Stream: implementers
Topic: profiling problem with slices
Sean McIlvenna (May 03 2017 at 17:32):
I'm following the examples here https://www.hl7.org/fhir/profiling-examples.html#extensions to define slices for different extension.
However, when I run this through the ig publisher, with some example resources, the example resources are throwing warnings saying "Profile http://hl7.org/fhir/ccda/StructureDefinition/CCDA-on-FHIR-Consultation-Note, Element matches more than one slice" for each of the extensions in the example.
Sean McIlvenna (May 03 2017 at 17:33):
CCDA-on-FHIR-Referral-Note.xml
CCDA-on-FHIR-US-Realm-Header.xml
CCDA-on-FHIR-Referral-Note-Example.xml
Sean McIlvenna (May 03 2017 at 17:34):
attached is the strucdef for referral note and its base profile us-realm-header, as well as the example instance it is failing on. can someone take a look and tell me what I'm doing wrong?
Sean McIlvenna (May 03 2017 at 17:34):
the extensions are defined in the base us-realm-header profile
Lloyd McKenzie (May 03 2017 at 20:17):
First, make sure you're using the absolute latest version of the IGPublisher (from build.fhir.org) as several corrections have been made recently.
Sean McIlvenna (May 03 2017 at 20:29):
so I shouldn't be using the ig publisher packaged with stu3?
Sean McIlvenna (May 03 2017 at 20:30):
seems like the stu3 should be updated (as though it were errata) with these kinds of important changes
Sean McIlvenna (May 03 2017 at 20:35):
just downloaded and ran with latest version of ig publisher
Sean McIlvenna (May 03 2017 at 20:35):
getting same errors
Lloyd McKenzie (May 03 2017 at 20:36):
Ok. I'll have a fix that should go in this weekend that will at least tell you which slices it's finding a match against that are a problem
Eric Haas (May 03 2017 at 20:37):
how about a screenshot of the spreadsheet to see how is defined.
Eric Haas (May 03 2017 at 20:37):
and a screenshot of the CD snippet of interest
Eric Haas (May 03 2017 at 20:38):
I find inspection of the SD usually helps ferret out the issue
Sean McIlvenna (May 03 2017 at 20:55):
Sean McIlvenna (May 03 2017 at 20:55):
interesting... it doesn't show anything about the profile that those extensions are for
Eric Haas (May 04 2017 at 02:36):
I didn't see a problem either except the error is for http://hl7.org/fhir/ccda/StructureDefinition/CCDA-on-FHIR-Consultation-Note and the example's declared profile is http://hl7.org/fhir/ccda/StructureDefinition/CCDA-on-FHIR-Referral-Note ..so the error must lie there
Sean McIlvenna (May 04 2017 at 04:47):
@Eric Haas sorry that was a copy/paste error. I'm getting the same error for ALL Composition.section slices across all of C-CDA on FHIR's composition profiles.
Sean McIlvenna (May 04 2017 at 04:48):
which makes the list of errors quite large...
Sean McIlvenna (May 04 2017 at 04:48):
but, if I can figure out why it is producing that error, then I can eliminate a TON of those errors with one code-fix
Sean McIlvenna (May 04 2017 at 04:49):
@Grahame Grieve, you mind taking a look at this thread when you get a sec, please?
Sean McIlvenna (May 04 2017 at 04:49):
maybe you'll see something...
Sean McIlvenna (May 05 2017 at 16:54):
Could still use a hand here if someone wouldn't mind... just not figuring out why these slices aren't working
Lloyd McKenzie (May 05 2017 at 19:14):
The discriminator should be expressed relative to the element being sliced. So, for example, when Slicing Composition.section, the discriminator is "code", not "Composition.section.code"
Lloyd McKenzie (May 05 2017 at 19:14):
See if that helps and if not, I'll dig deeper
Lloyd McKenzie (May 05 2017 at 19:14):
(And sorry for taking so long to take a quick glance at your instances :()
Sean McIlvenna (May 05 2017 at 19:15):
oh man, that's it?!
Sean McIlvenna (May 05 2017 at 19:15):
awesome find, if so!
Sean McIlvenna (May 05 2017 at 19:41):
Unfortunately it's still showing the same errors
Sean McIlvenna (May 05 2017 at 19:43):
Instance: CCDA-on-FHIR-Referral-Note-Example.xml
Sean McIlvenna (May 05 2017 at 19:43):
StructureDefinition: CCDA-on-FHIR-Referral-Note.xml
Sean McIlvenna (May 05 2017 at 19:44):
Errors: pasted image
Lloyd McKenzie (May 05 2017 at 19:47):
Hmmm. You're slicing via pattern. That's supposed to work, but I've never tested it. I usually specify code.coding.system and code.coding.code as dual discriminators and assign fixed values. It's possible pattern slicing doesn't work yet.
Sean McIlvenna (May 05 2017 at 19:48):
Well, we could fix the code
Sean McIlvenna (May 05 2017 at 19:48):
but, I think we did want to leave it open to pattern
Lloyd McKenzie (May 05 2017 at 19:52):
I understand exactly why you're doing what you're doing. It just may be that you're treading in unexplored territory
Sean McIlvenna (May 05 2017 at 19:55):
So, how do we go about getting the ig publisher updated to support pattern-based slices?
Sean McIlvenna (May 05 2017 at 19:55):
:)
Lloyd McKenzie (May 05 2017 at 20:22):
Well, at the moment, it means writing code to turn a pattern into a FHIRPath expression, which is not an easy thing to do. My guess is we'll just handle the obvious one of Coding with code + system and throw an exception when we see anything else. I may have time to do that tomorrow.
Sean McIlvenna (May 08 2017 at 15:51):
Note that when I change them to "fixedCodeableConcept" I get the following error:
"Unsupported fixed value type for discriminator(code) for slice"
Sean McIlvenna (May 08 2017 at 15:51):
actually... that message is a warning.
and then I still get "Element matches more than one slice" as an error a bunch of times.
Sean McIlvenna (May 23 2017 at 19:33):
@Grahame Grieve or @Lloyd McKenzie
Were either of you ever able to take a look at this? Still not sure if this is a limitation of the IG Publisher (or validation logic used by the publisher) or if it is something I am doing wrong with the samples...
Lloyd McKenzie (May 23 2017 at 19:49):
You won't be able to use a fixed codeable concept. You'll need to do a fixed system and a fixed code. (and use both as discriminators)
Rick Geimer (May 24 2017 at 19:59):
Would be better to get pattern-based slices working in the IG publisher. Lloyd, is this restricted to the IG publisher code, or do you think this is something deeper? If just in the IG publisher I'm happy to take a look early next week and try fixing the bug.
Rick Geimer (May 24 2017 at 20:06):
Basically I worry that we are restricting what can be legally expressed in profiles to what we currently support in the IG publisher.
Lloyd McKenzie (May 24 2017 at 21:01):
It's not the publisher, but the validator (which is used by the publisher as well as stand-alone)
Lloyd McKenzie (May 24 2017 at 21:02):
If you want to take a stab at it @Rick Geimer, you're most welcome. I apologize for not getting to it yet. And realisticly, I won't have bandwidth to look at it for at least another couple of weeks.
Rick Geimer (May 24 2017 at 22:11):
Thanks, I'll work with Sean and take a look next week. Worst case will fall back to fixed system and fixed code per your suggestion.
Grahame Grieve (May 25 2017 at 03:14):
I can have a look on Monday - need to do a run of validator stuff.
Sean McIlvenna (May 30 2017 at 18:03):
I think I found the section of the InstanceValidator that needs to be improved to support fixedCodeableConcept and patternCodeableConcept... and Coding:
https://github.com/hl7-fhir/fhir-svn/blob/master/implementations/java/org.hl7.fhir.validation/src/org/hl7/fhir/r4/validation/InstanceValidator.java
Line #3478
private boolean valueMatchesCriteria(Element value, ElementDefinition criteria) throws FHIRException { ... }
Grahame Grieve (May 30 2017 at 18:46):
can you turn this into a test case ?- that is: a structure definition, and an instance that fails and one that succeeds
Sean McIlvenna (May 30 2017 at 18:57):
well, I have examples of SD's that should succeed... how do you want me to modify it so that it fails? (it is already failing)
Grahame Grieve (May 30 2017 at 18:59):
an instance that fails, not a SD that fails
Grahame Grieve (May 30 2017 at 19:00):
e.g. that is not valid against the SD
Sean McIlvenna (May 30 2017 at 19:00):
Oh gotcha.. Yes, I will get those created and attach them here
Grahame Grieve (May 30 2017 at 19:02):
thanks
Sean McIlvenna (May 30 2017 at 19:18):
Structure Definition: CCDA-on-FHIR-Referral-Note-orig.xml
Uses patternCodeableConcept for Composition.section.code, discriminates on Composition.section.code
Sean McIlvenna (May 30 2017 at 19:18):
Bad example: CCDA-on-FHIR-Referral-Note-Example-bad.xml
Sean McIlvenna (May 30 2017 at 19:19):
Good example: CCDA-on-FHIR-Referral-Note-Example.xml
Sean McIlvenna (May 30 2017 at 19:19):
the bad example duplicates one of the section, when it should be 1..1
Sean McIlvenna (May 30 2017 at 19:19):
and it also has an improper code for one of the sections
Sean McIlvenna (May 30 2017 at 19:19):
that section should not be matched to any slices
Sean McIlvenna (May 31 2017 at 21:18):
Is this what you were looking for @Grahame Grieve ?
Grahame Grieve (May 31 2017 at 23:35):
probably - got distracted and haven't got to it yet. hopefully today
Sean McIlvenna (Jun 06 2017 at 19:40):
Did you ever get to this @Grahame Grieve ?
Grahame Grieve (Jun 07 2017 at 12:34):
so I've spent about 8 hours on this now. Lloyd, umm, 'improved' the validator with the upshot that it's been ignoring profiles altogther
Grahame Grieve (Jun 07 2017 at 12:34):
now I have figured that out, and also figured out your canonical url mismatch ('_' vs '-')
Grahame Grieve (Jun 07 2017 at 12:35):
I'm finally debugging the actual slice code at last
Sean McIlvenna (Jun 07 2017 at 17:11):
well, good to hear there is progress (I think?) :)
Sean McIlvenna (Jun 07 2017 at 17:11):
sounds frustrating, though...
Sean McIlvenna (Jun 07 2017 at 17:12):
and yes, there have been problems with _ vs. - cause at one point the ig publisher wasn't handling dashes very well, so we converted them to underscores. Been a mess since then. trying to normalize them all now.
Lloyd McKenzie (Jun 07 2017 at 17:44):
I'm pretty sure it wasn't ignoring all profiles, but appologize for contributing to the problem.
Grahame Grieve (Jun 07 2017 at 20:51):
ok I just committed an update to the validator. @Sean McIlvenna, the validator says that your example good resources is missing some required slices. I haven't had a chance to check that yet - can you have a look at that?
Grahame Grieve (Jun 07 2017 at 20:51):
I'll be back on it in a few hours time
Grahame Grieve (Jun 07 2017 at 22:04):
this i swrong:
<element id="111256">
<path value="Composition.section.title"/>
<sliceName value="section_slice_pos2"/>
<short value="title"/>
<definition value="SHALL contain exactly one [1..1] title (CONF:2219-2079)."/>
<min value="1"/>
<max value="1"/>
</element>
Grahame Grieve (Jun 07 2017 at 22:04):
don't repeat the slice name on the parts of the slice, just on the first entry
Grahame Grieve (Jun 07 2017 at 22:05):
(deleted)
Grahame Grieve (Jun 07 2017 at 22:10):
And you're still using the wrong system URL for LOINC
Grahame Grieve (Jun 07 2017 at 22:22):
and also, pattern is case sensitive. are you sure you want to specify the display in the pattern?
Grahame Grieve (Jun 07 2017 at 22:33):
ok, all fixed. You should compare what I ended up with (ccda*.xml in build\tests\validation-examples) with what you gave me - will show the minimum changes I had to make to get the tests working.
Last updated: Apr 12 2022 at 19:14 UTC
