Database tools
Extract literal rows, inspect each table, and export the selected table.
Parsed locally in your browser. SQL is never executed or sent to a database.
Only safe integers become JSON numbers. Other numeric literals remain strings; CSV preserves number spelling.
Explicit columns and literal VALUES only. Names match exactly. MySQL uses backslash escapes; PostgreSQL/SQLite preserve backslashes.
Max 200000 UTF-16 units · 500 statements · 100 tables · 100 columns/table · 5000 rows · 100000 cells. Export text is limited to 2000000 UTF-16 units to bound repeated keys.
Enter INSERT statements or load the example.
SQL to JSON helps when a seed script needs to become an API fixture. Keep the explicit column list: it supplies the object keys and lets repeated INSERT statements line up even when their column order changes.
SQL to JSON groups rows by their table identifier. An orders INSERT cannot silently inherit the people columns. Select a table to inspect its own headers and export just those rows; schema-qualified names and a quoted name containing a dot stay distinct.
A large identifier should not change its last digit on the way to a test fixture. Safe integers become numbers; larger integers, decimal literals and exponent notation become strings. A function call stops conversion with a position instead of turning into an invented value.
PostgreSQL and SQLite preserve backslashes in ordinary strings. MySQL interprets its supported escape sequences. The selector does not detect server SQL modes: use the documented subset and review strings containing backslashes.
The status reports all statements and tables, then the selected table’s rows and columns. The grid reveals 50 rows at a time. Copy and download include every row of that table, even when the grid shows only the first page.
Automatic conversion waits 350 ms after an edit. Turn it off to use the parse button or Ctrl/Cmd + Enter. Any change disables exports immediately; a failed conversion leaves an error and keeps the older result unavailable.
SQL INSERT to JSON uses explicitly listed column names as keys. Later statements may list those same names in another order: values are mapped back to the first statement’s column order. Duplicate columns and changed column sets fail rather than overwrite or fill invented values.
A scanner tracks quotes instead of splitting text on every comma. Doubled apostrophes, commas, parentheses, semicolons and line breaks inside a single-quoted value stay in that value. Ordinary comments between tokens are ignored; unterminated comments are errors.
SQL to CSV produces a comma-separated header and CRLF rows, quoting fields and doubling embedded double quotes. SQL NULL becomes an empty CSV field; JSON keeps null distinct from an empty string. Use JSON when downstream code needs those types.
This SQL INSERT converter shows the exact text copied or downloaded. CSV formula protection prefixes risky text and headers with an apostrophe by default; numeric negative values remain unchanged. A count identifies affected cells, and turning the option off restores original text.
Yes. Every parenthesized VALUES tuple becomes a row. A tuple must contain exactly as many values as the explicit column list. If any row is malformed, the entire conversion fails, so a shorter partial export cannot be mistaken for the whole input.
They form separate table groups in first-seen order. Select a group before exporting. Repeated statements for the same exact table identifier append rows; this does not deduplicate rows, merge relationships or flatten different tables into one dataset.
NULL becomes JSON null and TRUE/FALSE become booleans. Numbers 0 and 1 stay numbers, without schema-based boolean inference. A quoted date keeps its spelling. Decimal and large integer literals become strings to preserve their digits.
No. The parser recognizes the supported INSERT grammar and extracts literals. NOW(), subqueries and other expressions have no value here and produce an error. Successful extraction does not prove the script is valid for a particular database schema.
Without a column list, there is no reliable way to recover object keys from the text alone. Include the names in the original export. The tool neither reads CREATE TABLE nor assigns placeholder names that could hide an incorrect column mapping.
Found a supported literal that parses incorrectly?
Use SQL to JSON with a small representative INSERT first. Check quoted values, the selected table and the preservation count before exporting a larger batch.