SQL Playground: Query Files in Your Browser

After reading this you will know how to load CSV, Excel and Parquet files as SQL tables, write joins and aggregations that run entirely in your browser, and read the results without being tricked by type inference or memory limits.

What it is, with one query

The SQL Playground turns files into tables and runs real SQL against them. Drop a file named sales.csv and you get a table called sales. From there you write standard SQL. The engine is DuckDB compiled to WebAssembly, so everything happens inside the browser tab. No data is uploaded anywhere.

Here is the hook. Suppose sales.csv has 1,000 rows with columns region, product and amount. This query answers "how much did each region sell?" in one pass:

SELECT region, SUM(amount) AS total FROM sales GROUP BY region ORDER BY total DESC

That is not a screenshot of a database server. It runs on the file you just dropped, using about 35 MB of engine code that downloads once and then stays cached.

When to use it, and when not

Use the Playground when you have tabular files and a question that SQL expresses cleanly: filtering, grouping, joining two files on a shared key, ranking within groups, or computing running totals. SQL is compact for these. A three-table join with a filter and a sort is four lines here and a fiddly afternoon in a spreadsheet.

Do not reach for it when a simpler tool already fits. If you only want to look at a CSV, sort it and delete a column, the CSV Viewer & Editor is faster. If you want a quick cross-tab without writing SQL, the Pivot Table Maker does grouping through menus. If you just need to see inside a Parquet file, the Parquet & Feather Preview shows the schema and a sample without a query.

How files become tables

Each file is registered under a table name derived from its filename, with the extension stripped. orders_2024.csv becomes orders_2024. A leading digit or a space forces you to quote the name: write SELECT * FROM "2024 orders".

Column types are inferred, not declared. DuckDB reads a sample of rows and guesses. This is where surprises begin. A column of postal codes like 01234 may be read as the integer 1234, dropping the leading zero. A column mixing 12 and N/A becomes text, so SUM on it will error or return zero.

You can override inference by calling the reader function directly instead of relying on the auto-registered table:

SELECT * FROM read_csv_auto('sales.csv', types={'zip': 'VARCHAR'})

Here read_csv_auto is the CSV reader, 'sales.csv' is the file, and types pins the zip column to text so leading zeros survive.

Grouping and aggregation, the core idea

Most useful queries collapse many rows into few. GROUP BY region takes 1,000 sales rows and produces one row per distinct region. The aggregate functions decide what each group becomes:

Here g is one group, i ranges over the rows in that group, x_i is the value of the aggregated column, and n_g is the number of rows in the group. Count ignores the values and just tallies rows. Sum adds them. Average divides sum by count.

The mental model: SQL sorts rows into buckets by the GROUP BY key, then computes one number per bucket. A row with a NULL amount is skipped by SUM and AVG but still counted by COUNT(*). That mismatch explains many "the numbers do not add up" moments.

Window functions, the part spreadsheets cannot do easily

A window function computes across a set of rows without collapsing them. You keep all four sales rows but attach a per-region rank or running total to each. The clause OVER (PARTITION BY region ORDER BY amount DESC) defines the window.

RANK() OVER (PARTITION BY region ORDER BY amount DESC)

PARTITION BY region restarts the ranking for each region. ORDER BY amount DESC ranks largest first. In North, Widget (120) gets rank 1 and Gadget (80) rank 2. In South, Widget (200) gets rank 1 and Gadget (50) rank 2. All four rows survive; each just gains a rank column.

To keep only the top row per region, wrap it in QUALIFY rank = 1. That is DuckDB shorthand that filters on a window result without a subquery.

Joining across files

The Playground shines when two files share a key. Drop sales.csv and regions.csv where regions maps each region to a manager. Join them:

SELECT s.region, r.manager, SUM(s.amount) FROM sales s JOIN regions r ON s.region = r.region GROUP BY 1, 2

The join key is region. An inner JOIN keeps only rows where a match exists in both files. If regions is missing the South entry, South sales vanish silently from the result. Use LEFT JOIN to keep every sales row and get NULL for the missing manager, which makes the gap visible instead of hidden.

Common mistakes

Selecting a non-grouped column

Writing SELECT region, product, SUM(amount) ... GROUP BY region fails or returns an arbitrary product, because SQL cannot pick one product for a group of many. Either group by it too or aggregate it.

Trusting inferred types

A version column like 1.10 read as a number becomes 1.1, and IDs with leading zeros lose them. Pin the type with the reader function when the column is really text.

Counting rows instead of values

COUNT(*) counts all rows including NULLs; COUNT(amount) counts only non-NULL amounts. If 50 of 1,000 rows have a NULL amount, these differ by 50.

Filtering after grouping by mistake

WHERE filters rows before grouping; HAVING filters groups after. To keep regions whose total exceeds 200, use HAVING SUM(amount) > 200, not WHERE.

Related tools

Once you have an answer, other tools finish the job. Export the result and chart it in the Data Visualizer, or turn a CSV into JSON with the CSV ↔ JSON Converter. If your source is an Excel workbook, the Excel → CSV Converter gives you one clean CSV per sheet to drop in here. To view JSON as a table before querying, use the JSON Table Viewer. To mask sensitive columns before sharing a result, use the Data Anonymizer. To format a small result for a document, paste it into the Markdown Table Generator.

Related tools

Auto Analyst: Instant Statistical Report

Drop a CSV or Excel file and get a statistician’s report: data problems, distributions, every real relationship, key drivers and Simpson’s paradox warnings, in plain English.

CSV Viewer & Editor

View, edit, sort, filter and extract columns from CSV files.

CSV ↔ JSON Converter

Convert CSV to JSON and JSON back to CSV.

Data Visualizer

29 chart types — heatmap, treemap, dumbbell, slope, candlestick — with rich styling and annotations.

Excel → CSV Converter

Convert XLSX/XLS sheets to clean CSV, per sheet.

JSON Table Viewer

Paste JSON, get a sortable, searchable table.