I've noticed that while the docs for that protocol (rightly, IMO) mandate the timestamp and series attributes for all messages, the messages emitted by this tool do not contain either of those.
If we were to add them to the messages output by the tool, it would invalidate the sample data in testdata/good/*.ndjson. Since the values for series and timestamp will be dynamically generated at run-time and are non-deterministic, we won't be able to use static values for them in the test data.
I'm not sure how to resolve this. How are those testdata files used? Could the tools that check them against the actual output be made more sophisticated to be flexible about the exact values of timestamp / series? What other options do we have?
I'm working to bring the schemas in https://github.com/cucumber/cucumber/tree/master/event-protocol up to date with the messages emitted by the
gherkin
command-line tool.I've noticed that while the docs for that protocol (rightly, IMO) mandate the
timestamp
andseries
attributes for all messages, the messages emitted by this tool do not contain either of those.If we were to add them to the messages output by the tool, it would invalidate the sample data in
testdata/good/*.ndjson
. Since the values forseries
andtimestamp
will be dynamically generated at run-time and are non-deterministic, we won't be able to use static values for them in the test data.I'm not sure how to resolve this. How are those
testdata
files used? Could the tools that check them against the actual output be made more sophisticated to be flexible about the exact values of timestamp / series? What other options do we have?