Lazy State: Maybe use @ instead of $

[josephjclark]

The Lazy State operator uses $ to read from state. For example:

get($.url)

Some benefits of using $ are:

• It is valid JS syntax, so get($.url) will run in plain JS and not freak out any linters/validators
• It is like JSON path, which uses $ for root. And JSON path is sort of the model it's based on
• It's also a bit like a jquery selector, for people old enough to remember jquery
• $ is often used as a special variable in JS, devs are kind of used to it

Problems with $ are:

• Yes, the code will compile and even run - but you'll get type errors in native JS. Job code isn't really executable until it's been put through the compiler.
• It's really really ugly in template literals because of the double dollar, ie:

get(`resource/${$.data.id}`)

Benefits of using @ are:

• @ is a symbol related do addressing generally, which fits nicely with @path.to.state
• @ is is not valid JS, helping openfn code stand out as distinct from JS.
• It's stronger branding and more distinct than dollar
• It plays well with the compiler
• Works great in template literals
• It's quite easy to explain
• We can keep $ as an alias for the short-medium term

But the problems with @ are:

• Source editors like VSCode and Monaco (and so Lightning) won't recognise @ and will complain. Yes we can add support in Lightning an extend the syntax, but those using third party tools for local dev will have a worse experience
• We've already launched $ and are using it in production code. I don't want to re-write everything and retrain now! On the other hand, it IS still new and flagged as experimental, so maybe there's still time for a dramatic change

I think the bottom line with @ is that it's better branding and basically good for us, but it moves the openfn DSL a step closer to being a distinct formal language, meaning we need to support .ojs files and create OpenFn JavasScript language definitions.

To be honest actually $ is no different - it SHOULD be treated as a different language statement, with linting rules and syntax highlighting associated with it. And $ won't run without compilation anyway.

So maybe introducing @ just forces us to be more honest about our coding language.

However, I don't want to do all that stuff if we're going to throw it away next year in favour of a v3 langauge syntax, as is being discussed in the dark corridors of power here...

[josephjclark]

Petition to use .ofn files instead of .ojs files 🤔