Excel to SQL

How to convert an Excel sheet to SQL

1. Drop the file. A .xlsx with several sheets shows a sheet picker; a CSV is read with its encoding detected, so accented text and CJK survive.
2. Check the inferred types. The column table shows what each column became. A column of numbers with a leading zero is kept as text on purpose; change the header or the data if you want it numeric.
3. Pick the dialect and batch size. Quoting, boolean literals and the maximum rows per INSERT differ between MySQL, PostgreSQL, SQLite and SQL Server; the output follows the one you chose.
4. Copy or download. Copy the SQL to the clipboard for a console, or download the .sql file to run with your client of choice.

What the types are inferred from

Every value in a column is looked at, not just the first few rows. A column becomes an integer only if every non-empty value is a whole number without a leading zero; a decimal if every value parses as a number once currency symbols and thousands separators are removed; a date or timestamp if every value is ISO-shaped (2026-09-06, optionally with a time); a boolean if every value is true/false or yes/no. Anything else is text, sized to the longest value with headroom — VARCHAR(50), VARCHAR(100) — and switched to TEXT when that would exceed 255 characters. Empty cells become NULL in every type.

The one rule that surprises people is deliberate: 00012345 is text. Invoice numbers, account numbers, postcodes and product codes carry their leading zeros as part of the identifier, and a column that loses them on import is the first thing a reconciliation trips over later.

Four dialects, four sets of details

• MySQL / MariaDB — backtick-quoted identifiers, DECIMAL(18,4) for money, TRUE/FALSE, VARCHAR up to 255 then TEXT.
• PostgreSQL — double-quoted lower-case identifiers, NUMERIC(18,4), TEXT for strings (it costs nothing there), TIMESTAMP for dates.
• SQLite — its four storage classes: INTEGER, REAL, TEXT; dates stay text, which is what SQLite itself recommends.
• SQL Server — bracketed identifiers, NVARCHAR with N'…' literals so non-Latin text survives, BIT for booleans, and INSERTs capped at 1,000 rows each because that is the server's own limit per statement.

Column names become identifiers

Headers are turned into safe identifiers: spaces and punctuation become underscores, everything is lower-cased, a leading digit gets an underscore in front, and a repeated header gets a number (amount, amount_2). Empty headers are named by position. The original header text is not lost — it is visible in the column table above the SQL, so you can rename in the sheet and drop it again if the generated name is not what you want.

What it does not do

• No primary keys, indexes or constraints are guessed; add them after the load, when you know which column is actually unique.
• Formulas are converted as the value Excel last calculated, not as the formula.
• Merged cells, formatting and comments are ignored; only the grid is read.
• Very large sheets work, but a million-row INSERT script is a slow way to load a database — for that size, export CSV and use the database's bulk loader.

Nothing leaves this tab

The file is parsed and the SQL is written by JavaScript in your browser. The customer list or price table you are about to load into a database is not uploaded to this site or anywhere else.