Workflowr Versions Save

Organize your project into a research website

v1.7.1

8 months ago

This patch release includes improved documentation, bug fixes, and a reduction in the number of Suggested dependencies.

Updated documentation

  • Update single-page FAQ to explain how to remove navbar

  • Update badge links (implemented by @wyq977 in #276)

  • Explicitly label internal functions with roxygen2

  • Build online docs via GitHub Actions and push to gh-pages

  • Improve guidance on rendering a standalone HTML file in FAQ vignette (idea from @fkgruber in #283)

  • Add the appropriate PKGNAME-package \alias to the package overview help file as per "Documenting packages" in Writing R Extensions

  • Modernize CITATION. Replace citEntry() with bibentry(), replace personList() with c(), format as R code, and simplify.

Bug fixes

  • Fix missing figure version table on Windows (implemented by @warmdev in #275)

  • Use conditionMessage() to extract error message from callr subprocess r-lib/callr#228

  • Remove warning from sprintf() in print.wflow_start() when git = FALSE

Dependencies

  • Removed from Imports: xfun
  • Removed from Suggests: covr, devtools, spelling
  • Added to Suggests: sessioninfo

v1.7.0

2 years ago

This minor release includes some new features, improved documentation, and bug fixes.

Minor improvements

  • New argument combine for wflow_build() and wflow_publish(). When Rmd files are specified with the argument files, they are built in addition to any Rmd files that are automatically built when setting arguments like make = TRUE and republish = TRUE. If you would instead like to only build Rmd files that are included in all the filters, you can set combine = "and" to take the intersection. For example, if you ran wflow_build("analysis/example*.Rmd", make = TRUE, combine = "and"), then this would only build those Rmd files matched by the file glob analysis/example*.Rmd and had been modified more recently than their corresponding HTML file. With the default, combine = "or", this would have built all the files that matched the file glob in addition to any files that had been modified more recently than their corresponding HTML file, even if they didn't match the file glob pattern (idea from @bfairkun in #221, implementation by @zaynaib in #227, #228)

  • If wflow_publish() is called, but was not instructed which files to publish, it now throws an error. In other words, you must specify the files you wish to publish or use one of the convenience arguments like republish = TRUE or update = TRUE. It's previous behavior was to complete without having done anything, which was misleading (idea from @stephens999)

  • It is now easier to enter commit messages with a separate title and body. If you pass a character vector to the argument message to any of the functions that perform a commit, e.g. wflow_publish(), the first element will be used as the title, and any subsequent elements will be separate paragraphs in the commit body. Using a separate title and body will improve the display of your commit messages on GitHub/GitLab and git log --oneline since these only show the title (suggestion from @LearnUseZone in #222, implementation by @zaynaib in #225)

  • New argument only_published for wflow_toc(). If set to FALSE, then the table of contents will also include unpublished files (implemented by @giocomai in #234)

Updated documentation

  • Added more documentation to wflow_build() to explain when it does and doesn't load code defined in a .Rprofile file (idea from @pcarbo)

Bug fixes

  • Bug fix: Now workflowr will detect any problems with its dependencies when it is attached. All dependencies must be installed, loadable, and meet the minimum required version. Broken packages were causing cryptic, misleading errors (reported by in @markellekelly in #216 and @LearnUseZone in #217)

  • Bug fix: wflow_quickstart() can now handle relative paths to the Rmd files when the working directory is changed (change_wd = TRUE, which is the default)

  • Remove Rd warnings when installing package on Windows by explicitly specifying the topic page when cross-referencing an exported function from another package. Note that the links worked previously, so this change is just being proactive in case this warning starts getting strongly enforced. If the authors of the other package rearrange how they group functions into documentation topics, this will break the cross-references and require an update. See this thread for more details

  • Bug fix: wflow_use_github() and wflow_use_gitlab() now use Font Awesome 5 syntax to insert icons into the navigation bar when a recent version of rmarkdown is installed (>= 2.6) (bug report from @christianholland, #231)

  • Bug fix: wflow_open() no longer sends a warning if you are using bookdown::html_document2 as your primary output format in _site.yml with base_format: workflowr::wflow_html (bug report from @rgayler, #233)

Miscellaneous

  • Removed function wflow_update(). Its only purpose was to migrate projects created with [workflowrBeta][], which is now over 3 years old

  • Bump minimum required version of R from 3.2.5 to 3.3.0. While workflowr itself should be able to continue to work fine with R 3.2.5, it was becoming too much of a burden to regularly test workflowr with R 3.2.5 as the RStudio engineers have started updating their packages to require a minimum of R 3.3.0

  • Require minimum versions of callr 3.7.0, knitr 1.29, rmarkdown 1.18

  • Switched to the workflowr repository itself to use the default branch "main" and changed the owner to the workflowr organization. This has no effect on workflowr projects (future or existing). It mainly affects contributors to workflowr development. However, please update any links you might have bookmarked (e.g. to documentation)

v1.6.2

4 years ago

This patch release of workflowr includes minor improvements, updated documentation, and bug fixes. Users are encouraged to upgrade workflowr so that it is compatible with the next release of fs (>1.4.1).

Minor improvements

  • Make the HTTP calls by wflow_use_github() to the GitHub API more resilient to transient network issues by automatically retrying HTTP calls more than once (@jameslamb, #199, #201)
  • Enhance the RStudio Project Template to include more arguments from wflow_start() (@trannhatanh89, #193, #200)
  • Throw error early if user sets overwrite = TRUE to wflow_start() but existing = FALSE, since it isn't possible to overwrite non-existent files (@skpurdue, #194, #202, idea from @pcarbo)

Updated documentation

  • Update FAQ entry on external images to note that the website directory for GitLab Pages is public/, not docs/ (idea from @ryurko)

Bug fixes

  • Send warning if Python plots created using outdated version of reticulate (<1.15) (bug report from @lazappi, #181)
  • Make workflowr compatible with fs version 1.4.1.9000+ (@jimhester, #204)
  • Set minimum version requirement for dependency rprojroot to 1.2
  • Skip test of error handling of deleted working directory on Solaris

v1.6.1

4 years ago

This patch release of workflowr includes minor improvements, updated documentation, bug fixes, and lots of internal refactoring.

Minor improvements

  • Workflowr functions check if the working directory exists (idea from @pcarbo, #185)

  • wflow_use_github() can now create repositories for GitHub organizations using the new argument organization. However, GitLab Groups should still be specified with the argument username for wflow_use_gitlab() (bug report from @stephens999, #186)

  • Any workflowr function that prompts for user input will continue to re-prompt until valid input has been entered (or canceled by hitting the key Esc) (idea from @pcarbo)

  • All functions that commit to the Git repository first check for the availability of the Git variables user.name and user.email (either global or local), which are required for creating a commit (#85)

  • The workflowr icon is displayed in browser tabs of workflowr websites (idea from @pcarbo)

Updated documentation

  • Document that GitLab.com provides private repositories with access control to the source code repository and the website (#187)

  • Update URL to RStudio-specific instructions for installing pandoc (bug report from @PietrH, #190)

  • Document in the FAQ vignette how to include an external image using an absolute URL to the image hosted on another website (idea from @JiaxiangBU, #159)

  • Document how to embed Shiny apps with knitr::include_app() and shinyapps.io (idea from @rsimon64, #130)

  • Update the FAQ on including external images to account for the breaking change in knitr::include_graphics() introduced in knitr version 1.28 (bug report from @ditordccaa, #103)

  • Add FAQ entry on how to use a Git hosting site that uses the HTTP protocol (idea from @antass, #163)

Bug fixes

  • Fix Windows-specific bug that caused the table of past versions to be missing from the workflowr report (bug introduced in version 1.5.0)

Internal refactoring

  • Internal refactoring for increased speed and improved error handling of input arguments

  • Check for class with inherits() (#189)

  • Switch to use pandoc option --include-in-header to insert workflowr-specific CSS and other metadata (surprisingly --include-before-body works fine)

v1.6.0

4 years ago

New features

  • New function wflow_run(). It executes the code chunks of an R Markdown file in the current R session without affecting any of the website files. This is meant to be used while interactively developing an analysis. It does not change the working directory or isolate the computation from the current R session. This is analogous to the RStudio option "Run all" to run all the code chunks (idea from @pcarbo)

  • New autosave feature. The workflowr functions wflow_build(), wflow_publish(), and wflow_status() will autosave any unsaved files open in the RStudio editor pane. This is similar to the behavior of the Knit HTML button. This feature can be disabled by setting the package option workflowr.autosave to FALSE (idea from @xiangzhu in #179)

  • New vignette on using large data files in a workflowr project (motivated by @xiangzhu, #183)

  • If there are merge conflicts after running wflow_git_pull(), and the merge was allowed to proceed (fail = FALSE), then the conflicted files are listed and optionally opened in RStudio at the first line of the conflict that needs to be resolved

  • wflow_git_config() has a new argument overwrite. Previously wflow_git_config() would by default overwrite any previous settings. Now this will throw an error. To overwrite a previous setting, set overwrite = TRUE (idea from @petebaker)

Minor improvements and bug fixes

  • Warn user if only HTML file has been committed (and avoid throwing an error). Previously this threw an error because workflowr expects the R Markdown file to be committed to the Git repo if its corresponding HTML file is (bug report from @kevinlkx)

  • Warn user if a dependency does not meet the minimum required version. There are multiple ways this could happen. First, it is possible to install an old version after having installed workflowr. Second, when running install.packages(), if the minimum required version is available in any of the package libraries, it is not installed. However, if the version of the package in the first directory listed in .libPaths() does not meet the minimum required version, it is still the one that is loaded when workflowr is loaded (idea from @stephens999)

  • Fix off-by-one date bug by specifying the local timezone (see git2r Issue 407)

  • Fix bug when path to project includes a space. The bug was introduced in version 1.5.0 with the feature to use the system Git executable to run git ls-files. To fix the issue in version 1.5.0, set options(workflowr.sysgit = "") in the file .Rprofile (bug report from @wolfemd, #180)

  • Fix bug caused by an unset timezone. If the machine has no timezone set, workflowr will default to Etc/UTC

  • Handle missing title/pagetitle warning from pandoc 2+ and rmarkdown 1.18+ (see rmarkdown Issue 1355)

  • Improve speed of wflow_publish()/wflow_status() by using the system Git executable (if available) to obtain the last commit time of the analysis files (this is used to determine which published Rmd files are outdated and need to be republished)

  • Report exact command to run git push or git pull in the terminal if either wflow_git_push() or wflow_git_pull() fail (reported by @jennysjaarda, #182)

  • Update FAQ to include how to create a PDF using the RStudio Knit button (reported by @han16)

  • Update citation to workflowr publication

  • Properly quote the Git executable with shQuote() whenever Git is called from R

v1.5.0

4 years ago

This minor release of workflowr includes a new function, the introduction of options to control package-wide function behavior, the ability to suppress the workflowr report from the HTML file, a new vignette for teaching workflowr, and lots of error handling improvements.

New function wflow_quickstart()

The new function wflow_quickstart() provides a simple interface to effortlessly create a workflowr project from an existing data analysis. Pass it your existing R Markdown file(s), and it will start a new workflowr project, publish the analysis files, and configure GitHub (or GitLab).

wflow_quickstart(files = "existing-analysis.Rmd", username = "your-github-username")

Package options

This is the first release to include options for controlling the behavior of all workflowr functions. This makes it more convenient for you to create a consistent workflowr experience. You can set the options in your project's .Rprofile (using the function options()) instead of having to always remember to change a default argument every time you call a function.

Currently there are two workflowr package options. See ?workflowr for more details.

  • workflowr.git: Set the path to the system Git executable, which is occasionally used to increase the speed of Git operations performed by workflowr functions.

  • workflowr.view: Should workflowr functions open webpages for viewing in the browser? The default is set to interactive() (i.e. it is TRUE only if it is an interactive R session). This option is currently used by wflow_build(), wflow_git_push(), and wflow_publish().

Workshop tutorial for teaching workflowr

The new vignette "Reproducible research with workflowr" is designed to be taught as a tutorial in a workshop setting. It includes setup instructions, an example analysis to highlight the benefits of workflowr, and troubleshooting advice.

Suppress the HTML workflowr report

If you'd like to suppress the workflowr report at the top of an HTML page, you can set the option suppress_report to TRUE. To suppress the report in every HTML file, set the option in _workflowr.yml:

suppress_report: TRUE

To suppress the report in a specific HTML file, add the following to the YAML header of the corresponding Rmd file:

workflowr:
  suppress_report: TRUE

Many thanks to @kaneplusplus for implementing this feature! (#168)

Minor improvements and bug fixes

  • Require git2r >= 0.26.0 to support internal changes that increase speed and robustness of workflowr Git functionality

  • wflow_start() adds a .gitattributes file that classifies R Markdown files as R code for the official GitHub language statistics calculated via linguist. The default setting is to ignore R Markdown files.

  • Address callr 3.3.0 bug that writes objects to the global environment, causing the workflowr check of the global environment to fail. The failed check now explains that the problem can be fixed by updating the callr package.
  • Warn user from wflow_build() if index.Rmd is missing the workflowr-specific site generator wflow_site() (idea from @pcarbo, #177)

  • Fail early if missing required file index.Rmd

  • Add argument fail to wflow_git_pull() with default value of TRUE. Now if a pull generates a merge conflict, wflow_git_pull() will abort the pull. Thus no changes will be made to the local files. Users can set fail = FALSE to allow Git to add the merge conflicts to the local files.

  • Speed improvements for wflow_publish()

  • Improved error handling when files contain merge conflicts. Before workflowr only detected merge conflicts in Rmd files. Now it detects them for any file (since Git requires any merge conflicts to be resolved before it makes any new commits).

  • Warn user if knit_root_dir (the directory where the code in the Rmd files is executed) defined in _workflowr.yml is an absolute path. An absolute path would only work on the current computer, limiting reproducibility.

  • Include Git status in output of wflow_status(). Note that it purposefully excludes any files in the website directory since these generated files should only be committed by workflowr. You can omit the Git status by setting include_git_status = FALSE (idea from @pcarbo)

  • Make it clearer that you have two options for creating the remote repository on Github: 1) let wflow_use_github() do it automatically, or 2) create it yourself manually at https://github.com/new (idea from @pcarbo)

  • New FAQ "How can I save a figure in a vector graphics format (e.g. PDF)?"

  • Added citation to F1000Research paper. Run citation("workflowr") to obtain the new citation information.

  • Fixed wflow_start() infinite recursion bug by requiring stringr >=1.3.0

  • Added httpuv as imported dependency so that wflow_use_github() is able to automatically create the GitHub repository via the httr package

  • Set (or increased) minimum required versions for fs, git2r, httpuv, rstudioapi, stringr, whisker, clipr, shiny, testthat, and withr

  • Document possible error of a greyed out GitHub authentication button when trying to give permission for workflowr to create a repository for your account

  • Fixed bug in date displayed in table of past versions in the workflowr report. Depending on the time of day the commit was made, the displayed day may have been off by one.

v1.4.0

4 years ago

This minor release of workflowr features further GitHub integration, a new reproducibility check, and various improvements and bug fixes.

New features

  • The initial GitHub setup always included the manual step of creating the GitHub repository, but this is no longer the case! When you run wflow_use_github("username"), it will offer to create the new GitHub repository for you. If you sign-in to GitHub via your web browser and grant workflowr permission, it will be created automatically.

  • Absolute paths to files on your local computer are not reproducible. If you or someone else tries to execute the code on a different machine, it will fail. Now workflowr will automatically search for absolute paths to files that are inside of the workflowr project. If it detects any, it will fail the reproducibility check, and provide you with the equivalent relative paths to use.

  • Now every time you push your latest changes with wflow_git_push(), a new browser tab will automatically be opened to your online repository (idea from @pcarbo)

Minor improvements and bug fixes

  • Mention knitr::include_graphics() as an option for including external images (idea from @Zepeng-Mu, #162)
  • Check if Git repository is locked and produce error message that explains how to fix it (idea from @brimittleman)
  • Document that wflow_build() and wflow_publish() build the files in the given order if they are provided explicitly to the argument files (idea from @antass and @pcarbo, #164)
  • Handle spaces in chunk names for HTML targets. If a chunk name contains a space, the name of the figure file will also has a space. This broke the accompanying link for the HTML button with the table of previous versions of the figure (because it uses the filename to be unique). Now the spaces are removed in the HTML link.
  • Add FAQ entry on installing packages in a workflowr project (idea from @xiangzhu, #160)
  • wflow_use_github()/wflow_use_gitlab() provide better guesses if the arguments username or repository are left blank. If repository is NULL, it is set to the name of the workflowr project directory (idea from @pcarbo).
  • wflow_git_push() and wflow_git_pull() no longer accept direct URLs to remote repositories. The argument remote must be NULL or the name of an existing remote. The support for direct URLs was likely rarely used since it is rarely used with Git as well, and it likely never worked given how the underlying functions from git2r work.
  • Document in the GitLab vignette that the repository will be automatically created the first time the repository is pushed with wflow_git_push() (this is a feature unique to GitLab)

v1.3.0

5 years ago

This minor release of workflowr introduces two new functions, RStudio Addins, and various minor improvements.

New functions

  • The new function wflow_toc() builds a table of contents of the published R Markdown files in a workflowr project (@JiaxiangBU, #151, #155)

  • The new function wflow_rename_proj() renames a workflowr project throughout all its project files (idea from @frm1789 and @kbroman, #148)

RStudio Addins

RStudio Addins allow you to execute R code via the RStudio Addins menu. For extra convenience, you can bind the addins to keyboard shortcuts. The following workflowr functions have addins:

  • wflow_build()
  • wflow_publish()
  • wflow_status()
  • wflow_toc()
  • wflow_view()

Note that the addin for wflow_publish() is a Shiny Gadget that enables you to interactively choose which files to publish and write a detailed commit message (assistance from @zaynaib and @argdata, #143).

Minor improvements and bug fixes

  • wflow_build() fails early if pandoc is not installed (@zaynaib, #75)

  • wflow_git_push()/wflow_git_pull() fail early if user tries to use SSH authentication when it is not supported by the current installation of git2r/libgit2 (#144)

  • Fix support for knitr chunk option collapse and indent (reported by @pcarbo, #149)

  • Fix support for rmarkdown option keep_md (see rmarkdown Issue 1558)

  • Skip tests that only fail on CRAN servers (this is why there are no macOS binaries for 1.2.0)

  • Add a GitHub Pull Request template

  • Rename reproducibility tab "Report" to "Checks" (idea from @pcarbo)

  • Fix spacing issue with session information button (reported by @pcarbo, #157)

  • wflow_status() reports if the configuration files _workflowr.yml and _site.yml have been edited

  • Disable inline code chunks by default in R Markdown files created by workflowr. Document how to use inline code chunks in the FAQ (discussed with @rgayler and @Robinlovelace , #140)

v1.2.0

5 years ago

This release overhauls the layout of the reproducibility report, adds support for GitLab, introduces some safety checks and warnings related to caching, provides more documentation, and more.

Full support for GitLab

While it has always been possible to host workflowr projects using platforms other than GitHub, it was less convenient and not all the features were supported. With this release, a workflowr project hosted on GitLab has all the same features as a workflowr project hosted on GitHub, including links to past versions of the R Markdown, HTML, and figure files. It's also possible to use workflowr with GitHub Enterprise or a custom GitLab instance, but you'll have to coordinate with your administrators to make sure it's possible to host the website.

  • Create vignette "Hosting workflowr websites using GitLab" (written with @lazappi, #112)
  • Add argument domain to wflow_git_remote() to allow specifying any remote Git host, e.g. domain = "gitlab.com"
  • Create function wflow_use_gitlab() to automate GitLab setup

Reproducibility report and other layout changes

The layout of the reproducibility report and other content that workflowr inserts in the HTML output has been overhauled to be both highly informative but also collapsed by default. This way the information is there if you need it, but otherwise is minimally distracting.

  • Collapse the reproducibility report (suggested by @timtrice, #110)
  • Collapse the "Session information" chunk (suggested by @xiangzhu, #120)
  • Convert table of past figure versions to a collapsable button (suggested by @xiangzhu, #120)
  • Improve formatting of tables of past version of files using Bootstrap table classes
  • Remove the footer
  • Report the knit directory (where the code was executed) in the workflowr report
  • Link to workflowr GitHub repository no longer automatically inserted into navigation bar. Use either wflow_use_github() or wflow_use_gitlab() to insert a link to your workflowr project into the navigation bar

Improved support for caching

A popular knitr/rmarkdown feature is caching slow-running chunks. This can be problematic for workflowr because it assumes that the results are newly created when wflow_publish() publishes the results with a given version of the code. In this release, workflowr now provides warnings, safety checks, and some convience arguments for safely using caching.

  • Include a check in the reproducibility report that reports any existing cached chunks
  • A warning is inserted directly in the HTML file after any code chunk that is cached (cache=TRUE) but is not set to re-run if its upstream chunks are changed (autodep=FALSE)
  • Add argument clean_fig_files to wflow_build(). The default for wflow_build() is FALSE, so that old figures are not removed. This is useful for iterative development when plots from cached chunks may not be regenerated during a build. However, clean_fig_files is fixed to TRUE for wflow_publish() to ensure that the final results are produced during the build (suggested by @lazappi, #113)
  • Add argument delete_cache to wflow_build()/wflow_publish(). The defult is FALSE, but if set to TRUE it will delete the cache directory prior to building each R Markdown file. This helps ensure reproducibility of the published results
  • Have wflow_build() send message about status of cache directory

Documentation

In addition to the new vignette on GitLab, this release has multiple other new vignettes plus updates to existing ones.

  • Add vignette "Sharing common code across analyses" (written with @timtrice, #111, #142)
  • Add vignette "Alternative strategies for deploying workflowr websites"
    • Password-protected site with Amazon S3 (written by @edavidaja, #124)
    • Secure sharing with Beaker Browser (written by @johnsonlab, #59, #65)
  • Update getting started vignette to use new function wflow_use_github()
  • Add section "Style with custom CSS" to customization vignette
  • Add FAQ entry "How can I include external images in my website?"
  • Add code of conduct

Miscellaneous

  • Improve support for hosting on Shiny Server. Setting the option fig_path_ext: false in _workflowr.yml removes the file extension from the figure subdirectories, allowing them to be viewed on Shiny Server (implemented by @Tutuchan, #119, #122)
  • Insert warning into HTML if user configures the knitr chunk option fig.path, which workflowr ignores (idea from @lazappi, #114)
  • Add argument disable_remote to wflow_start(). It creates a Git pre-push hook that disables the ability to push to a remote repository. Useful for confidential projects. Currently only available for Linux and macOS (suggested by @rgayler, #141)
  • Refactor and export the individual functions of wflow_html() to facilitate integrating workflowr features into other R Markdown output formats such as blogdown (suggested by @docmanny, #126)
  • New function wflow_rename() to rename files and directories, including committing the change with Git
  • Use fs internally to improve cross-platform handling of file paths
  • Have wflow_publish()/wflow_git_commit() fail early if any of the files have merge conflicts
  • Switch from rawgit to raw.githack to serve past versions of HTML files (#134)
  • Have wflow_git_push() set the upstream tracking branch by default (see git2r Issue 375)
  • Have wflow_build() report the current working directory. If the knit directory (where the code is executed) is different than the working directory, have wflow_build() report where the code in each file is being executed
  • As usual, some minor bug fixes, improved error handling, and more tests

v1.1.1

5 years ago

Fix unit test on CRAN Windows servers.

See the release notes for v1.1.0 for the recent changes.