FHIR Chat · Canonical with version

Hey there everybody, we'd like to get some user input on a syntax decision. We are adding a Canonical keyword to SUSHI which allows the user to get the canonical URL of a FHIR resource, but FHIR allows canonical URLs to include a version by appending |<version> to the URL. For example, if we have a resource Foo with URL http://example.org/ValueSet/Foo and version 1.2.3, the URL as a canonical could be http://example.org/ValueSet/Foo|1.2.3. We are unsure how to best handle versions. There seem to be two paths forward:

Version included by default :

Canonical(Foo) is used to indicate that SUSHI should use the version it finds on the Foo resource, returns http://example.org/ValueSet/Foo|1.2.3
Canonical(Foo|4.5.6) is used when you want to manually specify a version different from the one that SUSHI finds on Foo, returns http://example.org/ValueSet/Foo|4.5.6
Canonical(Foo|) is used when you do not want the version, returns http://example.org/ValueSet/Foo

Version not included by default:

Canonical(Foo|.) is used to indicate that SUSHI should use the version it finds on the Foo resource, returns http://example.org/ValueSet/Foo|1.2.3.
Canonical(Foo|4.5.6) is used when you want to manually specify a version different from the one that SUSHI finds on Foo, returns http://example.org/ValueSet/Foo|4.5.6
Canonical(Foo) is used when you do not want the version, returns http://example.org/ValueSet/Foo

Chris Moesel (Jul 22 2020 at 15:11):

I think there are advantages to each, but in a FHIR canonical, no | means no specified version -- so I have concerns about making no | mean insert the current version in FSH.

David Pyke (Jul 22 2020 at 15:48):

I think Canonical(Foo) should return http://example.org/ValueSet/Foo
Canonical(Foo | x.y.z ) should return http://example.org/ValueSet/Foo|x.y.z
Anything else is a syntax error. That keeps it simple to use

Chris Moesel (Jul 22 2020 at 16:32):

So @David Pyke -- I want to be sure I understand what you're suggesting. You're suggesting that we should not support a way to say "insert the version in scope" -- but rather, if you want to specify the same version as is in your IG (or in a dependency) you have to explicitly put in whatever that version number is... Right?

David Pyke (Jul 22 2020 at 16:34):

Right. I'm a fan of brute force

Chris Moesel (Jul 22 2020 at 17:53):

I'm inclined toward your suggestion, @David Pyke. It keeps things simple and eliminates magic -- and it's often the magic that confuses people. If authors find that it's too burdensome and they really want a magic way of saying "the version in scope", then we can add it in STU2. And if it turns out nobody else needs/wants it, then we've kept FSH simpler!

But... we'll give this thread time to sit overnight in case anyone else wants to comment before we implement it.