OmniSelect FileSQL

← Back to App

Overview

OmniSelect FileSQL is a fully client-side SQL query tool. Your files are never uploaded to any server — all parsing and querying happens entirely in your browser. Load one or more files, write SQL, and get results instantly.

Step 1 — Select Files

Click Choose files (or anywhere in the File Select panel), or drag and drop files onto the page. You can load multiple files at once to join them with SQL.

Supported File Types

Format	Extensions	Notes
CSV / TSV / TXT	.csv, .tsv, .txt	Configurable delimiter, quote, record separator
Excel	.xlsx, .xls, .xlsm, .xlsb, .xltx, .xltm, .xlt	Choose sheet, skip rows, header row supported
Other spreadsheets	.ods, .dbf, .wks, .wk1, .wk3, .wk4	OpenDocument, dBase / FoxPro and Lotus 1-2-3; choose sheet, skip rows, header row supported
JSON	.json	Nested objects/arrays are flattened with dot-path columns
JSON Lines	.jsonl, .ndjson	One JSON record per line, flattened the same way as JSON; blank lines are skipped
XML	.xml	Nested elements flattened to dot-path columns
YAML	.yaml, .yml	Nested structures flattened
Avro	.avro	Parsed via WebAssembly; uncompressed, deflate and snappy files
Parquet	.parquet	Parsed via WebAssembly + Apache Arrow
Arrow / Feather	.arrow, .feather, .ipc, .arrows	Apache Arrow IPC files and streams. Compressed files (the pyarrow default) are refused: save with compression="uncompressed", or as Parquet
SQLite	.sqlite, .sqlite3, .db, .db3, .gpkg	Choose the table or view in the Sheet column. Read by SQLite itself (sql.js); encrypted databases are refused
Microsoft Access	.accdb, .mdb	Each table gets its own row, like the files in a ZIP, shown as shop.accdb › Orders with its own alias. Currency and decimal columns become numbers; password-protected databases are refused
SPSS	.sav, .zsav, .por	Dates become dates. A variable with value labels also gets a <name>_label column (1 → Control); user-missing values are empty
SAS	.sas7bdat, .xpt	Data sets and transport files; date, datetime and time formats become text dates and times
Stata	.dta	Dates (%td, %tc, %tm, %tq …) become dates; value labels get a _label column
MongoDB BSON	.bson	A mongodump file: ObjectIds become text, nested documents are flattened like JSON
MessagePack, CBOR	.msgpack, .mpk, .cbor	One list of records, or records one after another; flattened like JSON
Gzipped	.csv.gz, .json.gz, .jsonl.gz, .xml.gz, .yaml.gz …	Any format above, gzip-compressed. Unpacked by your browser’s built-in decompressor; name the file after the one inside (orders.csv.gz). The 50 MB limit applies to the unpacked size. More on .gz files
ZIP and TAR	.zip, .tar, .tar.gz, .tgz	Every readable file inside is added as its own row and table, shown as archive.zip › orders.csv. Up to 50 MB per file and 200 MB for the whole archive once unpacked; password-protected and ZIP64 archives are refused

File Row Options

Option	Applies To	Description
Alias	All	Single letter (A–Z) used as the table name in SQL. It is taken from the start of the filename, so orders.csv becomes O and you query it with SELECT * FROM O. If two files would claim the same letter, the second gets another free one. Editable — type whichever letter you prefer.
Sheet	Excel	Select which worksheet to load
Column Delimiter	CSV/TSV/TXT	Comma, Tab, Pipe, or custom character
Quote Character	CSV/TSV/TXT	Character used to wrap quoted fields
Record Delimiter	CSV/TSV/TXT	LF, CR, or CRLF line endings
Skip Rows	CSV/TSV/TXT/Excel	Number of rows to skip from the top before reading data
Header Row	CSV/TSV/TXT/Excel	When checked, first (non-skipped) row is used as column names

Step 2 — Write SQL

Use the SQL Editor to write standard SQL queries. The editor supports syntax highlighting. Each loaded file is available as a table using its alias letter.

Basic Examples

-- Select all rows from file A
SELECT * FROM A

-- Filter with a number column
SELECT * FROM A WHERE age > 25

-- Filter with a text column
SELECT * FROM A WHERE city = 'New York'

-- Select specific columns
SELECT first_name, last_name, salary FROM A WHERE salary > 50000

-- Order results
SELECT * FROM A ORDER BY last_name ASC

-- Aggregate
SELECT department, COUNT(*) as headcount, AVG(salary) as avg_salary
FROM A
GROUP BY department

Joining Multiple Files

Load two files — they get aliases A and B (or whatever letters are assigned). Then join them:

-- Inner join
SELECT A.order_id, A.amount, B.customer_name
FROM A
JOIN B ON A.customer_id = B.id

-- Left join
SELECT A.*, B.region
FROM A
LEFT JOIN B ON A.region_code = B.code

Nested JSON/XML/YAML Columns

Nested structures are flattened using dot-path names. For example, a JSON field address.city becomes a column. In the results table, the short alias is shown on top with the full path below it.

-- Query a nested field
SELECT name, address_city FROM A WHERE address_city = 'Austin'

SQL Functions Supported

Queries run on DuckDB, compiled to WebAssembly and running inside this browser tab. Commonly used functions include:

Category	Functions
Aggregate	COUNT, SUM, AVG, MIN, MAX
String	UPPER, LOWER, TRIM, SUBSTRING, CONCAT
Conditional	CASE WHEN … THEN … ELSE … END, COALESCE, IFNULL, NULLIF
Dates	year, month, date_trunc, strftime, and date arithmetic such as order_date + INTERVAL 7 DAY
Window	ROW_NUMBER, RANK, DENSE_RANK, LAG, LEAD, running totals with SUM(…) OVER (ORDER BY …), and QUALIFY to filter on them
Filtering	LIKE, IN, BETWEEN, IS NULL, IS NOT NULL
Clauses	WHERE, GROUP BY, HAVING, ORDER BY, LIMIT, OFFSET, WITH (common table expressions)

Text values go in single quotes: WHERE city = 'Austin'. Double quotes around a word that is not a column name are read as text too, so queries written for earlier versions of this tool keep working — except where only a column name makes sense (before = or IS, in the SELECT list, after ORDER BY): there a misspelt name is reported, not quietly compared as text.

Step 3 — Run the Query

Click the Run Query button. Results appear in the Query Results panel below. The Row limit box beside Export sets how many rows a query brings back: 1,000 unless you change it, and a note above the results says when there were more. Results over 1,000 rows are paginated at 1,000 rows per page — use the Previous / Next buttons to navigate. An export always holds every row of the result, whatever the Row limit. A query that is taking too long can be ended with the Stop button beside Run Query.

Step 4 — Export Results

Click the Export button in the Query Results panel to download results in your preferred format:

Format	Description
CSV	Comma-separated values, opens in Excel or any text editor
JSON	Array of objects, useful for APIs or further processing
Excel	.xlsx file with results on a single sheet
Parquet	Columnar .parquet file — numeric columns stay numeric, dates stay dates, everything else is written as text

Tips & Tricks

• You can reload a file with updated data using the ↻ (reload) button on each file row — no need to remove and re-add it.
• Use the X button to remove a file from the session.
• You can load up to 26 files simultaneously (one per alias letter A–Z).
• Drag and drop multiple files at once onto the File Select panel.
• String comparisons use single quotes: WHERE name = 'Alice'
• For columns that look like reserved SQL words (e.g. key, value, date), the app automatically bracket-quotes them so your query still works.

Privacy

All processing is done entirely in your browser. No data is sent to any server. Files remain on your machine at all times.