← Back to App

Turn Any File Into a SQL Table

The whole mental model is one sentence: one file becomes one table, and its alias letter is the table name. Once that clicks, everything else is just SQL you already know.

File in, table out

Drop a file on the File Select panel and it is parsed into rows and columns in memory, exactly as if you had loaded it into a database — except there is no database, no CREATE TABLE, and no import step. It is queryable the moment it appears.

This works identically for every supported format:

Format	Extensions	Becomes
Delimited text	.csv, .tsv, .txt	One table, columns from the header row
Excel	.xlsx, .xls, .xlsm	One table per selected sheet
JSON	.json	One table, nesting flattened to dot-paths
XML	.xml	One table, repeated elements become rows
YAML	.yaml, .yml	One table, nesting flattened
Avro	.avro	One table, read via WebAssembly
Parquet	.parquet	One table, read via WebAssembly

How the table gets its name

Each file is given a single-letter alias, shown in the Alias column. The letter comes from the start of the filename, so orders.csv becomes O and customers.csv becomes C. When two files would claim the same letter, the second is given another free one.

Seeing the schema

There is no separate schema browser; one query does the job:

SELECT * FROM O LIMIT 1

The result headers give you the full column list. For nested formats, each header shows the short name on top with the full dot-path underneath, so you can see where a flattened field came from.

How columns are named

• Spaces become underscores. A column headed first name is queried as first_name.
• Nesting becomes dot-paths. A JSON field at address.city gets its own column, with a short alias you can also use.
• Reserved words are bracket-quoted for you when they appear as column names, so a column called key or date still works.

Up to 26 tables at once

Load as many files as you have letters for — A to Z. They can be different formats, and they are all queryable in the same statement.

SELECT O.order_id, C.name, P.category
        FROM O
        JOIN C ON O.customer_id = C.customer_id
        JOIN P ON O.product_id = P.product_id

That joins a CSV, an Excel sheet and a Parquet file together, and the query does not care which was which.

Stacking files instead of joining them

Monthly extracts with identical columns are a common case. Use UNION ALL to treat them as one table:

SELECT * FROM J
        UNION ALL
        SELECT * FROM F
        UNION ALL
        SELECT * FROM M

Use UNION instead of UNION ALL if you want duplicate rows removed.

Limits worth knowing

Limit	Value
Files at once	26 (one per alias letter)
File size	50 MB each
Rows per file	1,000,000 — a safety backstop, and truncation is always reported
Results paging	1,000 rows per page

Past those, you want a real database, and that is the honest answer rather than a grudging one.

Related Guides

• Analyse a CSV file in three steps
• Join two CSV files with SQL
• Run SQL without installing a database
• All guides · Full How To Use guide