Creating packages based on the previous workshop for this user group.

Release process based on the useR!2021 talk

Maintaining packages based on the archived packages files.

Most common usage

Less used due to restrictions on size on CRAN

In my opinion the most important aspect.

Resources used to create this workshop

Maëlle post: how to write good packages is also recommended.

Name of the package, description of the package, maintainer, relationships with other packages... Language does not need to be on English, it can be in other languages

This can be done manually no need to do it via other tools (but recommended)

Important file!

In Rstudio you can insert the skeleton with Ctrl+Alt+Shift+R . In Rstudio you can convert the special comments into documentation with Ctrl+Shift+D .

Articles are like vignettes but not checked by R CMD check

With testthat, tinytest or other ways

Sites with pkgdown or other tools.

C++, C, Fortran or linking to other languages, Java, perl; or other programs (cURl, xmldev)

How to choose a name of the package.
How to write a good README and NEWS file. Set up continuous integration tests. How to pick up a license.

What does it mean? Building R CMD build and making it accessible to other people, example: install.packages("http://biodev.cea.fr/sgcca/gliomaData_0.4.tar.gz", repos = NULL)

What are the expectations?

Repositories keep versions of code

Archive do not delete content (normally) Keep versions "released". Only CRAN, Bioconductor and additional repositories like r-universe, drat, ... work via install.packages.

Able to combine them, not mutually-exclusive

Submissions are though specially if coming from places with poor training Lack of confidence/experience with reviews.

Differences in objectives but all looking for quality CRAN: Point errors, comments Bioconductor: In detail comment of style, classes, dependencies, structure… rOpenSci: guideline for reviewers (about style, tests, functions, description, documentation, …)

CRAN ~18000 packages, Bioconductor ~2000, rOpenSci ~300 To work with this slides use xaringan::infinite_moon_reader()

The different projects/archives have different setups. Read the table All of them first you need to pass the automatic checks in place before a human looks into it. Will use data from the three projects but mostly refer to CRAN.

One order of magnitude of difference between each other CRAN > Bioconductor > rOpenSci Many variability on month Also very few data collected from CRAN so far (Also there are some hiccups on CRAN collection, near the end of May the CRON job stopped working for a week. )

Many folders but these two are the most important. There isn't an explanation from CRAN about how do they work. Pretest is resubmission (newer versions of packages) and also for newbies

Submit when you are ready, better on the queue than outside.

Median time on submissions ~9 hours, mean time ~31.4254531 hours. 1, 2, 9, 31.4254531, 33, 2365

Different time, can be shorter or longer. Most longer need resubmission. Resubmit with different version (makes it easier to track how many are).

## # A tibble: 2 × 2
##   new     time
##   <chr>  <dbl>
## 1 New       58
## 2 Update     5

CRAN: 60h Bioconductor: most of them in 1 month rOpenSci: in 2 months (seeking 2 reviewers and posting them).

Bioconductor reviewers do a lot rOpenSci editors too Both organizations have a group of users involved on the package review system. Even if Bioconductor doesn't explicitly ask for reviewers from the community. Bioconductor are considering now how to improve the review system. Omitted bots bioc-issue-bot and ropensci-review-bot (new March 2021).

Non reviewers users on bioconductor still chime in to help.

Bot provides feedback of many issues and actions performed. It can be changed/adapted to change in requirements or errors. rOpenSci is going to have a bot too ropensci-review-bot.

On bioconductor most problems with the submissions are not the package itself but not replying or chosing another venue. rOpenSci provides more detailed questioning for scope of a package.

Bioconductor & rOpenSci 50%, some submissions are abandoned or do not fit the project. Different problems faced by new packages and older ones. More in depth review requires 1 month for each reviewer.

## # A tibble: 50 × 7
##    submission_n new    Accepted     n  perc suspended perc_suspended
##    <fct>        <chr>  <lgl>    <int> <dbl>     <dbl>          <dbl>
##  1 1            New    FALSE      652 17.8        652          31.0 
##  2 1            New    TRUE      3015 82.2          0           0   
##  3 1            Update FALSE      691 13.4        691          32.8 
##  4 1            Update TRUE      4456 86.6          0           0   
##  5 2            New    FALSE      148 23.4        148           7.03
##  6 2            New    TRUE       485 76.6          0           0   
##  7 2            Update FALSE      270  8.27       270          12.8 
##  8 2            Update TRUE      2993 91.7          0           0   
##  9 3            New    FALSE       41 20.4         41           1.95
## 10 3            New    TRUE       160 79.6          0           0   
## # … with 40 more rows

## # A tibble: 2 × 3
##   Approved     n  perc
##   <chr>    <int> <dbl>
## 1 No        1233 0.505
## 2 Yes       1208 0.495

## # A tibble: 1 × 4
##   `1. awaiting moderation` `2. review in progress` `3a. accepted` `3b. declined`
##                      <dbl>                   <dbl>          <dbl>          <dbl>
## 1                0.0000231                   0.299           35.7           19.1

Follow the detailed guidelines from Bioconductor and rOpenSci. Fix any problem that you haven't detected previously (double check the CRAN repository policy). Resubmit

Not in your hands to control, reactive work

Better quality => Less work

Check also your dependencies and be mindful to users

Be mindful that you now "need" to support this.

Deprecate and defunct are important steps to notify users.
Be mindful also to other developers that depend on your package: give them time, hear their concerns...

Thank also to the package authors (mainly tidyverse, ggplot2 and rhub, and gh). Maëlle Salmon and Stephanie Locke for the CRAN dashboard. And the organization. rOpenSci review: Video

## NULL