Protocol Buffers for FHIR
summary
Open-source toolkit converts FHIR STU3 to Protocol Buffers with auto-generated protos, a Java FHIR-JSON parser and examples that upload to BigQuery; internal proto for all resources is about contains 17,000-line resources.proto. Messages parse dates to epoch, store decimals as strings to preserve precision, inline extensions into fields, and claim about 10x smaller storage than JSON. Versioned releases (0.1, 0.2, upcoming 0.3) add bulk-download client that found vendor validation bugs and upcoming TensorFlow feature/label exporters for ML.
AI-written summary · about
chapters (21)
- 0:00Start
- 4:02Message Definitions compile into code
- 4:45Protocol Buffers - using generated code
- 5:34Protocol Buffers - supported languages
- 5:57Protocol Buffers: Serialization
- 9:09A Common Representation
- 9:33Our main use case
- 10:36Inside Google, the choice is Protobuf
- 11:57Outside Google, it's a bit more nuanced
- 12:38Approach
- 19:38Inline CodeSystems
- 20:06A Patient Example
- 21:27Primitive Types - Boolean
- 23:00Primitive Types - Double
- 24:01Primitive Types - Date Time
- 24:53Extensions - Patient Nationality
- 26:24As part of the Patient Resource
- 28:29Converting to/from JSON
- 29:28Version 0.2
- 32:54Goal: As Easy As Computer Vision
- 34:29Download and Try
related talks
48mProtocol Buffers- A Banked Journey
45mHow to solve scaling pains with FHIR and data streaming?
23mEffective Protobuf: Everything You Wanted To Know, But Never Dared To Ask
25mProtobuf Editions
47mMapping FHIR to OMOP Using Open Source Tools
35mMix, Match, Merge
46mThe FHIR Bulk Data API and what's new in v2!
20mDehydrated FHIR