hidden, but you can change this behaviour
You can also make use of numbers of downloads you obtained. There is an option to add labels to only certain percentage of most downloaded packages among those that are about to be plotted. This is meant to increase readability of the plot.
Finally, returned object is a ggplot
object, so you can
easily manipulate them with syntax known from ggplot2
package. We also use ggraph
enhancement for plotting
graphs.
As we notoriously indicated in the text, we are using caching to make everything a little bit faster. Functions that operate on the lowest level, after scrapping information from the repositories, store them in temporary files on the local machine. As a consequence, calling function the second time with the same set of parameters, should be faster.
State is also refreshed every 20 minutes to make sure you don’t miss any major update.
To make sure that you have the most recent data, you need to call
get_available_packages
and get_description
with parameter reset_cache = TRUE
.