Which SQL dialect is this?

DuckDB, compiled to WebAssembly and running inside the page. Window functions, common table expressions, PIVOT, and the usual analytical functions all work.

How are my files named as tables?

Each file becomes a table named after it, lowercased and simplified. The sidebar lists the exact names and columns so you can copy them.

Can I write to my files or reach the network from a query?

No. Queries are checked before they run and only reading from your own loaded files is permitted. Anything that would write a file or open a connection is refused.

How large a file can I query?

Filtering and counting work on files far larger than memory because they stream. Sorting or grouping a very large file needs to hold more at once, and you are warned before that becomes a problem.

Can I stop a query that is taking too long?

Yes, at any point, and the page stays usable afterwards. Nothing is left in a broken state.

Is my file uploaded anywhere?

No. When you run SQL on a CSV file here, the engine runs inside this page and the browser is told to block outbound connections, so there is no upload to intercept. Any request the Network tab of your developer tools shows while you run SQL on a CSV file is to this site — the engine’s own files — and none carries your data.

Run SQL on a CSV file

Drop files, they become tables, write SQL against them. A real analytical database running in the page, with nothing sent anywhere.

Query with SQLRunning on your device

Sources

Your files

or drag one here · it stays on your device

Preview rowsor press ⌘/Ctrl + Enter

Result

Write a query and run it — the rows come back here, with how long they took.

This browser build cannot write PARQUET files yet. CSV and Excel are available now, and hold the same data.

Have a look at the result before downloading it.

How this works

Add your files. Drop one or more spreadsheets onto the page.
Write a query. Reference each file by the table name shown in the sidebar.
Run it. Results appear below, with the time taken and the number of rows.
Download. Export the result in any supported format.

What it cannot do

Only reading from your own loaded files is allowed; statements that would write files, install extensions, or open network connections are refused before they run.
Sorting or grouping needs to hold much more in memory than filtering does, so the practical file size is smaller for those.
Previews are limited to a thousand rows by default so a mistyped query cannot lock up the page; exports run without that limit.

Questions

Which SQL dialect is this?: DuckDB, compiled to WebAssembly and running inside the page. Window functions, common table expressions, PIVOT, and the usual analytical functions all work.
How are my files named as tables?: Each file becomes a table named after it, lowercased and simplified. The sidebar lists the exact names and columns so you can copy them.
Can I write to my files or reach the network from a query?: No. Queries are checked before they run and only reading from your own loaded files is permitted. Anything that would write a file or open a connection is refused.
How large a file can I query?: Filtering and counting work on files far larger than memory because they stream. Sorting or grouping a very large file needs to hold more at once, and you are warned before that becomes a problem.
Can I stop a query that is taking too long?: Yes, at any point, and the page stays usable afterwards. Nothing is left in a broken state.
Is my file uploaded anywhere?: No. When you run SQL on a CSV file here, the engine runs inside this page and the browser is told to block outbound connections, so there is no upload to intercept. Any request the Network tab of your developer tools shows while you run SQL on a CSV file is to this site — the engine’s own files — and none carries your data.

Your file is processed in this page and never uploaded — the privacy page shows how to confirm that yourself.

Common problems this solves

Excel stops at 1,048,576 rows — here is how to open the rest — Excel silently drops every row past 1,048,576, and saving makes it permanent. Open, check and query the whole file in your browser — nothing uploaded.
Excel removed the leading zeros — get them back and keep them — Excel turns 00412 into 412 on open; saving makes it permanent. Check whether your file still has its zeros, keep them, or put them back — in your browser.
VLOOKUP between two files, without the formula — Bring columns from one spreadsheet into another by a shared ID, the way VLOOKUP does — every row at once, unmatched rows listed, nothing uploaded.
Get a CREATE TABLE statement from a CSV file — Turn a CSV into a CREATE TABLE statement and INSERTs for Postgres, MySQL, SQLite or SQL Server, column names from the header. In your browser, no upload.
Excel dates showing as numbers like 45292 — what they mean and how to convert them — A date column full of five-digit numbers is Excel counting days since 30 Dec 1899. The arithmetic, the traps, and a one-click conversion to real dates.
The file is too large to open — in Excel, Notepad, or anything else — A CSV Excel will not open, Notepad chokes on and Sheets refuses is usually only long. Read it whole, count it, keep the rows you need — on your machine.
Google Sheets says the file is too large to import — the ten million cell limit — Sheets refuses a file past ten million cells, and a wide file gets there before a million rows. Count what you have and drop the columns you do not need.
The Shopify orders CSV: one row per item, blank order fields, and how to make it one row per order — Shopify writes one row per line item and fills the order columns only on the first. Fill them down and keep phones and totals as written, in your browser.
Export Stripe data to CSV: which export to take, what it holds, and matching payouts to the bank — Which Stripe export to take, what the balance and payout files hold, how to match payouts to the bank, and loading one into a database or BI tool.
The PayPal activity CSV: gross, fee and net, currency conversions, and the rows that are not sales — The PayPal download mixes sales, fees, refunds, transfers and currency conversions. Which rows to keep, why fees are negative, and how to filter it here.
The QuickBooks report export: title lines above the heading, subtotal rows, and getting a plain table — A QuickBooks report export carries the company name above the heading and subtotal rows in the data. Strip them here and get a table any tool reads.
The Salesforce report CSV: the copyright footer, totals rows, and 15- versus 18-character IDs — A Salesforce report export ends in a copyright footer and may carry a totals row; IDs come in two lengths. Drop the footer and match IDs, in your browser.
The Google Analytics CSV: comment lines above the heading, a second table underneath, and numbers with commas — A Google Analytics download starts with # lines, ends with a second table, and writes numbers with commas and dates as 20240131. Make it one clean table.
The Etsy CSV downloads: orders in one file, items in another, and the statement that explains the fees — Etsy splits a sale across three downloads: orders, order items, and the monthly statement. Which holds what, and how to join them by order id here.
The Xero report export: organisation and date-range lines above the headings, and totals in the rows — A Xero report export carries the organisation name and date range above the headings and total rows in the data. Strip them here for a plain table.
The bank statement CSV: day-first dates, debit and credit in two columns, newest first — Bank downloads write dates day first, split money into debit and credit columns, and list the newest line first. One query here makes a clean ledger.
A French CSV: semicolons, decimal commas, and a space you cannot see — French exports use semicolons and group thousands with a space that is often a no-break space. What converts automatically here, and what does not.
A Dutch CSV: semicolons, decimal commas, and dates written with dashes — Dutch exports use semicolons and a decimal comma, and write dates 31-12-2026 — dashes that look like an ISO date in the wrong order. How to read one.
A Brazilian CSV: semicolons, decimal commas, and amounts with R$ attached — Brazilian exports use semicolons and a decimal comma, and often carry the currency inside the amount column, which stops any of it converting. What to do.
A Swiss CSV: semicolons, a decimal point, and apostrophes inside the numbers — Switzerland writes 1'234.56: apostrophe thousands, decimal point, no comma. Why the usual European conversion does nothing here, and what to do instead.
An American CSV abroad: the dates are the thing that breaks — An American CSV opens cleanly almost everywhere, which is the problem: its month-first dates are read as day-first by half the world, silently.
A British CSV: it looks American, and the dates are the other way round — A British CSV uses commas and a decimal point like an American one, but writes dates day first. The two look alike until a date passes the twelfth.
Upload a bank CSV to QuickBooks Online: three columns and one date format — QuickBooks Online takes bank CSVs as Date, Description and Amount, or with Credit and Debit apart. Turn a bank download into that shape here, no upload.

Sources

Result

Download

Sources

Result

Download