test: add NDJSON-specific edge case coverage

[membphis]

Background

qjson is a single-document parser — it does not provide a native NDJSON API. Callers split lines themselves and call qjson_parse() per record. This means most NDJSON edge cases (empty lines, CRLF, whitespace-only lines, error recovery) are caller responsibility.

Goal

Add systematic edge case coverage for scenarios that ARE qjson's responsibility when parsing individual NDJSON records.

Scope

Edge Cases to Cover

Edge Case	qjson Responsibility	Rationale
UTF-8 BOM at input start	Yes	RFC 8259 prohibits BOM; qjson should reject
Large single record (1MB+)	Yes	Verify no stack overflow or perf regression
Many small records (10K+)	Yes	Verify no memory leaks in parse/free cycle

Out of Scope (Caller Responsibility)

• Empty lines / whitespace-only lines → caller's split logic
• CRLF vs LF handling → caller's line trimming
• Malformed middle record error recovery → caller decides whether to continue
• Streaming NDJSON → qjson requires full buffer per record

Acceptance Criteria

• Add tests/ndjson_edge_cases.rs with:
  • BOM handling test (expect parse error per RFC 8259)
  • 1MB+ single-line record stress test
  • 10K+ small record batch test (parse/free cycle)
• All tests pass in both EAGER and LAZY modes
• Document that qjson is a single-document parser

Technical Notes

• Follow existing test patterns from ffi_depth_stress.rs
• Use DocGuard RAII pattern for cleanup
• BOM bytes: \xef\xbb\xbf (UTF-8 BOM)

References

• RFC 8259 §8.1: "Implementations MUST NOT add a byte order mark"
• NDJSON spec: https://github.com/ndjson/ndjson-spec

[coderabbitai]

🔗 Related PRs

#43 - test: add cjson and simdjson fixture coverage [merged]
#51 - perf: skip redundant validation in eager decode, inline depth check [closed]
#78 - test(fuzz): add nesting-depth stack-safety target [merged]
#141 - test: add FFI grammar corpus [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[membphis]

Scope Clarification

qjson does not provide a native NDJSON API — callers split lines themselves and call qjson_parse() per record. This means most edge cases listed in the issue (empty lines, CRLF, whitespace-only lines, error recovery) are caller responsibility, not qjson's.

What's actually in scope for qjson

Edge Case	qjson Responsibility	Current Coverage
UTF-8 BOM at input start	Yes — qjson should reject or handle	❌ None
Large single record (1MB+)	Yes — no stack overflow or perf regression	⚠️ Implicit (1.7MB fixture exists, no dedicated test)
Many small records (10K+)	Partial — verify no memory leaks in parse/free cycle	❌ None (current NDJSON test has 793 records)
Empty lines / CRLF / whitespace lines	No — caller's split logic	N/A
Malformed middle record	No — caller decides error handling	N/A

Revised acceptance criteria

1. Add BOM handling test (expect parse error per RFC 8259)
2. Add 1MB+ single-line record stress test
3. Add 10K+ small record batch test (parse/free cycle, check for leaks)
4. Document that qjson is a single-document parser; NDJSON requires caller to split lines