Skip to main content
Version: v0.1.10

Parse messages in a pipeline

cosyte parse reads one input, from a file or from stdin, and writes JSON to stdout. These steps cover the shapes it takes in a pipeline. The examples use the file names message.hl7, stream.mllp, and bulk.ndjson; use your own.

Read from stdin​

Pass - in place of a file name:

cat message.hl7 | cosyte parse -

Pick a field out of the JSON​

--json prints the result on one line. The model is the parsed message as the format's library builds it; for HL7 v2, @cosyte/hl7 keeps each segment's fields in order, with fields[3] holding PID-3:

cat message.hl7 | cosyte parse - --json \
  | jq -r '.model.segments[] | select(.name == "PID") | .fields[3].repetitions[0].components[0].subcomponents[0]'
MRN-000123

The output is the message's own data, so send it only where you would send the message.

Force a format​

Detection reads the content, not the file name, so a .txt that holds an HL7 v2 message parses as HL7 v2. When detection cannot decide, name the format with --format:

cosyte parse --format hl7 message.hl7 --json > /dev/null; echo "exit $?"
exit 0

A format that does not match the content is refused, not guessed around:

cosyte parse --format fhir message.hl7; echo "exit $?"
cosyte: CLI_PARSE_FAILED: the fhir parser rejected the input (MALFORMED_JSON)
exit 65

Stream an MLLP capture​

A file of MLLP frames is parsed one frame at a time. Each frame's HL7 v2 message becomes one line of JSON, written as soon as it is parsed. --quiet drops the summary line on stderr:

cosyte parse stream.mllp --quiet | jq -c '{record, format, segments: (.model.segments | length)}'
{"record":0,"format":"hl7","segments":2}
{"record":1,"format":"hl7","segments":2}

Stream an NDJSON file​

--ndjson treats each non-empty line as one record, the shape of a FHIR bulk data export. A record that fails becomes a line with a stable code in error, and the stream goes on; the run then exits 65:

cosyte parse bulk.ndjson --ndjson --quiet | jq -c '{record, error, resourceType: .model.resourceType}'
echo "exit ${PIPESTATUS[0]}"
{"record":0,"error":null,"resourceType":"Patient"}
{"record":1,"error":"MALFORMED_JSON","resourceType":null}
{"record":2,"error":null,"resourceType":"Observation"}
exit 65

If the reader on the other end of the pipe stops early (head, a pager you quit), the run ends with CLI_OUTPUT_WRITE_FAILED and exit 74, never 0.

Mind the size limit​

One invocation reads at most 64 MiB of input. Split a larger capture into files and parse each one.

Reference​