CSV to JSON Converter

Because turning it into 7 would destroy it. Type inference here is deliberately conservative: a leading zero, a leading plus, exponent notation, or a value beyond safe integer precision all keep the cell as text. That covers postcodes, phone numbers, product codes and long IDs, which are the columns people most often find quietly mangled by other converters. Switch the inference toggle off to keep every cell a string.

Every candidate - comma, semicolon, tab and pipe - is used to parse a sample of the first rows, and the one producing the most consistent column count across those rows wins, with a tie going to comma. Delimiters inside quoted fields are ignored, so a semicolon file whose cells contain commas is still detected correctly. You can always set it explicitly.

An empty cell becomes null, while a cell written as a pair of quotes becomes an empty string. That is the only place the distinction between quoted and unquoted matters, and it is preserved because a blank in a spreadsheet usually means missing rather than empty.

The second and later ones are suffixed - name, name_2, name_3 - so no data is lost to a silently overwritten key. Blank header cells become column_1, column_2 and so on by position.

Five megabytes and 100,000 rows. Both are named errors rather than a silent truncation, so a file that is too large tells you instead of converting only part of itself.