What each format is good at
Converting between data formats is rarely just a syntax change — the formats have genuinely different capabilities, and some conversions lose information. Knowing which direction loses what saves a lot of debugging.
- CSV and TSV are flat and tabular. Rows and columns, no nesting, no types — every value is text. Universally readable by spreadsheets, which is their main advantage.
- JSON is hierarchical and typed: objects, arrays, strings, numbers, booleans, null. The default for APIs and configuration.
- YAML is a superset of JSON with a much more readable syntax — indentation instead of braces, plus comments, which JSON lacks entirely. That comment support is why config files gravitate to it.
- XML is hierarchical with attributes, namespaces, and schema validation. Verbose, but still the standard in enterprise, publishing, and document formats.
Where conversions lose information
The important asymmetry is between flat and nested formats. Going from
JSON to CSV has to flatten nesting — usually by joining keys into
column names like user.address.city — and arrays of varying length
don't map cleanly onto fixed columns at all. Going the other way, CSV to JSON,
everything arrives as strings unless the tool infers types, and inference guesses
wrong on things like leading-zero postcodes and long ID numbers.
YAML to JSON silently drops comments, because JSON has no way to represent them. If the YAML is a config file, those comments are often the most valuable part — convert with care.
XML to JSON has no single correct mapping. XML distinguishes
attributes from child elements and JSON doesn't, so every converter picks a
convention (commonly an @ prefix for attributes). Round-tripping
through JSON and back rarely returns the original document.
The CSV traps
CSV looks simple and isn't. Values containing commas, quotes, or newlines must be quoted and escaped, delimiters vary by locale (much of Europe uses semicolons because the comma is a decimal separator), and line endings differ across platforms. Most "the conversion is wrong" problems trace back to one of these rather than to the data.
Choosing a direction
- Data for a spreadsheet or non-technical colleague? Convert to CSV.
- Data for an API or application? JSON.
- A config file humans will edit? YAML.
- A system that requires a schema? XML.
Everything runs locally
All conversions happen in your browser. Nothing is uploaded, which means you can safely convert customer exports, production data, and anything else you shouldn't be pasting into a remote service.