R Package Documentation on the Web:
Zelig use case
Current Setup
Homepage
About/
Community/
Etc
Package docs
Current Setup: Workflow
Sphyinx
OpenScholar
I want to update the Homepage/About/etc . . .
I want to update the package docs . . .
Roxygen
Login
Click on widget to edit
Save
Login (GitHub)
Update RestructuredText
Commit
Jenkins build
Login (GitHub)
Update LaTeX
Build
Commit/Push to CRAN
Current Setup: Workflow
Sphyinx
OpenScholar
I want to update the Homepage/About/etc . . .
I want to update the package docs . . .
Roxygen
Login
Click on widget to edit
Save
Login (GitHub)
Update RestructuredText
Commit
Jenkins build
Login (GitHub)
Update LaTeX
Build
Commit/Push to CRAN
Remember to keep in Sync (links, css)
Issues
- Lots of overlapping technologies
- Source is in three places (2 GitHub repos + OpenScholar)
- Reproducible code/previous versions spread over 2 repos, some not reproducible/version controlled
- Difficult for external contributions
Issues
- Lots of overlapping technologies
- Source is in three places (2 GitHub repos + OpenScholar)
- Reproducible code/previous versions spread over 2 repos, some not reproducible/version controlled
- Difficult for external contributions
Causes
- Large error surface -> more errors
- Hard to find error
- Hard to fix errors
- Not practicing what we preach
- Not a model for others
Ways Forward
1. Closest Analoge
2. Most Streamlined
3. Min OpenScholar
4. Max OpenScholar
Possibilities
Key Question
How should Open Scholar fit into a "best practices" software documentation workflow?
(Possible) Decision Criteria
- User friendly
- Reproducible
- Version controlled
- Minimise technology overlap
- Syncing updates across platforms
- Cost effective (open source first)
- Persistant
Closest Analogue
OpenScholar
I want to update the Homepage/About/etc . . .
I want to update the package docs . . .
Roxygen + pkgdown
Login
Click on widget to edit
Save
Login (GitHub)
Update rmarkdown
Build
Commit/Push to CRAN
Remember to keep in Sync (links, css)
Remaining Issues
- Fewer, overlapping technologies
- Source is in two places (1 GitHub repo + OpenScholar)
- Reproducible code/previous versions spread over 1 repo, some not reproducible/version controlled
- Difficult for external contributions
Most Streamlined
I want to change/add anything
IQSSdevtools (Roxygen + pkgdown)
Login (GitHub)
Update rmarkdown
Build
Site hosted on package's GitHub repo in docs/ with IQSS URL
OpenScholar link via Dataverse
Push to GitHub/ CRAN/Dataverse
Benefits
- Minimal number of technologies
- Source is in one place
- Reproducible code
- Full version control
- All docs stay in sync
- Easy for external contributions (pull request of R markdown changes)
- General solution to advocate
Min OpenScholar
I want to change/add anything
Roxygen + osdown
Login (GitHub)
Update rmarkdown
Build
Push to GitHub/ CRAN/Dataverse
Site hosted on package's GitHub repo in docs/ with IQSS URL
OpenScholar link via Dataverse
Min OpenScholar
I want to change/add anything
Roxygen + osdown
Login (GitHub)
Update rmarkdown
Build
Push to GitHub/ CRAN/Dataverse
osdown [OpenScholar-down]:
a fork of pkgdown with Open Scholar branding/best practices
Benefits
- Minimal number of technologies
- Source is in one place
- Reproducible code
- Full version control
- All docs stay in sync
- Easy for external contributions (pull request of rmarkdown)
- Promotes OpenScholar
Costs
R + web development resources needed to create osdown
Max OpenScholar
I want to change/add anything
Roxygen
Login (GitHub)
Update rmarkdown
Build
New OpenScholar Capability: Open Scholar listens to GitHub for changes to HTML files and renders the result using OpenScholar templates.
Push to GitHub/ CRAN/Dataverse
Benefits
- Small number of technologies for users
- Content source is in one place
- Reproducible content code
- Full version control of content
- All docs stay in sync
- Easy for external contributions (pull request of rmarkdown)
- Promotes Open Scholar
Costs
Potentially very resource intensive to setup and maintain
- Is OpenScholar WYSIWYG possible?
- If so, would require GitHub integration, becomes less general.
1. Closest Analoge
2. Most Streamlined
3. Min OpenScholar
4. Max OpenScholar
Possibilities
r package docs on the web
By Christopher Gandrud
r package docs on the web
- 1,265