Trailblazer DX - 2023

Some notes on the sessions from the Trailblazer DX conference in San Fransisco, CA in March 2023.

Keynote

• Einstein GPT
  • Takes in customer data
  • Use Salesforce Model, your model, or any other model
  • Einstein Assistant - Chat inside of Sales Cloud that you can use to ask different questions about your customers, including public data
  • Can be embedded in Slack
  • Can be embedded in Tableau
  • Can be used in Service Cloud for assistance in Chat sessions
  • Also available in VS Code like a “GitHub Copilot”-like feature
• Data Cloud
  • Foundation of the correct means for getting AI to work for you
    • If your data is bad, your AI will be bad
  • DataLake Object
    • Can store trillions of records
    • Can create thousands of these types of objects
  • Data Pipelines: The ETL tool (“no ETL?”) for Data Cloud
• Flow Updates
  • HTTP Callouts in Flows
  • Case Classification Flows
  • Mulesoft Composer + Flow for pre-built actions
  • Reactive Flow Components
    • Enhanced Screen flow components
  • Flow Orchestration on Mobile
• Slack - GA in Summer ‘23
  • Workflow builder

Apex Best Practices

• Consider using Platform Cache for speeding up long-running processes
• Platform Event delivery is not guaranteed - Consider moving certain client workflows to centralized Async handlers
  • Though in this next year we’ll have the ability to pass a callback class that will be executed upon successful or failed attempts to publish an event to the bus

DataWeave

• Functional programming language for use in processing records of one type and going into another type
• Created by MuleSoft
• Language / CLI can be downloaded and is open source
• VS Code extension allows for testing DW scripts with samples
• Could be used for loading large amounts of data via Apex
  • Can call to a DW script, passing in data from Apex
  • Can output SObjects
• Should be able to deploy DW scripts as part of the metadata api and store / invoke them in SFDC

CumulousCI Suite

• Problems that it solves:
  • Reduces Org Setup Cost
  • Improves Org Reliability
  • Helps with conflicts
• Helps to bootstrap orgs for development
• Steps to use it
  • Document exactly what you need in order to test and develop in the platform
    • Get exact steps as a setup guide
  • Take those steps and convert them to tasks in a YAML file that CumulusCI can ingest and run
• Tons of steps are pre-built and can be leveraged
• Can work by creating scratch orgs, or can be used in a new Dev sandbox
• Based on SFDX
• Free and OSS
• MetaDeploy
  • Just like CCI CLI but in a web-based platform

Resources

• Trailhead modules on CumulousCI
• sfdo-community-sprints.github.io

Slack Keynote

• ChatGPT app inside of slack
  • Beta open Today
• Use Cases
  • Summarize threads / channels
  • DM it for answering questions, about public information or information found in Slack
  • Draft messages
• Screen Flows can be run in Slack

2GP Packaging Updates

• Benefits of 2GP
  • Source-driving development
  • API/CLI-first deployment model
  • Multiple packages per namespace
    • Make sure you use namespace accessible for your Apex
  • Flexible versioning model
  • Declarative dependencies
  • Super-simple Patch Versioning
• 2023 Focus Areas
  • Making packaging operations faster
  • Enable everyone to move to 2GP by delivering on Package Migrations
• Package Performance Roadmap
  • See picture that I took of the roadmap
• Package Migrations (not GA yet)
  • Fully dedicated team to this effort
  • Package Conversion vs. Migration
    • Conversion: Convert 1GP package to 2GP
      • force:package:convert
    • Migration: Subscribers are migrated to 2GP version from 1GP version
  • What does this look like when GA? See picture that I took of process
• What does Conversion look like before GA?
  • Developer preview open to all partners available today
  • Roadmap: See the picture that I took of the roadmap
  • How to sign up and get more documentation? See the picture I took of the signup slide
• Main takeaways
  • Realistically it’s looking like full 1GP -> 2GP packaging won’t be available till at least Spring ‘24, so for Zingtree we just need to focus on tightening up on deployment automations / using CCI for better controls while continuing to use 1GP
  • Should still plan on using 2GP tools for Zingtree like Scratch Orgs, CLI, code, etc…
  • I need to write up SLDC for working on Managed Packages

AI Code Generation w/ CodeGen & Einstein

• Ananya is the PM for IDE interactions
• Main pain points today
  • See compilation errors without having to deploy Apex
  • Code profiling is limited and difficult
  • Too much Boilerplate to create, and lack of tools to auto-generate
• LLM used: CodeGen
  • 16B parameters
  • Largest Open Source LLM for Code
• Einstein GPT for Devs and Admins
  • Can generate code
    • Github Copilot-like experience
    • Can also have chat interface directly in VS Code to chat more about code
  • Can generate flows?
• Replit’s GhostWriter built on CodeGen
• Roadmap
  • Developer preview (I signed up, but don’t know really what that entails)
  • Invite-only Pilot to hopefully come out at Dreamforce this year
  • GA not even planned yet

Custom Property Editors in LWC

• Ability to create LWCs that would act as more customized property editors for another LWC
  • See pics of slides
• Also there are more OOTB types for more rich types / properties
  • Like Dates with a date picker
  • See pics of slides
• Can create custom property types
  • Defined in experiencePropertyTypeBundle metadata type
  • Series of JSON files, using JSON Schema formatting
  • Includes validation and display rules
  • See pics of slides
• Can use Accordions or Tabs for better organization of properties for an LWC
  • See pics of slides
• Roadmap
  • Available as a dev preview for Experience Cloud sites
  • Summer ‘23 - 2GP Packaging Support
  • Winter ‘24 - Experience Builder GA
  • Spring ‘24 - Conditional Properties / Lightning App Builder Dev Preview

Async vs. Sync Architectures

• Record-Triggered Flows can have an async path
• Limit Tracking
  • Apex and Flows track Async processing (250k maximum) the same, but independently
  • Remember that you should design the architecture to allow for things to fail and not be run
• Usually async actions won’t run immediately: be prepared for 20mins of latency of Async processing on a bad day
• Especially when you’re running async processes, really consider error handling and logging frameworks
• General recommendations / last thoughts
  • Don’t overuse Batch Apex, there are a lot of downsides if you don’t use it for its appropriate use case
  • Queueuable Apex is much better than @future methods; don’t bother using them any more
  • Use Async Record-triggered flows instead of 0-wait pause elements
  • Make sure you follow the correct order of general actions in async flows
    • Fetch all records
    • Callouts to external systems
    • Prepare records
    • CUD on records in the database
    • Check out “Rollback Node” in Flow for longer Screen flows with more advanced transaction management

DevOps Center + CLI

• DevOps Center can be great for teams that are hybrid (programmatic and declarative developers)
  • DOC will keep track of all changes made to branches and show them in an easy-to-digest format
• For programmatic developers, you don’t really have to work in DevOps center, you just need to start the work item from DevOps center
  • It’ll create a new feature branch for you, and then you can get started using VS Code and git / Github
  • “Environments Out of Sync” - What you’ll see when merging via Github. Basically means that it’s in the branch but not in the actual org
    • Could have Github action that auto-deploys on merge to a branch

Random Learnings from Booths

VS Code Booth

• Use the Apex replay debugger, it’s SO much better than using System.debug() statements
• They’re easy to use, and don’t require any other org setup, just set a checkpoint and run it
  • Difference between breakpoints vs. checkpoints
  • Use checkpoints instead
  • Limited to 5 checkpoints at any given time across your org

LWC Booth

CLI Booth

• sf CLI is already installed with sfdx CLI
• Can start using it if we want, no big rush right now

API Booth

• Use the Composite API (tree api) for creating records with nesting
• SOQL will allow for sub queries up to 5 levels deep soon (within the next release or so)
• Eventually (date TBD) there will be friendly URLs that allow for querying data across relationships, up to 5 levels deep
  • Ex. /v55.0/data/Account/001a32jfasldkjdsf/Contacts/003afjssdfljk23sflkjds9f/Cases/...

Future of Apex

• Session of the Apex roadmap
  • See picture of the roadmap
• WITH USER MODE - Better version of WITH SECURITY ENFORCED
• DataWeave support with Apex
• In a future release, Set will implement Iterable interface

Security

• Apex runs in System Mode
  • Can do whatever the highest privilege user in that org can do, full stop
• stripInaccessible()
  • See pic of slide
  • Still a really good thing to use
• WITH SECURITY_ENFORCED - didn’t age well
  • Can only do FLS/CRUD checks
  • Can’t work with restriction rules
  • Doesn’t apply to WHERE clause
  • Doesn’t work well with polymorphic lookups
  • Will likely be version this out
• User-Mode DB Ops - Available as of Spring ‘23
  • See pics of slides
  • CRUD as users
  • insert as user
  • Legitimately forces Apex to drop out of System mode directly into User mode, and then runs DB operations, then comes back when done
    • Will scale much better as the platform evolves
• User-Mode DB Ops Future
  • Allow for Developer-based Permission Sets that allow for User-Mode + additional permission sets
    • User doesn’t need to have permission set applied
  • See pics of slide for potential syntax, though could change
  • Potentially coming before EOY

Quality of Life (DX)

• Dynamic Bind Variables in SOQL (See slide for example)
• Label translation - GA in Summer ‘23
• Apex-based Sorting to become easier - No date yet
• Interface Implementer Discovery - GA now
  • Used largely for dynamic code generation, searching for all Apex classes that conform to an interface
  • I believe that this would be a good replacement for the Custom Metadata-driven centralized Trigger or Async frameworks shown in Advanced Apex Programming

Async Updates

• Controlling the job rate of Queueable Apex- GA now
  • Define a delay in when you’d like your Queueable job to run, up to 10 minutes
  • Can also choose to use 0 minutes as the option, meaning run as fast as possible
    • Be careful with this
• Ability to have more dynamic chained job protection (for when you’re in a dev org, e.g.) - Coming Soon
  • See slide for example, but syntax could change
• Deduplication control - Coming Later
  • Allow for setting a “Deduplication Key” when you enqueue a Queueable job to avoid from people enqueueing the same job accidentally
  • See Slide for example
• Extended Query Locators - Coming Later
  • Allows for Queueable to be similar to Batch Apex in pagination without a bunch of pagination custom logic
  • Would like to basically allow for Queueable to replace Batch Apex in the future
  • See slide for example syntax, though could change before GA

DataWeave and Apex

• Coming in Summer ‘23 as a Beta
  • But it’s available everywhere now!
• DON’T need to be a MuleSoft customer to use DataWeave in Apex

Generics in Apex

• Why?
  • To add better tools for dealing with bulkification / performance tuning of your Apex
• Coming in the future, no date specifically available
• Still gathering feedback on how to implement based on use cases and needs

Other Related Topics

• Salesforce
• Development
• Technology