Yuki Tanaka

My interest is the lossy step. Every conversion here is straightforward to write and quietly destructive, so I show the round trip and mark what came back different. JSON to CSV flattens nested objects into columns and flattens arrays into repeated rows or serialised strings, and the convention you pick decides whether the file can be read back into the original shape. Arrays of objects break this hardest, because there is no correct answer without knowing what the consumer expects.

JSON to YAML produces the file most people wanted, with two recurring problems. Long scalar values get line-wrapped into something that reads as several lines when parsed back, and any string containing a colon, a hash or a leading indicator character needs quoting. JSON to TypeScript is the one worth automating, and it is still guesswork. A field's inferred type comes from the sample values, so a field seen only as null becomes nullable, and one seen as an integer may arrive as a string from a different endpoint.

I cover inferring from several samples rather than one. Conversion has a direction worth naming. Going from a typed source, where a schema already describes the structure, into JSON is a projection you can repeat.

Coming back from JSON into a typed shape means inventing a schema nobody wrote, and that schema becomes an API you now maintain. I cover choosing between a strict schema and a loose record, because the answer depends on who reads the output next. There are security notes here too. A conversion turning a JSON file into a configuration file can carry untrusted keys straight into an interpreter, and a YAML loader using an unsafe schema will construct arbitrary objects from input.