dateback is an R package that works as a virtual CRAN snapshot for source packages. It automatically downloads and installs tar.gz files with dependencies, all of which were available on a specific day.
install.packages("dateback")
Download dateback_x.y.z.tar.gz from https://github.com/r-suzuki/dateback/releases
Install it locally on R:
install.packages("dateback_x.y.z.tar.gz", repos = NULL)
devtools::install_github("r-suzuki/dateback")
To install ranger package and it's dependencies on the date 2023-03-01:
dateback::install(pkgs = "ranger", date = "2023-03-01")
Note that library(dateback) is not required. This package::function notation is recommended when using this package.
Or you can collect packages first, and install them later on (maybe on another system).
dateback::collect(pkgs = "ranger", date = "2023-03-01", outdir = "local_repo")
It downloads the latest tar.gz source packages on the day. Here is an excerpt from the log:
package file date status
1 Rcpp Rcpp_1.0.10.tar.gz 2023-01-22 archive
2 RcppEigen RcppEigen_0.3.3.9.3.tar.gz 2022-11-05 latest
3 ranger ranger_0.14.1.tar.gz 2022-06-18 archive
The output directory can be used as a local package repository:
install.packages(pkgs = "ranger", repos = "file:local_repo")
This package was originally developed to (partially) substitute the "CRAN Time Machine" (or "MRAN Time Machine"), which no longer work because of the retirement in July 2023.
As mentioned in the above URL, miniCRAN package would be a better choice if you want to archive the current packages and will use them in the future.
dateback will be helpful if you haven't archived packages in advance. It will include the following cases:
Reproduce the result of old code without pre-archived R packages.
Work on an older version of R, on which recent versions of some packages do not work properly (or cannot be installed) due to compatibility issues.
Need source package files to make a Docker image stable and reproducible, especially when using an older version of R.
Suppose we want to install a package X and its dependencies, which were available on a specific DATE. This is done by dateback with:
dateback::install(pkgs = X, date = DATE)
Let DATE_X_LATEST be the date when the latest version of X was published on CRAN. dateback works as follows:
DATE_X_LATEST <= DATE, downloads the latest version.DATE_X_LATEST > DATE), searches the "Archive" section of CRAN for the latest version on DATE, and downloads it.Then it unpacks the downloaded tar.gz file to check the dependencies.
Suppose that X is dependent on Y and Z. Then dateback downloads them with dependencies in the same manner, recursively.
It may seem trivial, but doing this manually can be difficult, time-consuming and error-prone with complex and/or diverse dependencies. The example below is the result of dateback::install("rstan", date = "2023-03-01"), which illustrates such a case:
package file date status
1 RcppParallel RcppParallel_5.1.7.tar.gz 2023-02-27 latest
2 Rcpp Rcpp_1.0.10.tar.gz 2023-01-22 archive
3 RcppEigen RcppEigen_0.3.3.9.3.tar.gz 2022-11-05 latest
...
41 pkgbuild pkgbuild_1.4.0.tar.gz 2022-11-27 archive
42 BH BH_1.81.0-1.tar.gz 2023-01-22 latest
43 rstan rstan_2.21.8.tar.gz 2023-01-17 latest
rstan is dependent on 13 packages (excluding base/recommended packages), and some of them are dependent on other packages. dateback downloaded 43 packages in total to solve all the dependencies on the specified date.
Posit Package Manager (PPM) has a snapshot feature, which can be used as a direct replacement for MRAN Time Machine. It provides both source and binary packages, and can be used with checkpoint package which originally needs MRAN Time Machine as a backend.
Compared to PPM, dateback would be evaluated as follows:
dateback has no explicit limit)dateback tries to reproduce past state from current CRAN contents, so changes in CRAN can affect the behaviorMany users would benefit from PPM, especially if using Windows/Mac. dateback would be a good choice for collecting source packages to be copied to a Linux container.
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4