FHIR Chat · broken links in r4 ig

I'm getting a bunch of errors related to "link XXX cannot be resolved". However, the errors seems to be related to links generated by the ig publisher in "snapshot" views, that are for base resource elements that have not been constrained. For example:

The link 'address-type|4.0.0.html' for "address-type|4.0.0" cannot be resolved

The page produced by the IG publisher shows this: "Binding: address-type|4.0.0 (required)" for Organization.address.type, but the profile for the IG does not even constrain Organization.address.type. So, this seems to be related to the base resource's element.

Sean McIlvenna (Mar 11 2019 at 18:58):

I'm thinking this is a bug in the IG Publisher, but it's difficult for me to confirm. Does anyone else know?? OR does anyone else know how to resolve?

Sean McIlvenna (Mar 11 2019 at 19:14):

So, in my case, the base profile is a us-core-r4 profile for organization. It seems that only the elements that aren't constrained are the ones that have link resolution issues

Eric Haas (Mar 11 2019 at 20:48):

is it this issue? https://chat.fhir.org/#narrow/stream/179252-IG-creation/topic/bad.20links.20to.20core.20r3.20and.20r4.20specification/near/160428815

Lloyd McKenzie (Mar 11 2019 at 21:09):

A bunch of the references to v3 value sets accidentally got tagged with FHIR version numbers instead of v3 version numbers. FHIR-I approved a technical correction to fix them, but we don't know yet when all of the FHIR R4 technical corrections will get applied.

Sean McIlvenna (Mar 11 2019 at 21:36):

@Sarah Gaunt , fyi

Sean McIlvenna (Mar 11 2019 at 21:37):

So, @Lloyd McKenzie , what do we do? These are errors that are being produced by the ig publisher, that we have to have resolved in order to publish...

Sean McIlvenna (Mar 11 2019 at 21:37):

but I have no way to resolve them right now

Lloyd McKenzie (Mar 11 2019 at 21:50):

Grahame's chasing stuff down. Anything that's confirmed to be a tooling error won't stop you from going to ballot.

Sarah Gaunt (Mar 11 2019 at 21:58):

Is there a list of confirmed tooling errors? I have 130 broken links in an R3 IG such as:
The link 'http://hl7.org/fhir/STU3/' for "FHIR Version: 3.0.1" cannot be resolved

I don't know how to tell if it's a tooling error or something I am doing wrong on my end.

I mean, that link above does resolve, so I'm not sure how to proceed.

Grahame Grieve (Mar 12 2019 at 02:16):

why are you referencing it?

Grahame Grieve (Mar 12 2019 at 03:16):

so there's several issues with links. Some of them are markdown links which need to be turned from relative to absolute. I've made the changes to do that, but I need to regenerate dependent IGs - namely, US core past versions - to fix the problem in other IGs. I'm probably not going to do that before ballot

Another problem is around the link checker not know about versions of past IGs. I can't easily resolve this either because I have to regenerate past IGs to sort this out.

Lloyd McKenzie (Mar 12 2019 at 14:40):

The link checker is capable of going out and checking non-HL7 published pages and validating the links. Our assumption is that all referenced HL7 pages will be from imported IGs - and I don't think that's an accurate assumption. Is there anything driving that requirement other than performance?

Sean McIlvenna (Mar 13 2019 at 20:55):

For the record, the issue Sarah described is not the same as mine

Sean McIlvenna (Mar 13 2019 at 20:56):

In my case, my IG has a profile whose base profile is a CCDA-on-FHIR profile.
The CCDA-on-FHIR profile has some bindings in it, but my IG profile doesn't change those bindings

Sean McIlvenna (Mar 13 2019 at 20:56):

in this case, the IG publisher seems to be creating invalid links in the snapshot table for those un-constrained bindings

Sean McIlvenna (Mar 13 2019 at 20:57):

while bindings that ARE created in the profiles of my IG end up producing correct links

Sean McIlvenna (Mar 13 2019 at 20:57):

The error I am getting is this:
The link 'encounter-status|4.0.0.html' for "encounter-status|4.0.0" cannot be resolved
(amongst others that are like it)