Let UTF-8 and SSE framing survive arbitrary transport boundaries.
Type: Build Languages: TypeScript, Go Stage: 1 of 4 Prerequisites: Basic functions, collections, JSON and the previous stages when applicable. Time: ~2 hours
What you build
Build an incremental UTF-8 decoder and line parser. Handle LF, CRLF and CR, even when the separator or multibyte character spans chunks. A blank line dispatches data; EOF by itself does not. Comments are ignored. Multiple data fields join with a newline.
Worked example
Split the two bytes of é across pushes. No replacement character should appear. id: 1, data: Caf and a blank line dispatch one event. An event without a new id inherits the previous id; an empty id resets it.
pj streaming answer recovery 1
Implement the contract
new SSEDecoder().push(bytes:Uint8Array,final=false):Event[], where Event is {id,event,data}. Default event type is message. Ignore id values containing NUL. Final flush rejects truncated UTF-8 and discards un-dispatched data. This subset ignores retry fields.
Reject malformed input before producing a partial artifact. The tests include held-out records separate from the demonstration fixture.
Run your work
python3 scripts/project_test.py streaming-answer-recovery --init /tmp/streaming-answer-recovery-work
python3 scripts/project_test.py streaming-answer-recovery --stage 1 --path /tmp/streaming-answer-recovery-work --strictInitialization copies cumulative scaffolding without replacing your work. Implement the TODO functions in the first-stage starter; later-stage starter notes describe the new behavior. Tests import the learner workspace. Run the command in the README against your own input after passing the stage.
Inspect and extend
Try a trailing data line without a blank line. Explain why prematurely dispatching it can display incomplete JSON.