fhir-gen: valuesets can get rate limited

[josephjclark]

Hit quite an annoying problem in valueset generation.

During generationm we pull value set definitions down for core datatypes from hl7.org. There's a lot of data here.

What happens after we hit these endpoints too quickly is the server starts feeding us captchas - I'm pretty sure it's a rate limit. Which means requests in the generator start failing.

It would be far more effective to download the whole base spec as a zip and extract the datasets we need. But that's a bunch of work, too.

[josephjclark]

This is a blocker to #1545

I was hoping for eswatini we could just pull what we need, but I'm having to download the whole world here

[josephjclark]

Another angle on this is that a lot of the valuesets being downloaded don't seem to actually be used. If I was smarter about which we downloaded, would this work better?

EDIT: hmm, actually maybe all those value sets really are used...

[josephjclark]

I think my favourite solution here is to download all this into the fhir-4 adaptor (downloading the values from the base spec json), then import them into fhir-eswatini as needed.

I will need to do something about the duplication, and I'm still worried about the overall bundle size. But I think we've gotta give it a shot.

When it's done once for fhir-4, any other adaptor can easily inherit those values