Using the Ontoserver for FHIR Operations

37 views
Skip to first unread message

Travis Cummings

unread,
Sep 22, 2017, 1:00:15 PM9/22/17
to HSPC Platform
Hi Preston,

I see the server has a FHIR endpoint available and has a single coding system available.


That coding system has a URL defined at:


Further FHIR queries seem to have no representation though:


Should ValueSet and ConceptMap be returning results?

Or what is the correct way to use the coding system at http://ontoserver.hspconsortium.org/fhir/CodeSystem/32506021000036107-20170731 for FHIR operations?

Thanks!
Travis

Preston Lee

unread,
Sep 22, 2017, 1:21:58 PM9/22/17
to HSPC Platform
Jim Steel and Michael Lawley from CSIRO let me know last night that they have a 5.0.1 release to fix the indexing bug we’ve been experiencing and also have built and published a pre-computed index for us of the SCT international edition on an upstream syndication feed that should alleviate the need to bump the instance type to xlarge for this sort of thing. I’m going to update to the new build now and go through the content loading process again.

Jim also has some clarifying comments on how the implementation selects release versions during query processing. I’ve asked him to re-post those to this list and hopefully we’ll see him join to comment directly. :)

Standby..

Preston


--
You received this message because you are subscribed to the Google Groups "HSPC Platform" group.
To unsubscribe from this group and stop receiving emails from it, send an email to platform+u...@hspconsortium.org.
To post to this group, send email to plat...@hspconsortium.org.
To view this discussion on the web visit https://groups.google.com/a/hspconsortium.org/d/msgid/platform/4814d607-e079-4db3-8b4e-5c06354f6004%40hspconsortium.org.

susancastill...@gmail.com

unread,
Sep 22, 2017, 2:41:10 PM9/22/17
to HSPC Platform
That's great news Preston. Was this by email or an actual phone conversation?
Standby..

Preston


To post to this group, send email to pla...@hspconsortium.org.

Preston Lee

unread,
Sep 22, 2017, 2:47:13 PM9/22/17
to HSPC Platform
Travis: 5.0.1 is up and it seems to have successfully loaded the international version. Lookups now seem to be correctly using the international version by default, as well. Give it a shot!



Anonymous write access is still on … we’ll have to disable that ASAP. Server-side processing time for lookups seems to hover around 30ms without an additional caching layer, which is pretty good.

Preston



Jim Steel

unread,
Sep 24, 2017, 8:14:21 PM9/24/17
to HSPC Platform
Hi folks,

As Preston mentioned, 5.0.1 is now available, and includes the fix for the bug you found.
 
We’ve also published a generated index for the international edition of SNOMED on our CSIRO syndication endpoint. This syndication endpoint (https://ontoserver.csiro.au/synd/syndication.xml) is included as a secondary syndication endpoint by default, but there is a small bug/feature (left in for the temporary benefit of the Australian NRC) whereby a failure to authenticate with the primary syndication site will result in a failure to index, but if you configure your ontoserver deployment to point only at this endpoint (atom.syndication.feedLocation=https://ontoserver.csiro.au/synd/syndication.xml), it should be able to pull down the index rather than having to build it from scratch.
 
If you are building from scratch, there are actually two ways to do this. The “old way”, that you have been using based on your logs, is to
use /api/sct/addRf2 to add RF2 source into ontoserver’s database, then
call /api/indexCodeSystem for each version you want to index.
 
This is quite slow.
 
The alternative is to
construct your own syndication feed, a bit like the one I’m attaching to this message,
point your ontoserver at it, perhaps with our syndication endpoint as a fallback if you like (atom.syndication.feedLocation=file:///syndication.xml,https://ontoserver.csiro.synd/syndication.xml)
after starting ontoserver, copy in the feed file (docker cp syndication-intl.xml ontoserver:/syndication.xml) and the source zip file (docker cp SnomedCT_InternationalRF2_PRODUCTION_20170731T150000Z.zip ontoserver:/ SnomedCT_InternationalRF2_PRODUCTION_20170731T150000Z.zip)
call /api/indexCodeSystem for each version you want to index.
 
This should be faster. It should also be better from the point of view of database size, which I saw mentioned in one of the google group posts.
 
To answer some other questions that I saw on the google group:
 
What does the snomed.default.edition config do?
There are a bunch of ways to refer to a snomed code system resource through the FHIR api. For example, a $expand request can have a uri parameter of http://snomed.info/sct?fhir_vs, of http://snomed.info/sct/900000000000207008?fhir_vs, or of http://snomed.info/sct/900000000000207008/version/20170731?fhir_vs. The last one is explicit about which edition and version will be used, the second is explicit about edition and will use the most recent version it can find. The first one will use the latest version of whichever edition is configured as the default using snomed.default.edition.
 
authentication and authorization
We do have some support implemented for smart-on-fhir oauth-based authorization. It hasn’t had a lot of use, so we’d love to try and help you with us, so just email ontoserve...@csiro.au and we’ll go back and forth to make sure its working for you.
 
Getting rid of the “/fhir” path in the URL
This is a slightly surprising thing to want to do, and is probably best done using a proxy (nginx, apache, whatever), which has the additional benefit of giving performance improvements if you combine it with a web cache, in the style of the example one we have at https://github.com/aehrc/ontoserver-deploy
 
If you have any more queries about these or other issues, just post them here (or shoot them through to ontoserve...@csiro.au if you prefer). We’re generally pretty quick with turnaround (notwithstanding timezone delays).
 
Cheers,
 
Jim.
Standby..

Preston


To post to this group, send email to pla...@hspconsortium.org.

--
You received this message because you are subscribed to the Google Groups "HSPC Platform" group.
To unsubscribe from this group and stop receiving emails from it, send an email to platform+u...@hspconsortium.org.
To post to this group, send email to pla...@hspconsortium.org.
Reply all
Reply to author
Forward
0 new messages