Query CSV with SQL

How to use Query CSV with SQL

1. Press Add CSV files or drop one or more CSV/TSV files on the page. You can also use Paste CSV, give it a Table name and press Add table, or try Load sample data for a small customers and orders pair.
2. Check the Tables sidebar for each table’s name, row count and column types. Click a table or column name to insert it into the query.
3. Write SQL in the SQL query box, or pick one from Examples, then press Run or Ctrl/⌘ + Enter. Cancel stops a query that takes too long.
4. Browse the Results table, then use Export CSV or Export JSON to download the result.

How it works

Each file becomes a table in an in-memory SQLite database. It runs as WebAssembly (sql.js) in a background worker, so long queries don’t freeze the page. Table names come from the file name, lowercased with other characters turned into _ (Sales 2024.csv becomes sales_2024). Names that start with a digit get a t_ prefix, and a repeated name gets _2, _3…

The file is read with the same RFC 4180 CSV parser as the CSV ↔ JSON tool. It handles quoted fields with delimiters, "" and line breaks inside them, and ignores a leading byte-order mark. The delimiter is detected from comma, semicolon, tab and pipe; .tsv and .tab files always use tab. The first row is always the header. Blank column names become column_1, column_2…, and repeated names get _2, _3….

Each column’s type is chosen from up to 1,000 non-empty values. If all are whole numbers that fit exactly in a JavaScript number, it’s INTEGER. If all are decimal or exponent numbers, it’s REAL. Anything else makes it TEXT. Empty cells in number columns are stored as NULL. Import every column as TEXT (no type detection) turns this off.

You can use any SQL that SQLite supports, including joins, GROUP BY, WITH, window functions, views, and CREATE TABLE … AS, INSERT, UPDATE and DELETE. Several statements separated by ; run in order, and the last one that returns rows is shown. Errors are SQLite’s own messages, such as no such table: nope.

Limits

• Up to 200 MB of CSV text in total across all tables.
• The results table shows the first 1,000 rows. Exports include up to 500,000 rows of the last result.
• Queries are stopped after 30 seconds. Stopping or cancelling a query reloads the tables from the original files, so changes made by queries are lost. Changing Import every column as TEXT reloads them the same way.
• Rows with more fields than the header lose the extra fields (a notice gives the count). Missing fields become NULL. Up to 3 CSV parse problems, such as an unclosed quote, are listed per file.
• Type detection only looks at the first 1,000 non-empty values. A later value that doesn’t fit the type is stored as text. Numbers use the dot as decimal separator, and dates stay TEXT.
• The database lives only in this tab. Reloading or leaving the page discards it, and there is no way to download the database itself.
• The worker needs WebAssembly. If the browser blocks it, the tool shows an error instead of running.

Privacy

Files are read and queried inside your browser. The SQLite engine and its WebAssembly file load from this site, and the site’s security policy blocks requests to other servers, so your data is never uploaded. Nothing is saved: the tables live in memory and are gone when you close or reload the tab. There is no share link. CSV sent here from another tool with Send to… is passed through this tab’s session storage, removed as soon as it is read, and added as a table.

Frequently asked questions

Which SQL dialect does it use?

SQLite, compiled to WebAssembly (sql.js). Anything SQLite supports works, including JOIN, GROUP BY, WITH RECURSIVE, window functions and the built-in date and string functions. Tables use SQLite’s INTEGER, REAL and TEXT types.

How do I join two CSV files?

Add both files. Each becomes its own table, named after the file. Then write a normal JOIN, for example SELECT * FROM orders o JOIN customers c ON c.id = o.customer_id. With two tables loaded, Examples suggests a join on a likely key, such as a shared column name or customer_id = id.

Why is my number column typed as TEXT?

At least one of the first 1,000 non-empty values isn’t a plain number, for example 1,200, 12% or N/A. Whole numbers too large to store exactly also make a column REAL or TEXT rather than INTEGER. You can still convert values in SQL with CAST(col AS REAL).

Can I change the data with INSERT, UPDATE or DELETE?

Yes, and the Tables sidebar updates after statements that may change tables. The changes only exist in this tab’s in-memory database. They are lost if a query is cancelled or times out, if you change the TEXT import option, or if you reload the page. Export a result to keep it.

Why does the result say it shows only part of the rows?

Only the first 1,000 rows are drawn on the page. The row count is still the full total, and Export CSV or Export JSON includes up to 500,000 rows.