Cookies for analytics and advertising
We use cookies for analytics and advertising, both sent to Google. Refusing changes nothing you can see.Read the privacy page
Data converter
Tables and configuration files move between formats constantly and the conversion is usually trivial — which is exactly why it is not worth sending your data to a stranger to have it done. Everything here runs in your browser, so an export full of customer records or a config file with credentials in it never leaves your machine.
Up to 100 files at once. Mixed formats are fine.
They convert one after another and download together as a ZIP.
Most used Data converter conversions
Every data format we handle
10 formats in this category, 71 working conversions — 70 of them without your file leaving the device.
- JSONJSON
The data format of the web. Nested structures, read by every programming language.
- CSVCSV
A table as plain text. Every spreadsheet, database and analytics tool reads it.
- NDJSONNDJSON
One JSON object per line. What log pipelines and data exports stream.
- TSVTSV
Like CSV but tab separated, which avoids the comma-inside-a-field problem entirely.
- XMLXML
The verbose ancestor of JSON. Still the backbone of enterprise and government data exchange.
- YAMLYAML
Indentation-based configuration format. Readable by people, picky about whitespace.
- ParquetParquet
Columnar storage for large datasets. Far smaller and faster to query than CSV.
- TOMLTOML
A configuration format that stays readable without YAML’s indentation traps.
- INIINI
Sections and key-value pairs. The oldest configuration format still in daily use.
- SQLSQL
A table rewritten as INSERT statements, ready to load straight into a database.
Read only — we can open this format but not write it.
Tools for Data converter files
Somebody who has to set a password now — a new account, a forced reset, a router or a database user — and does not want to invent one, knowing that anything they invent will be a variation on one they already use.
Somebody verifying that a download arrived intact and unmodified, usually comparing against a hash published on the project page they downloaded it from.
Data converter
Where these conversions actually get interesting
CSV looks simple and is not. There is no single standard: separators differ by locale, quoting rules vary, and a field containing a comma or a newline is where most naive converters break. Numbers are the other trap — an identifier with leading zeros or a long number will be silently mangled by anything that guesses types too eagerly.
The structural mismatch matters more. JSON and YAML nest arbitrarily; CSV is flat. Converting a nested document to a table means deciding how to flatten it, and converting the other way means deciding what the columns become. There is no single right answer, which is why the pages say what they chose rather than pretending the question does not exist.
The bigger formats
Parquet is columnar and compressed, which makes it dramatically smaller and faster to query than the same data as CSV — often by an order of magnitude on anything large. It is the right format for data you will analyse repeatedly, and the wrong one for data a colleague needs to open in a spreadsheet.
Reading and writing it happens in your browser like everything else in this category, which for a dataset of any size is also simply faster than uploading it would be.
Nothing here guesses what a column means
A data conversion has one genuinely hard decision in it, and it is what to do with a value that could be read more than one way. `007`, `1,5`, `03/04/2026` and an empty cell all have at least two defensible readings, and picking wrong is silent — nothing errors, the file simply means something else.
The rule across these pages is to preserve rather than to interpret: a leading zero stays, an ambiguous number stays a string, a date is written in the form that cannot be misread. It produces slightly more work downstream and it never destroys anything, which is the right way round for a step that cannot ask you a question.
Where the file is going decides the format
The choice between these formats is rarely about the data and nearly always about the reader. A script or a database wants JSON or CSV with a plain point decimal. A colleague who will double-click it wants a spreadsheet. A column store or a machine-learning pipeline wants Parquet, and gains most of its advantage on files large enough that the other two have become awkward.
It is worth naming the recipient before choosing, because a correct file in the wrong shape fails just as thoroughly as a broken one. The most common version of this mistake is exporting CSV for a system that wanted typed values, then discovering downstream that every number arrived as text.
Data converter: common questions
Is my data uploaded anywhere?
No, and in this category that is often the whole reason to use it. Exports contain customer records, config files contain credentials, and neither should be pasted into a website. Everything here runs in your browser and nothing is transmitted.
Will long numbers and leading zeros survive?
That is the single most common way data conversion goes wrong, and it is why the pages state how types are handled rather than guessing eagerly. Where a value is ambiguous, preserving what you gave us is the priority.
How is nested JSON turned into a flat table?
By flattening, and the pages say how. There is no single correct answer to what a nested document should look like as columns, so the choice is stated rather than hidden.
Is there a row limit?
No row limit, but a file limit: 100 MB each and a hundred files a batch on the free tier. A 100 MB CSV is several million rows, so the ceiling you meet first is your own memory rather than the row count.