WIP: typed arguments

[garthk]

Work-in-progress on typespecs for the measurements and metadata. Would fix #5.

[garthk]

The code is a mess, but I like the results.

  @typedoc "Measurements we send via `:telemetry.execute/3`"
  @type measurements :: %{required(:count) => pos_integer(), optional(:ms) => pos_integer()}

  @typedoc "Metadata we send via `:telemetry.execute/3`"
  @type metadata :: %{required(:events) => list(Event.t()), optional(:payload) => binary()}

  defevent([:honeycomb, :start],
    measurements: measurements(),
    metadata: metadata(),
  )

→

I think the easiest way to make this clean is to:

• Remove the t_ from t_measurements and t_metadata
• Remove defaults
• Emit function_name/0 if we don't get a measurements or a metadata
• Emit function_name/1 if we get a measurements but no metadata
• Emit function_name/2 if we get both
• Error out if we get metadata only

[garthk]

Yeah, that feels better. As I say in the commit:

Determining the arity of the emitted function from the presence or of measurements and metadata might feel like we were "forced" to use typespecs, but map isn't so bad...

Reading my docs output as if I were consuming my own package, I liked how the lack of a \\ %{} made it clear to me that I'm not going to get any measurements.

Sorry for dragging #1 into this; I just noticed I'm using the :telemetry terms by default despite that not having been settled.

[garthk]

Meanwhile, I'm falling out of love with that defstruct pattern for options. I like how Kernel.struct!/2 blows up when the user gets a key wrong, but prefer the typings for keyword lists otherwise.

[garthk]

This is obsolete given your 0.2.0, no?

[coryodaniel]

I believe so, but I haven’t tested it with require/1, optional/1 on maps yet.

Sorry for the delay, been blowing through this project and a few others to tighten up some stuff to release a work project OS I’ve been working on.