# 1 Introduction

This package produces AnVIL workspaces from R packages. An example uses the new Gen3 package as a basis for the Bioconductor-Package-Gen3 workspace (permission to access this workspace is required, but there are no restrictions on granting permission).

## 1.1 Package installation

If necessary, install the AnVILPublish library

if (!"AnVILPublish" %in% rownames(installed.packages()))
BiocManager::install("AnVILPublish")

There are only a small number of functions in the package; it is likely best practice to invoke these using AnVILPublish::...() rather than attaching the package to the search path.

## 1.2 The gcloud SDK

It is necessary to have the gcloud SDK available to copy notebook files to the workspace. Test availability with

AnVIL::gcloud_exists()

and verify that the account and project are appropriate (consistent with AnVIL credentials) for use with AnVIL

AnVIL::gcloud_account()
AnVIL::gcloud_project()

Note that these be used to set, as well as interrogate, the acount and project.

## 1.3notedown software

Conversion of .Rmd vignettes to .ipynb notebooks uses notedown python software. It must be available from within R, e.g.,

system2("notedown", "--version")

# 2 Creating or updating workspaces

CAUTION updating an existing workspace will replace existing content in a way that cannot be undone – you will lose content!

Workspace creation or update uses information from the DESCRIPTION file, and from the YAML metadata at the top of vignettes. It is therefore worth-while to make sure this information is accurate.

In the DESCRIPTION file, the Title, Version, Authors@R (preferred) or Author / Maintainer fields, Description, and License fields are used.

In vignettes, the title: and author: name: fields are used; the abstract is a good candidate for future inclusion.

## 2.1 From package source

The one-stop route is to create a workspace from the package source (e.g., github checkout) directory use as_workspace().

AnVILPublish::as_workspace(
"path/to/package",
"bioconductor-rpci-anvil",     # i.e., billing account
create = TRUE                  # use update = TRUE for an existing workspace
)

Use create = TRUE to create a new workspace. Use update = TRUE to update (and potentially overwrite) an existing workspace. One of create and update must be TRUE. The command illustrated above does not specify the name = argument, so creates or updates a workspace "Bioconductor-Package-<pkgname>, where <pkgname> is the name of the package read from the DESCRIPTION file; provide an explicit name to create or update an arbitrary workspace. The option use_readme = TRUE appends a README.md file to the formatted content of DESCRIPTION file.

AnVILPublish::as_workspace() invokes as_notebook() so this step does not need to be performed ‘by hand’.

See the command add_access(), below, to make the workspace available to a wider audience.

## 2.2 From collections of Rmd files

Some R resources, e.g., [bookdown][] sites, are not in packages. These can be processed to workspaces with minor modifications.

1. Add a standard DESCRIPTION file (e.g., use_this::use_description()) to the directory containing the .Rmd files.

2. Use the Package: field to provide a one-word identifier (e.g., Package: Bioc2020_CNV) for your material. Add a key-value pair Type: Workshop or similar. The Pacakge: and Type: fields will be used to create the workspace name as, in the example here, Bioconductor-Workshop-Bioc2020_CNV.

3. Add a ‘yaml’ chunk to the top of each .Rmd file, if not already present, including the title and (optionally) name information, e.g.,

---
title: "01. Introduction to the workshop"
author:
- name: Iman Author
- name: Imanother Author
---

Publish the resources with

AnVILPublish::as_workspace(
"path/to/directory",      # directory containing DESCRIPTION file
"bioconductor-rpci-anvil",
create = TRUE
)

# 3 Updating notebooks or workspace permissions

These steps are performed automatically by as_workspace(), but may be useful when developing a new workspace or revising existing workspaces.

## 3.1 Updating workspace notebooks from vignettes

Transforming vignettes to notebooks may require several iterations, and is available as a separate operation. Use update = FALSE to create local copies for preview.

AnVIL::Publish::as_notebook(
"paths/to/files.Rmd",
"bioconductor-rpci-anvil",     # i.e., billing account
"Bioconductor-Package-Foo",    # Workspace name
update = FALSE                 # make notebooks, but do not update workspace
)

The vignette transformation process has several limitations. Only .Rmd vignettes are supported. Currently, the vignette is transformed first to a markdown document using the rmarkdown command render(..., md_document()). The markdown document is then translated to python notebook using notedown.

It is likely that some of the limitations of vignette rendering can be reduced.

## 3.2 Adding user access credentiials to share the notebook

The "Bioconductor_User" group can be added to the entities that can see the workspace. AnVIL users wishing to view the workspace should be added to the Bioconductor_User group, rather than to the workspace directly. To add the user group, use

AnVILPublish::add_access(
"bioconductor-rpci-anvil",
"Bioconductor-Package-Foo"
)

# 4 Vignette and .Rmd best practices

## 4.1 Orientation

.Rmd files need to be converted to jupyter notebooks. Currently there is not an ‘ideal’ solution, with details listed in the ‘Additional notes…’ section. Consequently, there are ‘best practices’ that lead to results that are more likely to be satisfactory, as outlined here.

## 4.2 Best practices

1. For packages, make sure the DESCRIPTION file is complete. Use the Authors@R notation for fully specifying authors. Add a Date: field indicating date of last modification. Follow other Bioconductor best practices, e.g., using and incrementing appropriate version numbers.

2. For collections of vignettes not in a package (e.g., a bookdown folder), add a DESCRIPTION file at the top level. An example is

Package: BCC2020
Type: Workshop
Title: R / Bioconductor in the AnVIL Cloud
Version: 1.0.0
Authors@R:
c(person(
given = "Martin",
family = "Morgan",
role = c("aut", "cre"),
email = "Martin.Morgan@RoswellPark.org",
comment = c(ORCID = "0000-0002-5874-8148")
),
person("Nitesh", "Turaga", role = "ctb"),
person("Lori", "Shepherd", role = "ctb"))
Description:
This book contains material for a 2 1/2 hour course offered at the
Bioinformatics Community Conference 2020. Bioconductor provides
more than 1900 R packages for the analysis and comprehension of
high-throughput genomic data. Most users install and run
Bioconductor on a personal computer or perhaps use an academic
cluster. Cloud-based solutions are increasing appealing, removing
better, scalable computing resources; and (b) large-scale
'consortium' and other reference data sets. This session
introduces the AnVIL cloud computing environment. We cover use of
the cloud as a replacement to desktop-style computing; integrating
workflows for 'upstream' processing of large data resources with
interactive 'downstream' analysis and comprehension, using Human
Cell Atlas single-cell datasets as an example; and querying
cloud-based consortium data for integration with a users own data
sets.
Date: 2020-07-17
Encoding: UTF-8
LazyData: true
Roxygen: list(markdown = TRUE)
RoxygenNote: 7.1.1

The Type and Package fields are used to construct the second and third elements of the workspace name (in this case, Bioconductor-Workshop-BCC2020). Title, Version, Authors@R, Description, License, and Date fields are used to construct the DASHBOARD page.

3. Start each vignette with ‘yaml’ containing essential metadata about the document – title and author(s). Include other information if desired, e.g., abstract, (static) date of last modification.

4. Use a file naming system AND a yaml title field that sorts files into the order in which the document content is to be presented, e.g., using file names 01-Setup.Rmd, 02-... and titles (in the yaml) title: "01 Setup", … Naming both files and titles in this way provides some chance that the Rmd files are presented, or can be made to be presented, sensibly across the Bioconductor package landing page and Workspace / NOTEBOOK interface.

5. All code chunks, regardless of annotations such as eval = FALSE or echo = FALSE are converted to visible, evaluated cells in jupyter notebooks. Replace code chunks that you do not wish the user to evaluate with HTML tags <pre></pre>.

6. Although both Rmarkdown and python notebooks support code chunks in multiple languages, there is no support for this in the conversion process – all cells are presented as R code.

## 4.3 Additional notes on .Rmd conversion

The current state of affairs with respect to notebook conversion is imperfect. Conversion is currently a two-step process: Rmarkdown to markdown, and markdown to ipynb.

• The conversion from Rmarkdown to markdown is currently accomplished with

knitr::opts_chunk\$set(eval=FALSE)
rmarkdown::render(..., md_document())

to create a markdown document from the .Rmd source.

This correctly processes the markdown content, including yaml metadata, but renders all code chunks identically.

Using other knitr options may allow, e.g., conditional inclusion of code chunks.

• Use notedown to convert from markdown to jupyter notebook, adding metadata to indicate that the notebook has an R kernel.

Here are some notes on alternative solutions.

• jupytext (version 1.5.1) but has difficulty with some markdown. For instance, reference-style links [foo][1] are only rendered correctly when the reference is in the same code chunk as the link. It is under active development and may mature into a possible alternative.

• pandoc (version 2.10.1) provides a one-step conversion from .Rmd to .ipynb, but code chunks are rendered as pre-formatted text rather than evaluable cell.

• notedown (version 1.5.1) also provides one-step conversion, but does not exclude yaml from vignettes. The project has not had commits for several years, and has several open issues.

# Session info

sessionInfo()
## R Under development (unstable) (2021-03-18 r80099)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 20.04.2 LTS
##
## Matrix products: default
## BLAS:   /home/biocbuild/bbs-3.13-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.13-bioc/R/lib/libRlapack.so
##
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=C
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base
##
## other attached packages:
## [1] BiocStyle_2.19.1
##
## loaded via a namespace (and not attached):
##  [1] bookdown_0.21       digest_0.6.27       R6_2.5.0
##  [4] jsonlite_1.7.2      magrittr_2.0.1      evaluate_0.14
##  [7] stringi_1.5.3       rlang_0.4.10        jquerylib_0.1.3
## [10] bslib_0.2.4         rmarkdown_2.7       tools_4.1.0
## [13] stringr_1.4.0       xfun_0.22           yaml_2.2.1
## [16] compiler_4.1.0      BiocManager_1.30.10 htmltools_0.5.1.1
## [19] knitr_1.31          sass_0.3.1