1.1 KiB
Architecture
The immutable catalog defines formats, domain properties, directed conversion edges, availability, and a per-property status. A bounded depth-first search enumerates acyclic paths of at most four conversions. Routes are sorted by the worst status for the user-selected properties and then by conversion count; catalog results remain explicitly expected rather than measured.
The executable data lab parses JSON, NDJSON, CSV, and a typed XML record dialect into a bounded canonical list of records. Every selected edge serializes and re-parses the canonical data, so losses accumulate along the actual path. The target is then serialized back to the source format. Stable property signatures compare record/field order, names, types, nulls, nesting, numeric values, and Unicode text. Metadata remains not-tested because the canonical model does not pretend to represent arbitrary wrappers, comments, namespaces, or schemas.
Limits are 2 MiB input, 10,000 records, 200 fields per record, 500,000 top-level values, JSON depth 64, 200,000 JSON nodes, and 8 MiB output.