Stream: terminology
Topic: Expansions in Specification
Grahame Grieve (Aug 07 2018 at 20:58):
GF#13232 proposes to reduce the size of expansions in the spec. The current method is that when the build tool asks for an expansion, it tells the server that it wants up to a 1000 codes. if the server returns an expansion with the correct extension, then the build tool adds a note that the expansion is incomplete.
the task proposes reducing that number down... to 20. Which is too far. Any opinions here?
Brian Postlethwaite (Aug 07 2018 at 21:13):
I assume that this is excluding any that are bound to a code type (required binding), and thus a target for code generators?
Grahame Grieve (Aug 07 2018 at 21:14):
there's none of them that are very long. but some are over 20.
Brian Postlethwaite (Aug 07 2018 at 21:16):
Its been such a good reference source for many of the v2/v3 lists too, and more of them will be truncated.
Language lists...
Paul Lynch (Aug 07 2018 at 21:16):
Maybe there could be a link to the full list (or up to some limit), if the shorter list on the page is truncated?
Grahame Grieve (Aug 07 2018 at 21:17):
if I'm going to generate the full set, i might as well just generate it in the page.
Grahame Grieve (Aug 07 2018 at 21:18):
here's a list of value sets with more than >100 codes and less than <2000 codes, which is actually the current limit:
Grahame Grieve (Aug 07 2018 at 21:18):
reason-medication-status-codes 120
resource-types 142
audit-entity-type 146
route-codes 150
security-labels 200
defined-types 205
all-types 207
relatedperson-relationshiptype 244
spdx-license 346
currencies 356
procedure-not-performed-reason 485
iso3166-1-N 498
iso3166-1-3 498
media-view 552
medication-form-codes 704
device-action 888
referencerange-appliesto 924
performer-role 988
practitioner-role 1000
allergyintolerance-code 1000
service-type 1192
nutrient-code 1398
jurisdiction 1552
Brian Postlethwaite (Aug 07 2018 at 21:18):
1,000 is better than the 10,000 that it was in DSTU2 right?
Grahame Grieve (Aug 07 2018 at 21:19):
not sure. it turns out that there's 2 paths for generating expansion... the internal path is not length limited, and generates one expansion longer than 2000 codes: http://build.fhir.org/valueset-use-context.html
Grahame Grieve (Aug 07 2018 at 21:20):
.. because the expansion is screwed up. I'll have to investigate that one
Brian Postlethwaite (Aug 07 2018 at 21:21):
That's not a huge list of them anyway right?
Grahame Grieve (Aug 07 2018 at 21:23):
Grahame Grieve (Aug 07 2018 at 21:23):
I think 2000 stands up to scrutiny.
Joel Schneider (Aug 07 2018 at 21:47):
Truncating the expansion for value sets such as resource-types and security-labels seems disadvantageous.
Grahame Grieve (Aug 07 2018 at 21:48):
here's the ones we do truncate:
Grahame Grieve (Aug 07 2018 at 21:48):
condition-stage-type
encounter-reason
substance-code
condition-code
condition-outcome
questionnaire-questions
medication-as-needed-reason
report-codes
supply-item
diagnostic-based-on-snomed
procedure-reason
manifestation-or-symptom
procedure-code
observation-methods
body-site
bodystructure-code
condition-stage
adverse-event-type
consent-content-code
flag-code
medication-codes
clinical-findings
observation-codes
doc-typecodes
approach-site-codes
device-kind
food-type
participant-role
Grahame Grieve (Aug 07 2018 at 21:48):
some of them are very large indeed
Rob Hausam (Aug 07 2018 at 22:03):
I think it might make sense to truncate to a smaller number than 1000 (or 2000), but 20 is definitely too small - given what I see in the list maybe 250 would be reasonable? I think that 1000 is more than anyone can really get much out of, but there's not a huge downside to having it that way, either, so I don't care that much about whether we change it (could be not persuasive?).
Grahame Grieve (Aug 07 2018 at 22:28):
so that would mean that the following value sets are truncated:
Grahame Grieve (Aug 07 2018 at 22:28):
v3-AmericanIndianAlaskaNativeLanguages
v3-ServiceDeliveryLocationRoleType
v3-ActRelationshipType
v3-EntityCode
v2-0003
v3-ObservationMethod
v2-0487
v3-orderableDrugForm
spdx-license
currencies
v2-0550
procedure-not-performed-reason
iso3166-1-N
iso3166-1-3
media-view
v3-ActReason
v3-TribalEntityUS
medication-form-codes
v3-RouteOfAdministration
v3-RoleCode
device-action
v3-Race
referencerange-appliesto
performer-role
practitioner-role
allergyintolerance-code
service-type
nutrient-code
v3-HumanLanguage
jurisdiction
Grahame Grieve (Aug 07 2018 at 22:28):
at least some of these, this is the only source I know of
Grahame Grieve (Aug 07 2018 at 22:29):
not - updated list includes v2/v3 codes that I didn't collect stats on originally. trucnated v3/v3 expansions:
v3-ActCode
v3-DocumentSectionType
v3-ObservationType
Rob Hausam (Aug 07 2018 at 22:34):
Along the lines of what I said earlier, if it's going to cause more pain to truncate some of these (as appears to be the case) then it would be better to leave it as it is.
Grahame Grieve (Aug 08 2018 at 11:12):
so some of those numbers are out by a factor of 2... I had a bad bug in my expansion algorithm
Rob Hausam (Aug 08 2018 at 13:40):
ok - should help to fix that, then
Eric Haas (Aug 08 2018 at 17:18):
based on this discussion - i can withdraw. Seems the effort is not worth it and i was only really talking about huge expansions base on intential definitions any way (eg, all of Snomed)
Grahame Grieve (Aug 08 2018 at 20:04):
ok I updated the task
Grahame Grieve (Aug 08 2018 at 20:05):
thanks
Last updated: Apr 12 2022 at 19:14 UTC
