Difference between revisions of "201601 PATCH Connectathon Track Proposal"
Jenni syed (talk | contribs) (Add tracking spreadsheet) |
Jenni syed (talk | contribs) (Update use cases from Chris Grenz) |
||
| Line 11: | Line 11: | ||
==Justification== | ==Justification== | ||
| − | There are real world use cases for some form of PATCH, for example, to update only the status of a specific resource, to update or add a comment, and many other single or limited field updates. With the way FHIR currently defines an update, the client should send in the full resource exactly as received, with the single field changed. This is sometimes challenging, both for the FHIR server and the Client. For example, if a client application does not support storing a specific field the server exposes, how will it send it back? If the FHIR server receives an update request where a field is missing, how does it know the intention was to blank/null that field out? | + | There are real world use cases for some form of PATCH, for example, to update only the status of a specific resource, to update or add a comment, and many other single or limited field updates. With the way FHIR currently defines an update, the client should send in the full resource exactly as received, with the single field changed. This is sometimes challenging, both for the FHIR server and the Client. For example, if a client application does not support storing a specific field the server exposes, how will it send it back? If the FHIR server receives an update request where a field is missing, how does it know the intention was to blank/null that field out? More Specifically: |
| + | |||
| + | *Sensitive Elements - a client may not have available to them all elements of a resource, or the some elements may be redacted or de-identified in some way. PATCH would allow the client to send updates to authorized elements without sending the unauthorized or de-identified. | ||
| + | |||
| + | *Targetted updates - explicit delineation of which elements a client is updating. In systems with complex dependency management (element A is dependent on elements B and C), this simplifies the triggering of dependency checks (prevents change detection on fields not patched). Also applies for security (don't have to authorize field updates that aren't patched). | ||
| + | |||
| + | *Bandwidth Contrained - a mobile or other bandwidth constrained client that creates many small updates in real time to a resource. For instance, a UI for updating (sharing) a care plan in real time. | ||
==Proposed Track Lead== | ==Proposed Track Lead== | ||
Latest revision as of 22:16, 9 January 2016
FHIR PATCH for MedicationStatement
Skype: https://join.skype.com/xUx16MpXljhv Connectathon 11 Tracking spreadsheet: https://docs.google.com/spreadsheets/d/13WB1dsQ3Dt8n7s-gOBG7HEpaWL_FFCV_QWM5ea64rV8/edit?usp=sharing
Submitting WG/Project/Implementer Group
This is something FHIR Core requested feedback on DSTU 2: http://hl7.org/fhir/dstu2/http.html#update
Justification
There are real world use cases for some form of PATCH, for example, to update only the status of a specific resource, to update or add a comment, and many other single or limited field updates. With the way FHIR currently defines an update, the client should send in the full resource exactly as received, with the single field changed. This is sometimes challenging, both for the FHIR server and the Client. For example, if a client application does not support storing a specific field the server exposes, how will it send it back? If the FHIR server receives an update request where a field is missing, how does it know the intention was to blank/null that field out? More Specifically:
- Sensitive Elements - a client may not have available to them all elements of a resource, or the some elements may be redacted or de-identified in some way. PATCH would allow the client to send updates to authorized elements without sending the unauthorized or de-identified.
- Targetted updates - explicit delineation of which elements a client is updating. In systems with complex dependency management (element A is dependent on elements B and C), this simplifies the triggering of dependency checks (prevents change detection on fields not patched). Also applies for security (don't have to authorize field updates that aren't patched).
- Bandwidth Contrained - a mobile or other bandwidth constrained client that creates many small updates in real time to a resource. For instance, a UI for updating (sharing) a care plan in real time.
Proposed Track Lead
Jenni Syed
Email: jenni dot syed at cerner dot com Skype: jenni.syed
Expected participants
Cerner (Jenni Syed)
Grahame Grieve
Roles
EHR
Exposes the FHIR MedicationStatement resource with GET (search) and PATCH methods available.
FHIR Client
Has the ability to GET and PATCH MedicationStatement.
Scenarios
Search MedicationStatement
- Action: FHIR Client sends GET to EHR /MedicationStatement resource with patient parameter
- Precondition: EHR should expose GET resource with the JSON format
- Success Criteria:
- Bonus point: MedicationStatement must expose the meta.versionId and meta.lastUpdated fields and support search by patient.
Patch MedicationStatement
- Action: FHIR Client sends PATCH to EHR /MedicationStatement resource using JSON Patch format to update the MedicationStatement.status
- Precondition: EHR should expose PATCH resource honoring the JSON Patch format.
- Success Criteria:
- Bonus point: Rejecting the update if there is a version conflict. Support for updating other fields (eg: note).
Grahame's comment: JSON Patch is pretty dumb. There should be better ways to patch. I'll propose one....
Example Patch
PATCH open/Patient/1 HTTP/1.1 Content-Type: application/json-patch+json
[{ "op": "add", "path": "/active", "value": "false" }]
