Art of the Feature Toggle

Exploring the use of feature toggles as a deployment and release pattern for shiny applications

Transcript

1. Art of the Feature Toggle Patterns for maintaining and improving Shiny applications over time
2. Solutions Engineering at RStudio
3. Shiny in Production Workshop - 2019 Bookdown: Shiny in Production (RStudio Conference 2019) Chapter 12 - DevOps Philosophy and Tooling - “Production” has many meanings - Are there lessons can be learned from the DevOps community?
4. Introducing new features over time Shiny Applications in Production
5. Adding new features to Shiny in Production Version 1 Version 2
6. Adding new features to Shiny in Production Version 1 Version 2
7. Adding new features to Shiny in Production Version 1 Version 2
8. Version 1 Version 2 Does pushing a code change feel risky?
9. DevOps Learning: Decouple deployment from release • Deployment is any push of code to an environment (test, prod) • Release is when that code (feature) is made available to users or customers Deployment on demand and thoughtful release strategies allow more control (and more success) over the delivery of features to end users. - Application-based release patterns (today!) - Environment-based release patterns (tomorrow)
10. Application-Based Release Patterns
11. Feature Toggle A mechanism to selectively enable and disable features, or control which features are visible to specific user segments. Enable dark launch: deploy a change to production and then perform testing while it’s invisible to most users. - Modify the code to make calls to new functions, log results without displaying - Have 1% of users make invisible calls to new feature, fix issues, progressively increase users to test production load
12. Starting from the Gallery
13. “Personalized Data Access” session$user
14. Feature Toggles Shiny applications can access the username and groups of the current user through the session parameter of the shinyServer function. Your application could use this information to display customized messages or to enable functionality for a specific subset of users.
15. Implementation
16. Implementation UI built through conditional session$user tests within server-side output objects output$subtitle output$salesPlot output$salesTbl
17. Dynamic UI Feature Toggles
18. The conditionalPanel function The conditionalPanel function, is used in ui.R and wraps a set of UI elements that need to be dynamically shown/hidden. Creates a panel that shows and hides its contents depending on the value of a JavaScript expression, usually input-based. This example shows how to define an output variable in the server code that you can use in the UI. Output condition example: Dean Attali session$user session$groups
19. conditionalPanel( condition = "output.isBeta == false", … UI Elements) conditionalPanel( condition = "output.isBeta == true", … UI Elements) output$isBeta <- reactive({ [Test for group membership] })
20. Feature Toggles in Production? Will I (personally) use feature toggles for Shiny application deployment and releases? No, not likely. - Not easy to manage, automate, test and ultimately “toggle” - Alternative: “Environment-Based Release Patterns” - Still a cool concept, useful for solving other types of problems - RStudio Connect Feature Hacks!
21. Mixing Access Controls & Feature Toggles on RStudio Connect
22. Content Access Control Options on RStudio Connect Access Control Setting Description Anyone, no login required Allows for anonymous viewer access All users, login required Everyone with an RStudio Connect account Specific users or groups Only users or groups you specify can access Collaborators & you Only you and your collaborators can access
23. People > Group Management UI in RStudio Connect (Admin Tools) Solutions Engineering group setup
24. Problem: Can you add me as a collaborator on this? 1. Self service code access - install.packages( "gitlink") 2. Slack Incoming Webhook integration If session$groups != solutions { Display an action button that will send me a slack alert when clicked. }
25. “Slack me if you need access” Problem: Can you add me as a collaborator on this? 1. Self service code access - install.packages( "gitlink") 2. Slack Incoming Webhook integration If session$groups != solutions { Display an action button that will send me a slack alert when clicked. }
26. “Slack me if you need (collaborator) access”
27. Action Button - Command Pattern
28. Recommendation: Create a Slack App There are a number of R packages and legacy methods you could use to talk to slack from your R code. I recommend creating a Slack App. Send data to Slack in real-time - Incoming Webhook Creating an Incoming Webhook gives you a unique URL to which you send a JSON payload with the message text and some options.
30. “Slack _me_ if you need access” - Find your slack member ID - Construct your message - Send JSON payload to Webhook URL
31. Action Button - Command Pattern message-handler.js server.R ui.R
32. Access Control Setting Group or User Conditional Solution All users, login required Group isSolutions() FALSE Element for requesting collaborator access
33. Left: Publisher view with access controls on RStudio Connect Signed in as Kelly, who is a member of solutions No button displayed Right: Vanity URL app access by a different user User is not a member of the solutions group Request Access button displayed Watch on YouTube
34. Access Control Setting Group or User Conditional Solution Anyone, no login required is.null(user()) TRUE Custom UI for anonymous viewers More Hacks for session$user Create a custom landing page for anonymous visitors and logged out users
35. Left: Publisher view with access controls on RStudio Connect Right: Custom UI for anonymous and logged-out visitors to the app Watch on YouTube
36. Code: github.com/kellobri/shiny-feature-toggle Slides: bit.ly/shiny-feature-toggle - solutions.rstudio.com - community.rstudio.com - kelly@rstudio.com