FI-2701: Migrate to HL7 validator wrapper

[dehall]

Summary

This PR migrates the test kit from the Inferno validator-wrapper to the HL7 validator-wrapper. This is similar to most of the other test kits, with one significant difference that his test kit used the inferno validator's FHIRPath endpoint, so this now includes the fhirpath-service in the docker-compose files and nginx.background.conf.

Other changes:

• add the hl7_validator service to docker-compose and nginx. The old validator components are left commented out in case someone really wants the old validator UI
• Bump the core dependency
• Migrate validator blocks to fhir_resource_validator and specify the IG
• Update the generator with ^ and re-generate the suite
• Update the fhirpath evaluation method to call the appropriate endpoint (based on Migrate to the FHIRPath service #2 )
• Add a README to the IGs folder explaining when it's necessary for the files to be there
• Fixes an issue where the custom validation logic tries to validate against native profile urls with the IG's version, in particular http://hl7.org/fhir/StructureDefinition/PractitionerRole|2.0.1 . The validator doesn't have this profile loaded which results in a 500 error and a purple test result
• Updates the preset and sample bundles to remove relative links, eg <a href="#something"> . A new check in the validator has strict requirements for these and produces "Hyperlink does not resolve" errors. The replacement was <a href=\\\\\\"#.*?">See above \((.*?)\)</a> --> \1, eg, <a href=\\\"#Patient_SubscriberExample\\\">See above (Patient/SubscriberExample)</a> became just Patient/SubscriberExample

Testing Guidance

From a user's perspective there should be no visible difference between this and using the old validator. Run through the tests and make sure everything works

[tstrass]

Changes look good. I just noticed one difference. With the new validator and preset, the server suite input bundles for approval and denial pass validation, and with the old validator and preset they fail validation with the messages below. Is this expected?

[dehall]

Failing on the old one and passing on the new one is unexpected. I can look more into that

[tstrass]

Here's my test run: https://inferno.healthit.gov/suites/davinci_pas_server_suite_v201/b8L0hB6WlyK#davinci_pas_server_suite_v201-Group01-pas_server_v201_approval_use_case

[dehall]

Ignore the previous (now-deleted) message if you saw that. I wonder if the prod validator instance is in an inconsistent state because running that same test locally on main passes for me:

[dehall]

Ok I see the difference now. That error we're seeing on prod is for a CRD extension http://hl7.org/fhir/us/davinci-crd/STU2/StructureDefinition-ext-coverage-information.html and so the validator will only check that when the CRD IG is loaded, either explicitly or as a dependency of another IG. When you run this test kit by itself on main, you don't get those errors. But on prod the old validator instance is running for a few test kits and one of them may be bringing in a different version of CRD. If you load all the same IGs into the validator locally as are running on prod, the same error happens. (davinci-pdex-2.0.0.tgz, davinci_pas_2.0.1.tgz, uds-plus-external-package.tgz, davinci_dtr_2.0.1.tgz, udsplus-101.tgz)

So, I think this should be good, in terms of matching the old behavior when running in isolation. Let me know if you disagree @tstrass or @karlnaden , otherwise I'll merge this in

[karlnaden]

your analysis looks sound to me. I'll defer to Tom having not looked closely.

[tstrass]

Agreed, thanks for investigating

[dehall]

Just for posterity, turns out that constraint that was failing is bad anyway. https://jira.hl7.org/browse/FHIR-46087 (thanks fhir-at-mitre slack channel)