Convert a table between CSV, JSON, Markdown and HTML in any direction. Everything is read into one common shape first and then written out, so the same rules apply whichever pair you pick. HTML is read with the browser's own parser rather than a regular expression, which is why a cell wrapped in a link or a bold tag gives you its text instead of its markup — the regex version of this breaks on the first nested tag. Markdown output escapes pipes inside cells, and Markdown input unescapes them, so a value like 'yes | no' survives a round trip instead of splitting into two columns. JSON input uses the union of keys across all objects, so a record missing a field leaves an empty cell rather than shifting every column after it. Nothing is uploaded.
| Format | Reading | Writing |
|---|---|---|
| CSV | RFC 4180, delimiter detected | Quotes only where needed |
| JSON | Union of keys, nesting stringified | Optional type inference |
| Markdown | Separator dropped, pipes unescaped | Pipes escaped, columns padded |
| HTML | DOMParser, text from nested tags | Markup characters escaped |
No. Every conversion happens in your browser.
With DOMParser, the browser's own HTML parser. That means malformed markup, attributes and tags nested inside a cell are all handled by the engine that renders web pages, and a cell's value is its text however it is wrapped. Regex-based table scrapers fail on the first nested tag.
Writing Markdown escapes it, and reading Markdown unescapes it, so a value like 'yes | no' survives the round trip. An unescaped pipe ends the cell, which silently splits the value into two columns — the most common way a generated Markdown table breaks.
The header is the union of every object's keys in first-seen order, so an object missing a field produces an empty cell rather than shifting the remaining values into the wrong columns.
It is serialised into the cell as JSON rather than dropped. A table has no place to put nesting, and dropping it would lose data without telling you.
Yes, in every format. For HTML that is the first row of the first table on the page, whether it uses th or td cells.
Every input is normalised to one grid of rows before anything is written out, which is what keeps twelve possible conversions to four readers and four writers rather than twelve separate code paths. HTML is read with the browser's own DOMParser rather than a regex, so malformed markup, attributes and inline tags inside a cell are handled by the same engine that renders pages — textContent yields the cell's value whatever is wrapped around it, and a regex approach fails on the first nested tag. Markdown reading drops the separator row by shape rather than by position, so a table with an unusual leading line still parses, and pipes escaped as backslash-pipe are unescaped back into the cell rather than splitting it. Writing Markdown does the reverse, escaping every pipe, because an unescaped one ends the cell and silently splits the value in two. JSON input takes the union of keys across all objects in first-seen order, so an object missing a field produces an empty cell instead of shifting the columns; a nested value is serialised rather than dropped, since dropping it loses data without saying so. HTML output escapes the four characters that matter, which means a cell containing markup round-trips instead of becoming part of the table. CSV writing quotes only the fields that need it and doubles internal quotes, per RFC 4180. Everything runs in your browser.